diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,56419 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.18549251221981258,
+  "eval_steps": 500,
+  "global_step": 47000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-06,
+      "loss": 716.2841,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 699.611,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5e-05,
+      "loss": 543.0736,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-05,
+      "loss": 345.0442,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5e-05,
+      "loss": 190.8229,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-05,
+      "loss": 129.2581,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5e-05,
+      "loss": 68.5645,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-05,
+      "loss": 28.87,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5e-05,
+      "loss": 14.7074,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-05,
+      "loss": 9.8457,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.500000000000001e-05,
+      "loss": 8.2193,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-05,
+      "loss": 8.074,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 8.8025,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7e-05,
+      "loss": 26.2587,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.500000000000001e-05,
+      "loss": 50.4178,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-05,
+      "loss": 16.1342,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.5e-05,
+      "loss": 8.8308,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9e-05,
+      "loss": 7.419,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.5e-05,
+      "loss": 7.2521,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001,
+      "loss": 7.0084,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99999999038429e-05,
+      "loss": 8.2626,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999961537156e-05,
+      "loss": 7.0366,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999913458602e-05,
+      "loss": 7.9765,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999846148626e-05,
+      "loss": 6.7673,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999759607231e-05,
+      "loss": 6.4718,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999653834413e-05,
+      "loss": 6.4201,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999528830176e-05,
+      "loss": 6.3429,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999384594516e-05,
+      "loss": 6.106,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999221127438e-05,
+      "loss": 6.1027,
+      "step": 145
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999038428942e-05,
+      "loss": 5.8923,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999998836499028e-05,
+      "loss": 5.8181,
+      "step": 155
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999998615337697e-05,
+      "loss": 5.7303,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999998374944947e-05,
+      "loss": 5.697,
+      "step": 165
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999998115320785e-05,
+      "loss": 5.5815,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999997836465207e-05,
+      "loss": 5.3889,
+      "step": 175
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999997538378215e-05,
+      "loss": 5.3269,
+      "step": 180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999997221059811e-05,
+      "loss": 5.307,
+      "step": 185
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999996884509996e-05,
+      "loss": 5.1735,
+      "step": 190
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999996528728773e-05,
+      "loss": 5.1296,
+      "step": 195
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999996153716138e-05,
+      "loss": 5.1319,
+      "step": 200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999995759472098e-05,
+      "loss": 5.102,
+      "step": 205
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999995345996653e-05,
+      "loss": 4.8224,
+      "step": 210
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999994913289802e-05,
+      "loss": 4.7975,
+      "step": 215
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999994461351552e-05,
+      "loss": 4.8552,
+      "step": 220
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999993990181899e-05,
+      "loss": 4.5189,
+      "step": 225
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999993499780848e-05,
+      "loss": 4.8096,
+      "step": 230
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999992990148401e-05,
+      "loss": 4.8069,
+      "step": 235
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99999246128456e-05,
+      "loss": 4.8597,
+      "step": 240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999991913189324e-05,
+      "loss": 4.5138,
+      "step": 245
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999991345862698e-05,
+      "loss": 4.5089,
+      "step": 250
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999990759304684e-05,
+      "loss": 4.5595,
+      "step": 255
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999990153515283e-05,
+      "loss": 4.4312,
+      "step": 260
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999989528494499e-05,
+      "loss": 4.4037,
+      "step": 265
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999988884242333e-05,
+      "loss": 4.484,
+      "step": 270
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999988220758788e-05,
+      "loss": 4.4117,
+      "step": 275
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999987538043867e-05,
+      "loss": 4.506,
+      "step": 280
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999986836097573e-05,
+      "loss": 4.3508,
+      "step": 285
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999986114919906e-05,
+      "loss": 4.411,
+      "step": 290
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999985374510871e-05,
+      "loss": 4.2754,
+      "step": 295
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99998461487047e-05,
+      "loss": 4.1845,
+      "step": 300
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999983835998708e-05,
+      "loss": 4.2648,
+      "step": 305
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999983037895585e-05,
+      "loss": 4.312,
+      "step": 310
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999982220561107e-05,
+      "loss": 4.3153,
+      "step": 315
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999981383995273e-05,
+      "loss": 4.3693,
+      "step": 320
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999980528198091e-05,
+      "loss": 4.2449,
+      "step": 325
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999979653169563e-05,
+      "loss": 4.4409,
+      "step": 330
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999978758909689e-05,
+      "loss": 4.2739,
+      "step": 335
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999977845418476e-05,
+      "loss": 4.0231,
+      "step": 340
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999976912695928e-05,
+      "loss": 4.0508,
+      "step": 345
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999975960742045e-05,
+      "loss": 4.0531,
+      "step": 350
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999974989556833e-05,
+      "loss": 4.1512,
+      "step": 355
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999973999140295e-05,
+      "loss": 3.9009,
+      "step": 360
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999972989492437e-05,
+      "loss": 4.3107,
+      "step": 365
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999971960613259e-05,
+      "loss": 3.9073,
+      "step": 370
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999970912502769e-05,
+      "loss": 4.1516,
+      "step": 375
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999969845160968e-05,
+      "loss": 4.1611,
+      "step": 380
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99996875858786e-05,
+      "loss": 3.9599,
+      "step": 385
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999967652783453e-05,
+      "loss": 3.9219,
+      "step": 390
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999966527747748e-05,
+      "loss": 3.9158,
+      "step": 395
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999965383480749e-05,
+      "loss": 3.9322,
+      "step": 400
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999964219982461e-05,
+      "loss": 3.8471,
+      "step": 405
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99996303725289e-05,
+      "loss": 3.9856,
+      "step": 410
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99996183529204e-05,
+      "loss": 3.7377,
+      "step": 415
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999960614099913e-05,
+      "loss": 3.8324,
+      "step": 420
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999959373676517e-05,
+      "loss": 3.9418,
+      "step": 425
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999958114021855e-05,
+      "loss": 3.8771,
+      "step": 430
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999956835135933e-05,
+      "loss": 3.7497,
+      "step": 435
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999955537018754e-05,
+      "loss": 3.8777,
+      "step": 440
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999954219670327e-05,
+      "loss": 3.952,
+      "step": 445
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999952883090652e-05,
+      "loss": 3.7888,
+      "step": 450
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999951527279736e-05,
+      "loss": 3.8492,
+      "step": 455
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999950152237586e-05,
+      "loss": 3.8447,
+      "step": 460
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999948757964206e-05,
+      "loss": 3.7508,
+      "step": 465
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.9999473444596e-05,
+      "loss": 3.7305,
+      "step": 470
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999945911723776e-05,
+      "loss": 3.7131,
+      "step": 475
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999944459756738e-05,
+      "loss": 3.7908,
+      "step": 480
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999942988558492e-05,
+      "loss": 3.8519,
+      "step": 485
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999941498129042e-05,
+      "loss": 3.6184,
+      "step": 490
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999939988468398e-05,
+      "loss": 3.76,
+      "step": 495
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999938459576561e-05,
+      "loss": 3.7592,
+      "step": 500
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999936911453541e-05,
+      "loss": 3.6597,
+      "step": 505
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99993534409934e-05,
+      "loss": 3.8063,
+      "step": 510
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999933757513967e-05,
+      "loss": 3.7862,
+      "step": 515
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999932151697425e-05,
+      "loss": 3.6837,
+      "step": 520
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999930526649724e-05,
+      "loss": 3.6057,
+      "step": 525
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999928882370867e-05,
+      "loss": 3.6215,
+      "step": 530
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999927218860862e-05,
+      "loss": 3.7247,
+      "step": 535
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999925536119717e-05,
+      "loss": 3.7103,
+      "step": 540
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999923834147435e-05,
+      "loss": 3.8163,
+      "step": 545
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999922112944024e-05,
+      "loss": 3.6233,
+      "step": 550
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999920372509493e-05,
+      "loss": 3.543,
+      "step": 555
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999918612843844e-05,
+      "loss": 3.7653,
+      "step": 560
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999916833947088e-05,
+      "loss": 3.6019,
+      "step": 565
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999915035819229e-05,
+      "loss": 3.5738,
+      "step": 570
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999913218460276e-05,
+      "loss": 3.5387,
+      "step": 575
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999911381870235e-05,
+      "loss": 3.5544,
+      "step": 580
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999909526049112e-05,
+      "loss": 3.7664,
+      "step": 585
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999907650996915e-05,
+      "loss": 3.5455,
+      "step": 590
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999905756713653e-05,
+      "loss": 3.695,
+      "step": 595
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999903843199331e-05,
+      "loss": 3.7061,
+      "step": 600
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999901910453957e-05,
+      "loss": 3.6527,
+      "step": 605
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999899958477539e-05,
+      "loss": 3.6827,
+      "step": 610
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999897987270083e-05,
+      "loss": 3.5501,
+      "step": 615
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999895996831598e-05,
+      "loss": 3.5967,
+      "step": 620
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99989398716209e-05,
+      "loss": 3.7557,
+      "step": 625
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999891958261572e-05,
+      "loss": 3.7793,
+      "step": 630
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999889910130044e-05,
+      "loss": 3.727,
+      "step": 635
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999887842767519e-05,
+      "loss": 3.6344,
+      "step": 640
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999885756174003e-05,
+      "loss": 3.664,
+      "step": 645
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999883650349505e-05,
+      "loss": 3.6003,
+      "step": 650
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999881525294032e-05,
+      "loss": 3.6123,
+      "step": 655
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999879381007595e-05,
+      "loss": 3.577,
+      "step": 660
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.9998772174902e-05,
+      "loss": 3.5509,
+      "step": 665
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999875034741852e-05,
+      "loss": 3.5339,
+      "step": 670
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999872832762567e-05,
+      "loss": 3.6058,
+      "step": 675
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999870611552346e-05,
+      "loss": 3.5768,
+      "step": 680
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999868371111203e-05,
+      "loss": 3.5923,
+      "step": 685
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999866111439145e-05,
+      "loss": 3.6136,
+      "step": 690
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999863832536178e-05,
+      "loss": 3.5217,
+      "step": 695
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999861534402315e-05,
+      "loss": 3.5816,
+      "step": 700
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999859217037562e-05,
+      "loss": 3.5321,
+      "step": 705
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99985688044193e-05,
+      "loss": 3.6159,
+      "step": 710
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999854524615424e-05,
+      "loss": 3.4857,
+      "step": 715
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999852149558058e-05,
+      "loss": 3.4135,
+      "step": 720
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999849755269837e-05,
+      "loss": 3.6419,
+      "step": 725
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999847341750774e-05,
+      "loss": 3.4044,
+      "step": 730
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999844909000875e-05,
+      "loss": 3.5022,
+      "step": 735
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999842457020152e-05,
+      "loss": 3.495,
+      "step": 740
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999839985808612e-05,
+      "loss": 3.5878,
+      "step": 745
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999837495366267e-05,
+      "loss": 3.4944,
+      "step": 750
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999834985693124e-05,
+      "loss": 3.5807,
+      "step": 755
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999832456789196e-05,
+      "loss": 3.5956,
+      "step": 760
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999829908654488e-05,
+      "loss": 3.5206,
+      "step": 765
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999827341289013e-05,
+      "loss": 3.5688,
+      "step": 770
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999824754692781e-05,
+      "loss": 3.5244,
+      "step": 775
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999822148865802e-05,
+      "loss": 3.3485,
+      "step": 780
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999819523808084e-05,
+      "loss": 3.4923,
+      "step": 785
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99981687951964e-05,
+      "loss": 3.4259,
+      "step": 790
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999814216000476e-05,
+      "loss": 3.5631,
+      "step": 795
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999811533250606e-05,
+      "loss": 3.6357,
+      "step": 800
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99980883127004e-05,
+      "loss": 3.5568,
+      "step": 805
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999806110058787e-05,
+      "loss": 3.4238,
+      "step": 810
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999803369616858e-05,
+      "loss": 3.4805,
+      "step": 815
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999800609944263e-05,
+      "loss": 3.3312,
+      "step": 820
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999797831041014e-05,
+      "loss": 3.4652,
+      "step": 825
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99979503290712e-05,
+      "loss": 3.4987,
+      "step": 830
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999792215542592e-05,
+      "loss": 3.6739,
+      "step": 835
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999789378947444e-05,
+      "loss": 3.4512,
+      "step": 840
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999786523121682e-05,
+      "loss": 3.5399,
+      "step": 845
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99978364806532e-05,
+      "loss": 3.4159,
+      "step": 850
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999780753778368e-05,
+      "loss": 3.3952,
+      "step": 855
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999777840260837e-05,
+      "loss": 3.3957,
+      "step": 860
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99977490751274e-05,
+      "loss": 3.2901,
+      "step": 865
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999771955534087e-05,
+      "loss": 3.3914,
+      "step": 870
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99976898432489e-05,
+      "loss": 3.4547,
+      "step": 875
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999765993885158e-05,
+      "loss": 3.4125,
+      "step": 880
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999762984214907e-05,
+      "loss": 3.2652,
+      "step": 885
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999759955314143e-05,
+      "loss": 3.257,
+      "step": 890
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999756907182882e-05,
+      "loss": 3.4659,
+      "step": 895
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999753839821133e-05,
+      "loss": 3.2861,
+      "step": 900
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999750753228909e-05,
+      "loss": 3.4078,
+      "step": 905
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999747647406223e-05,
+      "loss": 3.3789,
+      "step": 910
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999744522353085e-05,
+      "loss": 3.3178,
+      "step": 915
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999741378069509e-05,
+      "loss": 3.3318,
+      "step": 920
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999738214555505e-05,
+      "loss": 3.4214,
+      "step": 925
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999735031811088e-05,
+      "loss": 3.4165,
+      "step": 930
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999731829836265e-05,
+      "loss": 3.2615,
+      "step": 935
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999728608631054e-05,
+      "loss": 3.3593,
+      "step": 940
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999725368195465e-05,
+      "loss": 3.4622,
+      "step": 945
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99972210852951e-05,
+      "loss": 3.3699,
+      "step": 950
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999718829633203e-05,
+      "loss": 3.4333,
+      "step": 955
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999715531506555e-05,
+      "loss": 3.2514,
+      "step": 960
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99971221414958e-05,
+      "loss": 3.4681,
+      "step": 965
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999708877562288e-05,
+      "loss": 3.5241,
+      "step": 970
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999705521744696e-05,
+      "loss": 3.4144,
+      "step": 975
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999702146696815e-05,
+      "loss": 3.4672,
+      "step": 980
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999698752418655e-05,
+      "loss": 3.2687,
+      "step": 985
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999695338910235e-05,
+      "loss": 3.3031,
+      "step": 990
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999691906171564e-05,
+      "loss": 3.4671,
+      "step": 995
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999688454202656e-05,
+      "loss": 3.3356,
+      "step": 1000
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999684983003525e-05,
+      "loss": 3.3278,
+      "step": 1005
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999681492574185e-05,
+      "loss": 3.2737,
+      "step": 1010
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999677982914646e-05,
+      "loss": 3.3198,
+      "step": 1015
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999674454024926e-05,
+      "loss": 3.4587,
+      "step": 1020
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999670905905035e-05,
+      "loss": 3.4254,
+      "step": 1025
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999667338554987e-05,
+      "loss": 3.2285,
+      "step": 1030
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999663751974798e-05,
+      "loss": 3.2885,
+      "step": 1035
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99966014616448e-05,
+      "loss": 3.2673,
+      "step": 1040
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999656521124049e-05,
+      "loss": 3.338,
+      "step": 1045
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999652876853515e-05,
+      "loss": 3.3025,
+      "step": 1050
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999649213352898e-05,
+      "loss": 3.3108,
+      "step": 1055
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999645530622204e-05,
+      "loss": 3.435,
+      "step": 1060
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999641828661454e-05,
+      "loss": 3.4411,
+      "step": 1065
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99963810747066e-05,
+      "loss": 3.4376,
+      "step": 1070
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999634367049836e-05,
+      "loss": 3.3785,
+      "step": 1075
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999630607398995e-05,
+      "loss": 3.2375,
+      "step": 1080
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999626828518156e-05,
+      "loss": 3.345,
+      "step": 1085
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999623030407328e-05,
+      "loss": 3.4732,
+      "step": 1090
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99961921306653e-05,
+      "loss": 3.4064,
+      "step": 1095
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999615376495774e-05,
+      "loss": 3.2769,
+      "step": 1100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999611520695075e-05,
+      "loss": 3.2561,
+      "step": 1105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999607645664448e-05,
+      "loss": 3.3256,
+      "step": 1110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99960375140391e-05,
+      "loss": 3.325,
+      "step": 1115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999599837913474e-05,
+      "loss": 3.3318,
+      "step": 1120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999595905193156e-05,
+      "loss": 3.4583,
+      "step": 1125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999591953242969e-05,
+      "loss": 3.1966,
+      "step": 1130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99958798206293e-05,
+      "loss": 3.2154,
+      "step": 1135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999583991653053e-05,
+      "loss": 3.3572,
+      "step": 1140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999579982013357e-05,
+      "loss": 3.2882,
+      "step": 1145
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999575953143851e-05,
+      "loss": 3.3011,
+      "step": 1150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999571905044558e-05,
+      "loss": 3.3243,
+      "step": 1155
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999567837715487e-05,
+      "loss": 3.2902,
+      "step": 1160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999563751156657e-05,
+      "loss": 3.2472,
+      "step": 1165
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999559645368085e-05,
+      "loss": 3.3439,
+      "step": 1170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999555520349783e-05,
+      "loss": 3.3707,
+      "step": 1175
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99955137610177e-05,
+      "loss": 3.2101,
+      "step": 1180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99954721262406e-05,
+      "loss": 3.3127,
+      "step": 1185
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999543029916672e-05,
+      "loss": 3.2641,
+      "step": 1190
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999538827979618e-05,
+      "loss": 3.2214,
+      "step": 1195
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999534606812914e-05,
+      "loss": 3.2951,
+      "step": 1200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99953036641658e-05,
+      "loss": 3.4081,
+      "step": 1205
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999526106790631e-05,
+      "loss": 3.1939,
+      "step": 1210
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999521827935085e-05,
+      "loss": 3.3903,
+      "step": 1215
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999517529849953e-05,
+      "loss": 3.2708,
+      "step": 1220
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999513212535256e-05,
+      "loss": 3.1319,
+      "step": 1225
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99950887599101e-05,
+      "loss": 3.3906,
+      "step": 1230
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999504520217231e-05,
+      "loss": 3.2221,
+      "step": 1235
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999500145213936e-05,
+      "loss": 3.2366,
+      "step": 1240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999495750981142e-05,
+      "loss": 3.2526,
+      "step": 1245
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999491337518865e-05,
+      "loss": 3.2328,
+      "step": 1250
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999486904827124e-05,
+      "loss": 3.1628,
+      "step": 1255
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999482452905933e-05,
+      "loss": 3.2805,
+      "step": 1260
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999477981755313e-05,
+      "loss": 3.3916,
+      "step": 1265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999473491375277e-05,
+      "loss": 3.1894,
+      "step": 1270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999468981765846e-05,
+      "loss": 3.2753,
+      "step": 1275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999464452927035e-05,
+      "loss": 3.1456,
+      "step": 1280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999459904858861e-05,
+      "loss": 3.2511,
+      "step": 1285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999455337561344e-05,
+      "loss": 3.2846,
+      "step": 1290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9994507510345e-05,
+      "loss": 3.0858,
+      "step": 1295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999446145278346e-05,
+      "loss": 3.3178,
+      "step": 1300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999441520292902e-05,
+      "loss": 3.2925,
+      "step": 1305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999436876078183e-05,
+      "loss": 3.3261,
+      "step": 1310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999432212634209e-05,
+      "loss": 3.1822,
+      "step": 1315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999427529960996e-05,
+      "loss": 3.1962,
+      "step": 1320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999422828058564e-05,
+      "loss": 3.2941,
+      "step": 1325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99941810692693e-05,
+      "loss": 3.3664,
+      "step": 1330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999413366566113e-05,
+      "loss": 3.3429,
+      "step": 1335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999408606976128e-05,
+      "loss": 3.212,
+      "step": 1340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999403828156999e-05,
+      "loss": 3.1874,
+      "step": 1345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99939903010874e-05,
+      "loss": 3.2996,
+      "step": 1350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99939421283137e-05,
+      "loss": 3.1601,
+      "step": 1355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99938937632491e-05,
+      "loss": 3.2744,
+      "step": 1360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999384520589376e-05,
+      "loss": 3.189,
+      "step": 1365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999379645624787e-05,
+      "loss": 3.2307,
+      "step": 1370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999374751431163e-05,
+      "loss": 3.2247,
+      "step": 1375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999369838008522e-05,
+      "loss": 3.1591,
+      "step": 1380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999364905356883e-05,
+      "loss": 3.1941,
+      "step": 1385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999359953476265e-05,
+      "loss": 3.1062,
+      "step": 1390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999354982366686e-05,
+      "loss": 3.1071,
+      "step": 1395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999349992028167e-05,
+      "loss": 3.3272,
+      "step": 1400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999344982460728e-05,
+      "loss": 3.2959,
+      "step": 1405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999339953664384e-05,
+      "loss": 3.2,
+      "step": 1410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999334905639157e-05,
+      "loss": 3.2104,
+      "step": 1415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999329838385067e-05,
+      "loss": 3.2455,
+      "step": 1420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999324751902135e-05,
+      "loss": 3.1855,
+      "step": 1425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999319646190376e-05,
+      "loss": 3.3188,
+      "step": 1430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999314521249813e-05,
+      "loss": 3.0674,
+      "step": 1435
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999309377080464e-05,
+      "loss": 3.2032,
+      "step": 1440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99930421368235e-05,
+      "loss": 3.0309,
+      "step": 1445
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999299031055492e-05,
+      "loss": 3.1836,
+      "step": 1450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999293829199904e-05,
+      "loss": 3.252,
+      "step": 1455
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999288608115615e-05,
+      "loss": 3.2843,
+      "step": 1460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99928336780264e-05,
+      "loss": 3.1226,
+      "step": 1465
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999278108260997e-05,
+      "loss": 3.2395,
+      "step": 1470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999272829490709e-05,
+      "loss": 3.2298,
+      "step": 1475
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999267531491799e-05,
+      "loss": 3.2292,
+      "step": 1480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999262214264282e-05,
+      "loss": 3.334,
+      "step": 1485
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99925687780818e-05,
+      "loss": 3.395,
+      "step": 1490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999251522123517e-05,
+      "loss": 3.1668,
+      "step": 1495
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99924614721031e-05,
+      "loss": 3.2879,
+      "step": 1500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999240753068582e-05,
+      "loss": 3.2204,
+      "step": 1505
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999235339698351e-05,
+      "loss": 3.2902,
+      "step": 1510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99922990709964e-05,
+      "loss": 3.2524,
+      "step": 1515
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999224455272468e-05,
+      "loss": 2.9775,
+      "step": 1520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99921898421686e-05,
+      "loss": 3.2531,
+      "step": 1525
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999213493932832e-05,
+      "loss": 3.0939,
+      "step": 1530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999207984420408e-05,
+      "loss": 3.1433,
+      "step": 1535
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999202455679608e-05,
+      "loss": 3.2009,
+      "step": 1540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999196907710456e-05,
+      "loss": 3.1226,
+      "step": 1545
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999191340512967e-05,
+      "loss": 3.2458,
+      "step": 1550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99918575408717e-05,
+      "loss": 3.1211,
+      "step": 1555
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999180148433082e-05,
+      "loss": 3.3063,
+      "step": 1560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999174523550726e-05,
+      "loss": 3.1646,
+      "step": 1565
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999168879440122e-05,
+      "loss": 3.1005,
+      "step": 1570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999163216101295e-05,
+      "loss": 3.2049,
+      "step": 1575
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999157533534263e-05,
+      "loss": 3.2438,
+      "step": 1580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99915183173905e-05,
+      "loss": 3.051,
+      "step": 1585
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999146110715678e-05,
+      "loss": 3.2616,
+      "step": 1590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999140370464168e-05,
+      "loss": 3.1501,
+      "step": 1595
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999134610984543e-05,
+      "loss": 3.3399,
+      "step": 1600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999128832276825e-05,
+      "loss": 3.2702,
+      "step": 1605
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999123034341035e-05,
+      "loss": 3.2332,
+      "step": 1610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999117217177197e-05,
+      "loss": 3.1417,
+      "step": 1615
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999111380785332e-05,
+      "loss": 3.2529,
+      "step": 1620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999105525165464e-05,
+      "loss": 3.2571,
+      "step": 1625
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999099650317614e-05,
+      "loss": 3.1014,
+      "step": 1630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999093756241807e-05,
+      "loss": 3.2111,
+      "step": 1635
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999087842938062e-05,
+      "loss": 3.2437,
+      "step": 1640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999081910406403e-05,
+      "loss": 3.1051,
+      "step": 1645
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999075958646855e-05,
+      "loss": 3.1446,
+      "step": 1650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999069987659438e-05,
+      "loss": 3.0355,
+      "step": 1655
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999063997444178e-05,
+      "loss": 3.1102,
+      "step": 1660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999057988001095e-05,
+      "loss": 3.1113,
+      "step": 1665
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999051959330215e-05,
+      "loss": 3.4744,
+      "step": 1670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999045911431559e-05,
+      "loss": 3.2147,
+      "step": 1675
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99903984430515e-05,
+      "loss": 3.227,
+      "step": 1680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999033757951013e-05,
+      "loss": 3.0937,
+      "step": 1685
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99902765236917e-05,
+      "loss": 3.2238,
+      "step": 1690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999021527559647e-05,
+      "loss": 3.0363,
+      "step": 1695
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999015383522463e-05,
+      "loss": 3.0911,
+      "step": 1700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999009220257646e-05,
+      "loss": 3.1832,
+      "step": 1705
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999003037765217e-05,
+      "loss": 3.3829,
+      "step": 1710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998996836045203e-05,
+      "loss": 3.0736,
+      "step": 1715
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998990615097624e-05,
+      "loss": 3.1726,
+      "step": 1720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998984374922506e-05,
+      "loss": 3.1802,
+      "step": 1725
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998978115519872e-05,
+      "loss": 3.1189,
+      "step": 1730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998971836889748e-05,
+      "loss": 3.2107,
+      "step": 1735
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998965539032157e-05,
+      "loss": 3.061,
+      "step": 1740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998959221947121e-05,
+      "loss": 3.1489,
+      "step": 1745
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998952885634668e-05,
+      "loss": 3.2177,
+      "step": 1750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998946530094821e-05,
+      "loss": 2.9971,
+      "step": 1755
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998940155327603e-05,
+      "loss": 3.0335,
+      "step": 1760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998933761333041e-05,
+      "loss": 3.209,
+      "step": 1765
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998927348111156e-05,
+      "loss": 3.2136,
+      "step": 1770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998920915661977e-05,
+      "loss": 3.1268,
+      "step": 1775
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998914463985526e-05,
+      "loss": 3.115,
+      "step": 1780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998907993081829e-05,
+      "loss": 3.056,
+      "step": 1785
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99890150295091e-05,
+      "loss": 3.0728,
+      "step": 1790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998894993592794e-05,
+      "loss": 3.1122,
+      "step": 1795
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998888465007507e-05,
+      "loss": 3.0688,
+      "step": 1800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998881917195074e-05,
+      "loss": 3.1029,
+      "step": 1805
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998875350155518e-05,
+      "loss": 3.1511,
+      "step": 1810
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998868763888867e-05,
+      "loss": 3.1069,
+      "step": 1815
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998862158395144e-05,
+      "loss": 3.169,
+      "step": 1820
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998855533674379e-05,
+      "loss": 3.0911,
+      "step": 1825
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998848889726591e-05,
+      "loss": 3.2131,
+      "step": 1830
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99884222655181e-05,
+      "loss": 3.1529,
+      "step": 1835
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99883554415006e-05,
+      "loss": 3.1461,
+      "step": 1840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998828842521368e-05,
+      "loss": 3.0577,
+      "step": 1845
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998822121665757e-05,
+      "loss": 3.0225,
+      "step": 1850
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998815381583255e-05,
+      "loss": 3.1291,
+      "step": 1855
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99880862227389e-05,
+      "loss": 3.1253,
+      "step": 1860
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998801843737683e-05,
+      "loss": 2.9886,
+      "step": 1865
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998795045974664e-05,
+      "loss": 3.2402,
+      "step": 1870
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998788228984857e-05,
+      "loss": 3.1631,
+      "step": 1875
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99878139276829e-05,
+      "loss": 3.0777,
+      "step": 1880
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998774537324987e-05,
+      "loss": 2.9883,
+      "step": 1885
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998767662654977e-05,
+      "loss": 3.2159,
+      "step": 1890
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998760768758285e-05,
+      "loss": 3.0846,
+      "step": 1895
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998753855634937e-05,
+      "loss": 3.0736,
+      "step": 1900
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99874692328496e-05,
+      "loss": 3.0781,
+      "step": 1905
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998739971708382e-05,
+      "loss": 3.1528,
+      "step": 1910
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998733000905229e-05,
+      "loss": 3.2124,
+      "step": 1915
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998726010875527e-05,
+      "loss": 3.0524,
+      "step": 1920
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998719001619304e-05,
+      "loss": 3.0928,
+      "step": 1925
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998711973136584e-05,
+      "loss": 3.1544,
+      "step": 1930
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998704925427398e-05,
+      "loss": 2.9552,
+      "step": 1935
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99869785849177e-05,
+      "loss": 3.074,
+      "step": 1940
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998690772329731e-05,
+      "loss": 3.0739,
+      "step": 1945
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998683666941305e-05,
+      "loss": 3.09,
+      "step": 1950
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998676542326519e-05,
+      "loss": 3.1415,
+      "step": 1955
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998669398485403e-05,
+      "loss": 2.9659,
+      "step": 1960
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998662235417982e-05,
+      "loss": 3.0495,
+      "step": 1965
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998655053124287e-05,
+      "loss": 3.1948,
+      "step": 1970
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99864785160434e-05,
+      "loss": 3.0888,
+      "step": 1975
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998640630858174e-05,
+      "loss": 3.1249,
+      "step": 1980
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998633390885814e-05,
+      "loss": 3.0892,
+      "step": 1985
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99862613168729e-05,
+      "loss": 3.0172,
+      "step": 1990
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998618853262627e-05,
+      "loss": 3.0274,
+      "step": 1995
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998611555611855e-05,
+      "loss": 3.0957,
+      "step": 2000
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998604238735e-05,
+      "loss": 3.1453,
+      "step": 2005
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998596902632094e-05,
+      "loss": 3.0487,
+      "step": 2010
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998589547303162e-05,
+      "loss": 3.1755,
+      "step": 2015
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998582172748235e-05,
+      "loss": 3.056,
+      "step": 2020
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998574778967336e-05,
+      "loss": 3.0649,
+      "step": 2025
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9985673659605e-05,
+      "loss": 3.1405,
+      "step": 2030
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998559933727752e-05,
+      "loss": 2.9374,
+      "step": 2035
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99855248226912e-05,
+      "loss": 3.1847,
+      "step": 2040
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998545011584634e-05,
+      "loss": 3.2155,
+      "step": 2045
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998537521674324e-05,
+      "loss": 3.082,
+      "step": 2050
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998530012538216e-05,
+      "loss": 3.1767,
+      "step": 2055
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99852248417634e-05,
+      "loss": 3.1336,
+      "step": 2060
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998514936588727e-05,
+      "loss": 3.0851,
+      "step": 2065
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998507369775403e-05,
+      "loss": 3.0343,
+      "step": 2070
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998499783736398e-05,
+      "loss": 3.0336,
+      "step": 2075
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998492178471741e-05,
+      "loss": 3.0263,
+      "step": 2080
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998484553981463e-05,
+      "loss": 3.0562,
+      "step": 2085
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998476910265592e-05,
+      "loss": 3.1234,
+      "step": 2090
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998469247324157e-05,
+      "loss": 3.3007,
+      "step": 2095
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998461565157187e-05,
+      "loss": 3.07,
+      "step": 2100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998453863764714e-05,
+      "loss": 3.1745,
+      "step": 2105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998446143146767e-05,
+      "loss": 3.0176,
+      "step": 2110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998438403303372e-05,
+      "loss": 3.0887,
+      "step": 2115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998430644234564e-05,
+      "loss": 3.1905,
+      "step": 2120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998422865940369e-05,
+      "loss": 3.1964,
+      "step": 2125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998415068420819e-05,
+      "loss": 3.0743,
+      "step": 2130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998407251675943e-05,
+      "loss": 2.992,
+      "step": 2135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998399415705771e-05,
+      "loss": 3.0476,
+      "step": 2140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998391560510334e-05,
+      "loss": 3.1099,
+      "step": 2145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998383686089663e-05,
+      "loss": 3.2379,
+      "step": 2150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998375792443785e-05,
+      "loss": 3.0703,
+      "step": 2155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998367879572733e-05,
+      "loss": 3.1361,
+      "step": 2160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998359947476537e-05,
+      "loss": 2.9603,
+      "step": 2165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998351996155227e-05,
+      "loss": 3.3157,
+      "step": 2170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998344025608835e-05,
+      "loss": 3.0359,
+      "step": 2175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99833603583739e-05,
+      "loss": 3.0545,
+      "step": 2180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998328026840922e-05,
+      "loss": 3.163,
+      "step": 2185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998319998619466e-05,
+      "loss": 2.8271,
+      "step": 2190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998311951173048e-05,
+      "loss": 3.0281,
+      "step": 2195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9983038845017e-05,
+      "loss": 3.1644,
+      "step": 2200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998295798605456e-05,
+      "loss": 3.0977,
+      "step": 2205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998287693484345e-05,
+      "loss": 3.0378,
+      "step": 2210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998279569138396e-05,
+      "loss": 2.9875,
+      "step": 2215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998271425567645e-05,
+      "loss": 3.0556,
+      "step": 2220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99826326277212e-05,
+      "loss": 3.167,
+      "step": 2225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998255080751853e-05,
+      "loss": 3.1431,
+      "step": 2230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998246879506876e-05,
+      "loss": 3.2236,
+      "step": 2235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99823865903722e-05,
+      "loss": 2.9233,
+      "step": 2240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998230419342916e-05,
+      "loss": 3.007,
+      "step": 2245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998222160423998e-05,
+      "loss": 3.0215,
+      "step": 2250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998213882280496e-05,
+      "loss": 3.0457,
+      "step": 2255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998205584912442e-05,
+      "loss": 3.0833,
+      "step": 2260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998197268319866e-05,
+      "loss": 3.1484,
+      "step": 2265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998188932502805e-05,
+      "loss": 2.9882,
+      "step": 2270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998180577461286e-05,
+      "loss": 3.066,
+      "step": 2275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998172203195345e-05,
+      "loss": 3.2224,
+      "step": 2280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998163809705012e-05,
+      "loss": 3.0801,
+      "step": 2285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99815539699032e-05,
+      "loss": 2.9881,
+      "step": 2290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998146965051301e-05,
+      "loss": 3.0504,
+      "step": 2295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998138513887987e-05,
+      "loss": 3.0293,
+      "step": 2300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998130043500411e-05,
+      "loss": 3.0291,
+      "step": 2305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998121553888606e-05,
+      "loss": 3.0811,
+      "step": 2310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998113045052604e-05,
+      "loss": 2.9213,
+      "step": 2315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998104516992436e-05,
+      "loss": 3.1569,
+      "step": 2320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99809596970814e-05,
+      "loss": 3.0974,
+      "step": 2325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998087403199745e-05,
+      "loss": 3.1625,
+      "step": 2330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998078817467284e-05,
+      "loss": 3.1312,
+      "step": 2335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99807021251079e-05,
+      "loss": 2.9758,
+      "step": 2340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998061588330299e-05,
+      "loss": 3.1046,
+      "step": 2345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99805294492584e-05,
+      "loss": 3.0321,
+      "step": 2350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998044282297447e-05,
+      "loss": 2.9867,
+      "step": 2355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998035600445157e-05,
+      "loss": 2.9097,
+      "step": 2360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998026899369e-05,
+      "loss": 3.1072,
+      "step": 2365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998018179069011e-05,
+      "loss": 3.0588,
+      "step": 2370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998009439545222e-05,
+      "loss": 3.0902,
+      "step": 2375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998000680797668e-05,
+      "loss": 2.9894,
+      "step": 2380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997991902826381e-05,
+      "loss": 3.1105,
+      "step": 2385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997983105631396e-05,
+      "loss": 3.0562,
+      "step": 2390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997974289212748e-05,
+      "loss": 3.0621,
+      "step": 2395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99796545357047e-05,
+      "loss": 3.0411,
+      "step": 2400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997956598704594e-05,
+      "loss": 3.028,
+      "step": 2405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997947724615157e-05,
+      "loss": 3.1136,
+      "step": 2410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997938831302191e-05,
+      "loss": 3.13,
+      "step": 2415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997929918765731e-05,
+      "loss": 3.151,
+      "step": 2420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997920987005813e-05,
+      "loss": 3.0431,
+      "step": 2425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997912036022469e-05,
+      "loss": 3.0306,
+      "step": 2430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997903065815734e-05,
+      "loss": 2.9331,
+      "step": 2435
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997894076385642e-05,
+      "loss": 3.0912,
+      "step": 2440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997885067732228e-05,
+      "loss": 3.1764,
+      "step": 2445
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997876039855528e-05,
+      "loss": 3.1066,
+      "step": 2450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997866992755576e-05,
+      "loss": 3.0018,
+      "step": 2455
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997857926432406e-05,
+      "loss": 3.1288,
+      "step": 2460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997848840886051e-05,
+      "loss": 2.9873,
+      "step": 2465
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99783973611655e-05,
+      "loss": 3.031,
+      "step": 2470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997830612123937e-05,
+      "loss": 3.0885,
+      "step": 2475
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997821468908245e-05,
+      "loss": 3.0257,
+      "step": 2480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997812306469512e-05,
+      "loss": 2.954,
+      "step": 2485
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997803124807769e-05,
+      "loss": 2.9539,
+      "step": 2490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997793923923056e-05,
+      "loss": 2.9983,
+      "step": 2495
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997784703815408e-05,
+      "loss": 3.1665,
+      "step": 2500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997775464484856e-05,
+      "loss": 3.0844,
+      "step": 2505
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997766205931439e-05,
+      "loss": 2.9777,
+      "step": 2510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997756928155193e-05,
+      "loss": 3.1405,
+      "step": 2515
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997747631156152e-05,
+      "loss": 3.0687,
+      "step": 2520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997738314934353e-05,
+      "loss": 3.008,
+      "step": 2525
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99772897948983e-05,
+      "loss": 3.1787,
+      "step": 2530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997719624822622e-05,
+      "loss": 3.0424,
+      "step": 2535
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997710250932761e-05,
+      "loss": 3.0007,
+      "step": 2540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997700857820287e-05,
+      "loss": 3.0611,
+      "step": 2545
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997691445485236e-05,
+      "loss": 2.9193,
+      "step": 2550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997682013927639e-05,
+      "loss": 2.9781,
+      "step": 2555
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997672563147538e-05,
+      "loss": 2.9307,
+      "step": 2560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997663093144967e-05,
+      "loss": 2.9735,
+      "step": 2565
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997653603919962e-05,
+      "loss": 2.9583,
+      "step": 2570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997644095472561e-05,
+      "loss": 2.9902,
+      "step": 2575
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997634567802798e-05,
+      "loss": 2.8588,
+      "step": 2580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997625020910713e-05,
+      "loss": 3.0293,
+      "step": 2585
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997615454796341e-05,
+      "loss": 3.03,
+      "step": 2590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997605869459719e-05,
+      "loss": 3.0432,
+      "step": 2595
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997596264900882e-05,
+      "loss": 3.0813,
+      "step": 2600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99758664111987e-05,
+      "loss": 3.0963,
+      "step": 2605
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99757699811672e-05,
+      "loss": 3.0172,
+      "step": 2610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997567335891465e-05,
+      "loss": 3.0422,
+      "step": 2615
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997557654444148e-05,
+      "loss": 3.0659,
+      "step": 2620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9975479537748e-05,
+      "loss": 3.0265,
+      "step": 2625
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997538233883463e-05,
+      "loss": 3.0182,
+      "step": 2630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997528494770174e-05,
+      "loss": 3.0681,
+      "step": 2635
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997518736434968e-05,
+      "loss": 3.1108,
+      "step": 2640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997508958877884e-05,
+      "loss": 3.1159,
+      "step": 2645
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99749916209896e-05,
+      "loss": 2.9908,
+      "step": 2650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997489346098232e-05,
+      "loss": 3.0677,
+      "step": 2655
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99747951087574e-05,
+      "loss": 2.9963,
+      "step": 2660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997469656431521e-05,
+      "loss": 3.163,
+      "step": 2665
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997459782765613e-05,
+      "loss": 3.067,
+      "step": 2670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997449889878053e-05,
+      "loss": 3.0481,
+      "step": 2675
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997439977768879e-05,
+      "loss": 3.0929,
+      "step": 2680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99743004643813e-05,
+      "loss": 3.0945,
+      "step": 2685
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997420095885843e-05,
+      "loss": 3.0011,
+      "step": 2690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997410126112058e-05,
+      "loss": 3.0403,
+      "step": 2695
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997400137116814e-05,
+      "loss": 3.0929,
+      "step": 2700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997390128900148e-05,
+      "loss": 2.9846,
+      "step": 2705
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997380101462097e-05,
+      "loss": 2.9595,
+      "step": 2710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997370054802702e-05,
+      "loss": 3.0691,
+      "step": 2715
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997359988922e-05,
+      "loss": 2.9539,
+      "step": 2720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997349903820032e-05,
+      "loss": 2.909,
+      "step": 2725
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997339799496835e-05,
+      "loss": 3.0498,
+      "step": 2730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997329675952447e-05,
+      "loss": 2.8829,
+      "step": 2735
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997319533186907e-05,
+      "loss": 3.1035,
+      "step": 2740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997309371200258e-05,
+      "loss": 2.9939,
+      "step": 2745
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997299189992534e-05,
+      "loss": 2.9376,
+      "step": 2750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997288989563777e-05,
+      "loss": 3.0272,
+      "step": 2755
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997278769914025e-05,
+      "loss": 3.1062,
+      "step": 2760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99726853104332e-05,
+      "loss": 2.8455,
+      "step": 2765
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997258272951697e-05,
+      "loss": 2.9995,
+      "step": 2770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997247995639198e-05,
+      "loss": 2.9894,
+      "step": 2775
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997237699105862e-05,
+      "loss": 2.9567,
+      "step": 2780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99722738335173e-05,
+      "loss": 3.0591,
+      "step": 2785
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997217048376839e-05,
+      "loss": 2.9647,
+      "step": 2790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997206694181232e-05,
+      "loss": 2.9484,
+      "step": 2795
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997196320764946e-05,
+      "loss": 3.1434,
+      "step": 2800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997185928128023e-05,
+      "loss": 3.028,
+      "step": 2805
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9971755162705e-05,
+      "loss": 2.9272,
+      "step": 2810
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99716508519242e-05,
+      "loss": 2.9792,
+      "step": 2815
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997154634893823e-05,
+      "loss": 3.0378,
+      "step": 2820
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997144165374748e-05,
+      "loss": 2.9958,
+      "step": 2825
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997133676635235e-05,
+      "loss": 2.8298,
+      "step": 2830
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997123168675325e-05,
+      "loss": 3.1673,
+      "step": 2835
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997112641495059e-05,
+      "loss": 2.9602,
+      "step": 2840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997102095094475e-05,
+      "loss": 2.9999,
+      "step": 2845
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997091529473617e-05,
+      "loss": 2.862,
+      "step": 2850
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997080944632524e-05,
+      "loss": 2.8886,
+      "step": 2855
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997070340571236e-05,
+      "loss": 3.0119,
+      "step": 2860
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997059717289796e-05,
+      "loss": 2.9387,
+      "step": 2865
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997049074788242e-05,
+      "loss": 2.983,
+      "step": 2870
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997038413066617e-05,
+      "loss": 2.9671,
+      "step": 2875
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99702773212496e-05,
+      "loss": 3.1065,
+      "step": 2880
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997017031963315e-05,
+      "loss": 3.0717,
+      "step": 2885
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99700631258172e-05,
+      "loss": 3.0588,
+      "step": 2890
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996995573980219e-05,
+      "loss": 3.0656,
+      "step": 2895
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996984816158852e-05,
+      "loss": 2.9431,
+      "step": 2900
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996974039117661e-05,
+      "loss": 3.1271,
+      "step": 2905
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996963242856686e-05,
+      "loss": 3.0755,
+      "step": 2910
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996952427375969e-05,
+      "loss": 3.0249,
+      "step": 2915
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996941592675554e-05,
+      "loss": 2.982,
+      "step": 2920
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99693073875548e-05,
+      "loss": 3.036,
+      "step": 2925
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996919865615788e-05,
+      "loss": 2.9621,
+      "step": 2930
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996908973256524e-05,
+      "loss": 3.0802,
+      "step": 2935
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996898061677724e-05,
+      "loss": 3.0021,
+      "step": 2940
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996887130879435e-05,
+      "loss": 2.971,
+      "step": 2945
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996876180861698e-05,
+      "loss": 3.0793,
+      "step": 2950
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996865211624553e-05,
+      "loss": 2.9306,
+      "step": 2955
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996854223168044e-05,
+      "loss": 3.1775,
+      "step": 2960
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996843215492213e-05,
+      "loss": 2.8381,
+      "step": 2965
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9968321885971e-05,
+      "loss": 2.8186,
+      "step": 2970
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996821142482753e-05,
+      "loss": 2.972,
+      "step": 2975
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996810077149209e-05,
+      "loss": 2.9275,
+      "step": 2980
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996798992596512e-05,
+      "loss": 3.0649,
+      "step": 2985
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996787888824707e-05,
+      "loss": 2.996,
+      "step": 2990
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996776765833835e-05,
+      "loss": 2.9119,
+      "step": 2995
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996765623623935e-05,
+      "loss": 2.9854,
+      "step": 3000
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996754462195057e-05,
+      "loss": 3.0353,
+      "step": 3005
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99674328154724e-05,
+      "loss": 3.0521,
+      "step": 3010
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996732081680528e-05,
+      "loss": 2.9823,
+      "step": 3015
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996720862594963e-05,
+      "loss": 2.9751,
+      "step": 3020
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996709624290588e-05,
+      "loss": 2.9864,
+      "step": 3025
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996698366767448e-05,
+      "loss": 2.9163,
+      "step": 3030
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996687090025585e-05,
+      "loss": 2.9331,
+      "step": 3035
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996675794065043e-05,
+      "loss": 2.9456,
+      "step": 3040
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996664478885864e-05,
+      "loss": 3.0654,
+      "step": 3045
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996653144488093e-05,
+      "loss": 2.9068,
+      "step": 3050
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996641790871772e-05,
+      "loss": 3.1059,
+      "step": 3055
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996630418036948e-05,
+      "loss": 3.0286,
+      "step": 3060
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996619025983662e-05,
+      "loss": 2.8734,
+      "step": 3065
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996607614711959e-05,
+      "loss": 2.8923,
+      "step": 3070
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99659618422188e-05,
+      "loss": 3.0052,
+      "step": 3075
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996584734513473e-05,
+      "loss": 3.0417,
+      "step": 3080
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99657326558678e-05,
+      "loss": 2.855,
+      "step": 3085
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996561777441846e-05,
+      "loss": 2.9128,
+      "step": 3090
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996550270078713e-05,
+      "loss": 3.0786,
+      "step": 3095
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996538743497428e-05,
+      "loss": 3.1021,
+      "step": 3100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996527197698035e-05,
+      "loss": 3.0133,
+      "step": 3105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996515632680577e-05,
+      "loss": 2.9493,
+      "step": 3110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996504048445099e-05,
+      "loss": 3.0326,
+      "step": 3115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996492444991646e-05,
+      "loss": 2.9906,
+      "step": 3120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99648082232026e-05,
+      "loss": 2.8989,
+      "step": 3125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99646918043099e-05,
+      "loss": 3.0664,
+      "step": 3130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996457519323879e-05,
+      "loss": 2.9872,
+      "step": 3135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996445838998971e-05,
+      "loss": 3.0031,
+      "step": 3140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996434139456313e-05,
+      "loss": 2.9665,
+      "step": 3145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996422420695946e-05,
+      "loss": 2.8623,
+      "step": 3150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996410682717919e-05,
+      "loss": 2.9544,
+      "step": 3155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996398925522276e-05,
+      "loss": 3.0457,
+      "step": 3160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996387149109061e-05,
+      "loss": 2.9199,
+      "step": 3165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996375353478323e-05,
+      "loss": 2.9197,
+      "step": 3170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9963635386301e-05,
+      "loss": 2.9216,
+      "step": 3175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996351704564446e-05,
+      "loss": 2.9656,
+      "step": 3180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996339851281401e-05,
+      "loss": 2.8536,
+      "step": 3185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996327978781013e-05,
+      "loss": 3.0652,
+      "step": 3190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996316087063327e-05,
+      "loss": 3.1083,
+      "step": 3195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996304176128386e-05,
+      "loss": 3.0493,
+      "step": 3200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996292245976241e-05,
+      "loss": 2.8686,
+      "step": 3205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996280296606934e-05,
+      "loss": 2.9044,
+      "step": 3210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996268328020514e-05,
+      "loss": 2.9402,
+      "step": 3215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996256340217024e-05,
+      "loss": 2.9844,
+      "step": 3220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996244333196512e-05,
+      "loss": 2.9016,
+      "step": 3225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996232306959023e-05,
+      "loss": 2.9768,
+      "step": 3230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996220261504604e-05,
+      "loss": 3.0186,
+      "step": 3235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996208196833302e-05,
+      "loss": 3.0875,
+      "step": 3240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996196112945162e-05,
+      "loss": 2.8911,
+      "step": 3245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99618400984023e-05,
+      "loss": 2.8954,
+      "step": 3250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996171887518555e-05,
+      "loss": 3.1472,
+      "step": 3255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996159745980182e-05,
+      "loss": 2.9414,
+      "step": 3260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996147585225158e-05,
+      "loss": 2.9011,
+      "step": 3265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996135405253528e-05,
+      "loss": 2.9499,
+      "step": 3270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996123206065344e-05,
+      "loss": 3.031,
+      "step": 3275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996110987660646e-05,
+      "loss": 2.9591,
+      "step": 3280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996098750039486e-05,
+      "loss": 2.8978,
+      "step": 3285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99608649320191e-05,
+      "loss": 2.9656,
+      "step": 3290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996074217147964e-05,
+      "loss": 2.8397,
+      "step": 3295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996061921877696e-05,
+      "loss": 2.9753,
+      "step": 3300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996049607391153e-05,
+      "loss": 2.8962,
+      "step": 3305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996037273688384e-05,
+      "loss": 2.8723,
+      "step": 3310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996024920769433e-05,
+      "loss": 3.0633,
+      "step": 3315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996012548634351e-05,
+      "loss": 2.9734,
+      "step": 3320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996000157283183e-05,
+      "loss": 2.9704,
+      "step": 3325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995987746715977e-05,
+      "loss": 3.1029,
+      "step": 3330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995975316932785e-05,
+      "loss": 2.9785,
+      "step": 3335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995962867933648e-05,
+      "loss": 2.9739,
+      "step": 3340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995950399718619e-05,
+      "loss": 2.8747,
+      "step": 3345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995937912287743e-05,
+      "loss": 2.8958,
+      "step": 3350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995925405641068e-05,
+      "loss": 2.9812,
+      "step": 3355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995912879778646e-05,
+      "loss": 3.0432,
+      "step": 3360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995900334700522e-05,
+      "loss": 3.0119,
+      "step": 3365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995887770406745e-05,
+      "loss": 2.8706,
+      "step": 3370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995875186897361e-05,
+      "loss": 2.8969,
+      "step": 3375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995862584172422e-05,
+      "loss": 3.0507,
+      "step": 3380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995849962231973e-05,
+      "loss": 2.9064,
+      "step": 3385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995837321076067e-05,
+      "loss": 2.9102,
+      "step": 3390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99582466070475e-05,
+      "loss": 2.7253,
+      "step": 3395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995811981118068e-05,
+      "loss": 2.9038,
+      "step": 3400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995799282316074e-05,
+      "loss": 2.7778,
+      "step": 3405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995786564298816e-05,
+      "loss": 2.9654,
+      "step": 3410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995773827066343e-05,
+      "loss": 2.8384,
+      "step": 3415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995761070618702e-05,
+      "loss": 2.9263,
+      "step": 3420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995748294955943e-05,
+      "loss": 2.9436,
+      "step": 3425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995735500078116e-05,
+      "loss": 2.7846,
+      "step": 3430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99572268598527e-05,
+      "loss": 3.0521,
+      "step": 3435
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995709852677454e-05,
+      "loss": 3.0054,
+      "step": 3440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995697000154717e-05,
+      "loss": 3.1224,
+      "step": 3445
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995684128417109e-05,
+      "loss": 2.9929,
+      "step": 3450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995671237464678e-05,
+      "loss": 3.0295,
+      "step": 3455
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995658327297476e-05,
+      "loss": 2.7867,
+      "step": 3460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995645397915554e-05,
+      "loss": 2.866,
+      "step": 3465
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995632449318956e-05,
+      "loss": 3.0005,
+      "step": 3470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995619481507736e-05,
+      "loss": 2.8469,
+      "step": 3475
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995606494481944e-05,
+      "loss": 2.8668,
+      "step": 3480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995593488241627e-05,
+      "loss": 3.0087,
+      "step": 3485
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995580462786838e-05,
+      "loss": 3.0151,
+      "step": 3490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995567418117626e-05,
+      "loss": 2.8603,
+      "step": 3495
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99555435423404e-05,
+      "loss": 2.8283,
+      "step": 3500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995541271136133e-05,
+      "loss": 2.9228,
+      "step": 3505
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995528168823953e-05,
+      "loss": 3.0148,
+      "step": 3510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995515047297552e-05,
+      "loss": 3.0812,
+      "step": 3515
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995501906556979e-05,
+      "loss": 2.9919,
+      "step": 3520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995488746602286e-05,
+      "loss": 3.0645,
+      "step": 3525
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995475567433521e-05,
+      "loss": 3.0459,
+      "step": 3530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995462369050737e-05,
+      "loss": 2.9637,
+      "step": 3535
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995449151453986e-05,
+      "loss": 2.9818,
+      "step": 3540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995435914643315e-05,
+      "loss": 2.9173,
+      "step": 3545
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995422658618779e-05,
+      "loss": 2.9242,
+      "step": 3550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995409383380425e-05,
+      "loss": 2.965,
+      "step": 3555
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995396088928307e-05,
+      "loss": 3.0302,
+      "step": 3560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995382775262475e-05,
+      "loss": 2.9297,
+      "step": 3565
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99536944238298e-05,
+      "loss": 3.0348,
+      "step": 3570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995356090289875e-05,
+      "loss": 2.8781,
+      "step": 3575
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995342718983209e-05,
+      "loss": 2.8363,
+      "step": 3580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995329328463034e-05,
+      "loss": 2.861,
+      "step": 3585
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995315918729403e-05,
+      "loss": 2.9402,
+      "step": 3590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995302489782365e-05,
+      "loss": 3.0611,
+      "step": 3595
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995289041621975e-05,
+      "loss": 3.0562,
+      "step": 3600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995275574248283e-05,
+      "loss": 2.9117,
+      "step": 3605
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99526208766134e-05,
+      "loss": 2.837,
+      "step": 3610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995248581861199e-05,
+      "loss": 2.9863,
+      "step": 3615
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995235056847911e-05,
+      "loss": 2.8425,
+      "step": 3620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995221512621527e-05,
+      "loss": 2.8012,
+      "step": 3625
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995207949182104e-05,
+      "loss": 2.9657,
+      "step": 3630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995194366529687e-05,
+      "loss": 3.0224,
+      "step": 3635
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995180764664336e-05,
+      "loss": 2.8736,
+      "step": 3640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995167143586098e-05,
+      "loss": 2.8935,
+      "step": 3645
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995153503295026e-05,
+      "loss": 2.9034,
+      "step": 3650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995139843791173e-05,
+      "loss": 2.8983,
+      "step": 3655
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995126165074592e-05,
+      "loss": 2.9201,
+      "step": 3660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995112467145336e-05,
+      "loss": 2.997,
+      "step": 3665
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995098750003456e-05,
+      "loss": 2.8996,
+      "step": 3670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995085013649007e-05,
+      "loss": 2.8433,
+      "step": 3675
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99507125808204e-05,
+      "loss": 3.0431,
+      "step": 3680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99505748330261e-05,
+      "loss": 3.1236,
+      "step": 3685
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995043689310767e-05,
+      "loss": 2.8728,
+      "step": 3690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995029876106565e-05,
+      "loss": 2.8317,
+      "step": 3695
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99501604369006e-05,
+      "loss": 2.7119,
+      "step": 3700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9950021920613e-05,
+      "loss": 2.9074,
+      "step": 3705
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994988321220345e-05,
+      "loss": 2.9848,
+      "step": 3710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994974431167241e-05,
+      "loss": 2.7338,
+      "step": 3715
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994960521902048e-05,
+      "loss": 2.8224,
+      "step": 3720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994946593424815e-05,
+      "loss": 3.0732,
+      "step": 3725
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994932645735597e-05,
+      "loss": 3.0064,
+      "step": 3730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994918678834446e-05,
+      "loss": 2.9211,
+      "step": 3735
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99490469272142e-05,
+      "loss": 2.844,
+      "step": 3740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994890687396569e-05,
+      "loss": 3.117,
+      "step": 3745
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994876662859948e-05,
+      "loss": 2.9603,
+      "step": 3750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994862619111613e-05,
+      "loss": 3.0004,
+      "step": 3755
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994848556151615e-05,
+      "loss": 2.9558,
+      "step": 3760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994834473980009e-05,
+      "loss": 2.864,
+      "step": 3765
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994820372596848e-05,
+      "loss": 2.7642,
+      "step": 3770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994806252002189e-05,
+      "loss": 2.8008,
+      "step": 3775
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994792112196085e-05,
+      "loss": 2.9414,
+      "step": 3780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99477795317859e-05,
+      "loss": 2.9503,
+      "step": 3785
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994763774949758e-05,
+      "loss": 2.999,
+      "step": 3790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994749577509644e-05,
+      "loss": 2.8423,
+      "step": 3795
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994735360858304e-05,
+      "loss": 2.9063,
+      "step": 3800
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994721124995791e-05,
+      "loss": 2.7602,
+      "step": 3805
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994706869922161e-05,
+      "loss": 3.0747,
+      "step": 3810
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994692595637467e-05,
+      "loss": 2.9135,
+      "step": 3815
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994678302141765e-05,
+      "loss": 2.933,
+      "step": 3820
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994663989435111e-05,
+      "loss": 2.9197,
+      "step": 3825
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994649657517559e-05,
+      "loss": 2.8764,
+      "step": 3830
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994635306389163e-05,
+      "loss": 3.0742,
+      "step": 3835
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994620936049981e-05,
+      "loss": 2.9927,
+      "step": 3840
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994606546500064e-05,
+      "loss": 2.9222,
+      "step": 3845
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994592137739474e-05,
+      "loss": 2.8627,
+      "step": 3850
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99457770976826e-05,
+      "loss": 2.9105,
+      "step": 3855
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99456326258648e-05,
+      "loss": 2.7945,
+      "step": 3860
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994548796194188e-05,
+      "loss": 3.051,
+      "step": 3865
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994534310591444e-05,
+      "loss": 2.8925,
+      "step": 3870
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994519805778299e-05,
+      "loss": 3.108,
+      "step": 3875
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994505281754811e-05,
+      "loss": 2.9647,
+      "step": 3880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994490738521035e-05,
+      "loss": 2.8922,
+      "step": 3885
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994476176077028e-05,
+      "loss": 2.9536,
+      "step": 3890
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994461594422846e-05,
+      "loss": 2.7955,
+      "step": 3895
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994446993558544e-05,
+      "loss": 2.9858,
+      "step": 3900
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994432373484178e-05,
+      "loss": 2.8623,
+      "step": 3905
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994417734199805e-05,
+      "loss": 2.8655,
+      "step": 3910
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994403075705483e-05,
+      "loss": 3.0108,
+      "step": 3915
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994388398001264e-05,
+      "loss": 2.9468,
+      "step": 3920
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994373701087209e-05,
+      "loss": 2.8976,
+      "step": 3925
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994358984963372e-05,
+      "loss": 2.8425,
+      "step": 3930
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99434424962981e-05,
+      "loss": 2.9477,
+      "step": 3935
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994329495086579e-05,
+      "loss": 2.7548,
+      "step": 3940
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994314721333739e-05,
+      "loss": 2.8532,
+      "step": 3945
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994299928371342e-05,
+      "loss": 2.9597,
+      "step": 3950
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994285116199448e-05,
+      "loss": 2.8252,
+      "step": 3955
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994270284818114e-05,
+      "loss": 2.8548,
+      "step": 3960
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994255434227396e-05,
+      "loss": 2.9002,
+      "step": 3965
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994240564427351e-05,
+      "loss": 2.9655,
+      "step": 3970
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994225675418038e-05,
+      "loss": 2.9334,
+      "step": 3975
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994210767199513e-05,
+      "loss": 2.8629,
+      "step": 3980
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994195839771832e-05,
+      "loss": 2.8862,
+      "step": 3985
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994180893135052e-05,
+      "loss": 2.9174,
+      "step": 3990
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994165927289235e-05,
+      "loss": 2.8629,
+      "step": 3995
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994150942234435e-05,
+      "loss": 2.8059,
+      "step": 4000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99413593797071e-05,
+      "loss": 2.8777,
+      "step": 4005
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994120914498118e-05,
+      "loss": 2.7953,
+      "step": 4010
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994105871816717e-05,
+      "loss": 2.9581,
+      "step": 4015
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994090809926565e-05,
+      "loss": 2.9255,
+      "step": 4020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99407572882772e-05,
+      "loss": 3.0222,
+      "step": 4025
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994060628520239e-05,
+      "loss": 2.931,
+      "step": 4030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994045509004182e-05,
+      "loss": 2.9042,
+      "step": 4035
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994030370279603e-05,
+      "loss": 2.9237,
+      "step": 4040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994015212346566e-05,
+      "loss": 3.0383,
+      "step": 4045
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994000035205126e-05,
+      "loss": 2.9757,
+      "step": 4050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99398483885534e-05,
+      "loss": 2.7862,
+      "step": 4055
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99396962329727e-05,
+      "loss": 2.8734,
+      "step": 4060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993954388530973e-05,
+      "loss": 2.8215,
+      "step": 4065
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993939134556506e-05,
+      "loss": 2.7965,
+      "step": 4070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993923861373929e-05,
+      "loss": 2.9514,
+      "step": 4075
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993908568983302e-05,
+      "loss": 2.862,
+      "step": 4080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99389325738468e-05,
+      "loss": 2.9218,
+      "step": 4085
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993877926578128e-05,
+      "loss": 2.8987,
+      "step": 4090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993862576563701e-05,
+      "loss": 2.9184,
+      "step": 4095
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993847207341457e-05,
+      "loss": 3.005,
+      "step": 4100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993831818911456e-05,
+      "loss": 2.8927,
+      "step": 4105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993816411273759e-05,
+      "loss": 2.9092,
+      "step": 4110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993800984428425e-05,
+      "loss": 2.8178,
+      "step": 4115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993785538375511e-05,
+      "loss": 2.933,
+      "step": 4120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993770073115079e-05,
+      "loss": 2.8886,
+      "step": 4125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993754588647186e-05,
+      "loss": 2.6325,
+      "step": 4130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993739084971894e-05,
+      "loss": 3.0143,
+      "step": 4135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99372356208926e-05,
+      "loss": 2.8982,
+      "step": 4140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993708019999346e-05,
+      "loss": 2.8282,
+      "step": 4145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993692458702211e-05,
+      "loss": 3.0452,
+      "step": 4150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993676878197916e-05,
+      "loss": 2.9051,
+      "step": 4155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993661278486518e-05,
+      "loss": 2.9052,
+      "step": 4160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993645659568079e-05,
+      "loss": 2.9753,
+      "step": 4165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993630021442659e-05,
+      "loss": 2.8703,
+      "step": 4170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993614364110319e-05,
+      "loss": 2.8597,
+      "step": 4175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993598687571118e-05,
+      "loss": 2.8084,
+      "step": 4180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993582991825114e-05,
+      "loss": 2.8753,
+      "step": 4185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993567276872373e-05,
+      "loss": 2.9646,
+      "step": 4190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993551542712951e-05,
+      "loss": 3.039,
+      "step": 4195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99353578934691e-05,
+      "loss": 2.8552,
+      "step": 4200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993520016774311e-05,
+      "loss": 2.9043,
+      "step": 4205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993504224995212e-05,
+      "loss": 2.9262,
+      "step": 4210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993488414009677e-05,
+      "loss": 2.8822,
+      "step": 4215
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993472583817767e-05,
+      "loss": 2.8048,
+      "step": 4220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99345673441954e-05,
+      "loss": 2.9189,
+      "step": 4225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99344086581506e-05,
+      "loss": 2.875,
+      "step": 4230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993424978004384e-05,
+      "loss": 2.9068,
+      "step": 4235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993409070987578e-05,
+      "loss": 2.8754,
+      "step": 4240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.9933931447647e-05,
+      "loss": 2.891,
+      "step": 4245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993377199335812e-05,
+      "loss": 2.7851,
+      "step": 4250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993361234700975e-05,
+      "loss": 2.877,
+      "step": 4255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993345250860251e-05,
+      "loss": 2.8454,
+      "step": 4260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993329247813701e-05,
+      "loss": 2.9549,
+      "step": 4265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993313225561388e-05,
+      "loss": 2.9104,
+      "step": 4270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993297184103372e-05,
+      "loss": 2.8908,
+      "step": 4275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993281123439715e-05,
+      "loss": 2.8366,
+      "step": 4280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993265043570479e-05,
+      "loss": 2.9557,
+      "step": 4285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993248944495725e-05,
+      "loss": 2.9455,
+      "step": 4290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993232826215517e-05,
+      "loss": 2.8195,
+      "step": 4295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993216688729915e-05,
+      "loss": 2.9566,
+      "step": 4300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993200532038982e-05,
+      "loss": 2.9423,
+      "step": 4305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99318435614278e-05,
+      "loss": 2.8711,
+      "step": 4310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99316816104137e-05,
+      "loss": 2.7991,
+      "step": 4315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993151946734817e-05,
+      "loss": 2.8641,
+      "step": 4320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99313571322318e-05,
+      "loss": 2.9935,
+      "step": 4325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993119460506526e-05,
+      "loss": 3.0888,
+      "step": 4330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993103188584912e-05,
+      "loss": 2.7793,
+      "step": 4335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993086897458406e-05,
+      "loss": 2.887,
+      "step": 4340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993070587127064e-05,
+      "loss": 2.8533,
+      "step": 4345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993054257590958e-05,
+      "loss": 2.9726,
+      "step": 4350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993037908850141e-05,
+      "loss": 2.8357,
+      "step": 4355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993021540904683e-05,
+      "loss": 2.8525,
+      "step": 4360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993005153754642e-05,
+      "loss": 2.8359,
+      "step": 4365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992988747400085e-05,
+      "loss": 2.9078,
+      "step": 4370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992972321841075e-05,
+      "loss": 3.0275,
+      "step": 4375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99295587707767e-05,
+      "loss": 2.8292,
+      "step": 4380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99293941310994e-05,
+      "loss": 2.9929,
+      "step": 4385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992922929937945e-05,
+      "loss": 2.8618,
+      "step": 4390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992906427561747e-05,
+      "loss": 2.8389,
+      "step": 4395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992889905981413e-05,
+      "loss": 2.7753,
+      "step": 4400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992873365197004e-05,
+      "loss": 2.9134,
+      "step": 4405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992856805208583e-05,
+      "loss": 2.9741,
+      "step": 4410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992840226016216e-05,
+      "loss": 2.919,
+      "step": 4415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992823627619966e-05,
+      "loss": 2.8932,
+      "step": 4420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992807010019896e-05,
+      "loss": 2.9778,
+      "step": 4425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992790373216071e-05,
+      "loss": 2.8878,
+      "step": 4430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992773717208555e-05,
+      "loss": 3.0102,
+      "step": 4435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992757041997411e-05,
+      "loss": 2.8754,
+      "step": 4440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992740347582703e-05,
+      "loss": 2.9329,
+      "step": 4445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992723633964496e-05,
+      "loss": 2.8539,
+      "step": 4450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992706901142855e-05,
+      "loss": 2.9462,
+      "step": 4455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992690149117844e-05,
+      "loss": 3.0054,
+      "step": 4460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992673377889525e-05,
+      "loss": 2.9757,
+      "step": 4465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992656587457968e-05,
+      "loss": 2.8828,
+      "step": 4470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99263977782323e-05,
+      "loss": 2.9127,
+      "step": 4475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992622948985382e-05,
+      "loss": 2.8061,
+      "step": 4480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992606100944486e-05,
+      "loss": 2.9212,
+      "step": 4485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992589233700607e-05,
+      "loss": 2.9007,
+      "step": 4490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99257234725381e-05,
+      "loss": 2.7598,
+      "step": 4495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992555441604162e-05,
+      "loss": 2.9011,
+      "step": 4500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992538516751725e-05,
+      "loss": 2.9784,
+      "step": 4505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992521572696565e-05,
+      "loss": 2.7545,
+      "step": 4510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992504609438747e-05,
+      "loss": 2.9484,
+      "step": 4515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992487626978338e-05,
+      "loss": 2.7525,
+      "step": 4520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.9924706253154e-05,
+      "loss": 2.8493,
+      "step": 4525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992453604450001e-05,
+      "loss": 2.8149,
+      "step": 4530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992436564382206e-05,
+      "loss": 2.9744,
+      "step": 4535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99241950511208e-05,
+      "loss": 2.7981,
+      "step": 4540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99240242663969e-05,
+      "loss": 2.789,
+      "step": 4545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.9923853289651e-05,
+      "loss": 2.8568,
+      "step": 4550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992368212088377e-05,
+      "loss": 2.8879,
+      "step": 4555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992351076009585e-05,
+      "loss": 2.698,
+      "step": 4560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992333920728792e-05,
+      "loss": 2.8686,
+      "step": 4565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992316746246065e-05,
+      "loss": 2.7768,
+      "step": 4570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992299552561465e-05,
+      "loss": 2.8642,
+      "step": 4575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992282339675063e-05,
+      "loss": 2.8224,
+      "step": 4580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992265107586923e-05,
+      "loss": 2.8072,
+      "step": 4585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992247856297112e-05,
+      "loss": 3.0023,
+      "step": 4590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992230585805695e-05,
+      "loss": 2.8443,
+      "step": 4595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992213296112743e-05,
+      "loss": 2.9238,
+      "step": 4600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992195987218316e-05,
+      "loss": 2.8344,
+      "step": 4605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992178659122485e-05,
+      "loss": 2.9624,
+      "step": 4610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992161311825314e-05,
+      "loss": 2.8404,
+      "step": 4615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992143945326873e-05,
+      "loss": 2.7673,
+      "step": 4620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992126559627225e-05,
+      "loss": 2.8661,
+      "step": 4625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992109154726439e-05,
+      "loss": 2.8625,
+      "step": 4630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992091730624582e-05,
+      "loss": 2.9644,
+      "step": 4635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992074287321722e-05,
+      "loss": 2.9968,
+      "step": 4640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992056824817923e-05,
+      "loss": 2.8742,
+      "step": 4645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992039343113255e-05,
+      "loss": 2.9069,
+      "step": 4650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992021842207783e-05,
+      "loss": 2.8728,
+      "step": 4655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992004322101577e-05,
+      "loss": 3.0079,
+      "step": 4660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991986782794703e-05,
+      "loss": 2.8492,
+      "step": 4665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991969224287227e-05,
+      "loss": 3.0581,
+      "step": 4670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991951646579219e-05,
+      "loss": 2.9742,
+      "step": 4675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991934049670744e-05,
+      "loss": 2.7371,
+      "step": 4680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991916433561872e-05,
+      "loss": 2.9754,
+      "step": 4685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99189879825267e-05,
+      "loss": 2.7925,
+      "step": 4690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991881143743205e-05,
+      "loss": 2.8926,
+      "step": 4695
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991863470033546e-05,
+      "loss": 2.8813,
+      "step": 4700
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991845777123761e-05,
+      "loss": 2.9417,
+      "step": 4705
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991828065013918e-05,
+      "loss": 2.8718,
+      "step": 4710
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991810333704086e-05,
+      "loss": 2.7872,
+      "step": 4715
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991792583194331e-05,
+      "loss": 2.8953,
+      "step": 4720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99177481348472e-05,
+      "loss": 2.8947,
+      "step": 4725
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991757024575326e-05,
+      "loss": 2.7719,
+      "step": 4730
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991739216466214e-05,
+      "loss": 2.8671,
+      "step": 4735
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991721389157455e-05,
+      "loss": 2.8587,
+      "step": 4740
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991703542649115e-05,
+      "loss": 2.9225,
+      "step": 4745
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991685676941264e-05,
+      "loss": 2.772,
+      "step": 4750
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99166779203397e-05,
+      "loss": 3.0555,
+      "step": 4755
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991649887927304e-05,
+      "loss": 2.9068,
+      "step": 4760
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99163196462133e-05,
+      "loss": 2.8922,
+      "step": 4765
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991614022116123e-05,
+      "loss": 2.746,
+      "step": 4770
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991596060411747e-05,
+      "loss": 2.8224,
+      "step": 4775
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991578079508274e-05,
+      "loss": 2.8176,
+      "step": 4780
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991560079405773e-05,
+      "loss": 2.8898,
+      "step": 4785
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991542060104312e-05,
+      "loss": 2.9647,
+      "step": 4790
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991524021603961e-05,
+      "loss": 2.8396,
+      "step": 4795
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99150596390479e-05,
+      "loss": 2.9884,
+      "step": 4800
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991487887006867e-05,
+      "loss": 2.9154,
+      "step": 4805
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991469790910261e-05,
+      "loss": 2.7634,
+      "step": 4810
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991451675615045e-05,
+      "loss": 2.956,
+      "step": 4815
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991433541121286e-05,
+      "loss": 3.0876,
+      "step": 4820
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991415387429053e-05,
+      "loss": 2.7038,
+      "step": 4825
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991397214538419e-05,
+      "loss": 2.9156,
+      "step": 4830
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99137902244945e-05,
+      "loss": 2.81,
+      "step": 4835
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991360811162219e-05,
+      "loss": 3.0065,
+      "step": 4840
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991342580676794e-05,
+      "loss": 2.8665,
+      "step": 4845
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991324330993248e-05,
+      "loss": 2.8536,
+      "step": 4850
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991306062111649e-05,
+      "loss": 2.7937,
+      "step": 4855
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991287774032065e-05,
+      "loss": 2.8707,
+      "step": 4860
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991269466754571e-05,
+      "loss": 2.8644,
+      "step": 4865
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991251140279236e-05,
+      "loss": 2.7943,
+      "step": 4870
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991232794606129e-05,
+      "loss": 2.7979,
+      "step": 4875
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991214429735322e-05,
+      "loss": 2.9018,
+      "step": 4880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991196045666884e-05,
+      "loss": 2.9494,
+      "step": 4885
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991177642400887e-05,
+      "loss": 2.8961,
+      "step": 4890
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991159219937401e-05,
+      "loss": 2.7802,
+      "step": 4895
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991140778276499e-05,
+      "loss": 2.971,
+      "step": 4900
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991122317418249e-05,
+      "loss": 2.7642,
+      "step": 4905
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991103837362725e-05,
+      "loss": 2.8931,
+      "step": 4910
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991085338109993e-05,
+      "loss": 2.7499,
+      "step": 4915
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99106681966013e-05,
+      "loss": 2.9113,
+      "step": 4920
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991048282013204e-05,
+      "loss": 2.8981,
+      "step": 4925
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991029725169289e-05,
+      "loss": 2.7878,
+      "step": 4930
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991011149128453e-05,
+      "loss": 2.8287,
+      "step": 4935
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990992553890769e-05,
+      "loss": 2.791,
+      "step": 4940
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990973939456307e-05,
+      "loss": 2.8164,
+      "step": 4945
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990955305825143e-05,
+      "loss": 2.8767,
+      "step": 4950
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990936652997344e-05,
+      "loss": 3.0286,
+      "step": 4955
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990917980972983e-05,
+      "loss": 2.828,
+      "step": 4960
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990899289752133e-05,
+      "loss": 2.8306,
+      "step": 4965
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990880579334865e-05,
+      "loss": 2.79,
+      "step": 4970
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990861849721252e-05,
+      "loss": 2.8997,
+      "step": 4975
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990843100911362e-05,
+      "loss": 2.7331,
+      "step": 4980
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990824332905273e-05,
+      "loss": 2.9343,
+      "step": 4985
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990805545703055e-05,
+      "loss": 2.983,
+      "step": 4990
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990786739304779e-05,
+      "loss": 2.8792,
+      "step": 4995
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990767913710517e-05,
+      "loss": 2.7881,
+      "step": 5000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990749068920343e-05,
+      "loss": 2.8737,
+      "step": 5005
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990730204934329e-05,
+      "loss": 2.6785,
+      "step": 5010
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990711321752547e-05,
+      "loss": 2.9362,
+      "step": 5015
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990692419375072e-05,
+      "loss": 2.9137,
+      "step": 5020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990673497801972e-05,
+      "loss": 2.7605,
+      "step": 5025
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990654557033326e-05,
+      "loss": 2.7289,
+      "step": 5030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.9906355970692e-05,
+      "loss": 3.0102,
+      "step": 5035
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990616617909672e-05,
+      "loss": 2.8382,
+      "step": 5040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990597619554813e-05,
+      "loss": 2.7506,
+      "step": 5045
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990578602004698e-05,
+      "loss": 3.0032,
+      "step": 5050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990559565259397e-05,
+      "loss": 2.8493,
+      "step": 5055
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990540509318985e-05,
+      "loss": 2.9364,
+      "step": 5060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990521434183535e-05,
+      "loss": 2.7998,
+      "step": 5065
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990502339853121e-05,
+      "loss": 2.8367,
+      "step": 5070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990483226327816e-05,
+      "loss": 2.8306,
+      "step": 5075
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990464093607694e-05,
+      "loss": 2.9189,
+      "step": 5080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990444941692827e-05,
+      "loss": 2.9497,
+      "step": 5085
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99042577058329e-05,
+      "loss": 2.812,
+      "step": 5090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990406580279157e-05,
+      "loss": 2.8871,
+      "step": 5095
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990387370780501e-05,
+      "loss": 2.97,
+      "step": 5100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990368142087395e-05,
+      "loss": 2.7219,
+      "step": 5105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990348894199915e-05,
+      "loss": 2.6505,
+      "step": 5110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990329627118133e-05,
+      "loss": 2.8628,
+      "step": 5115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990310340842127e-05,
+      "loss": 2.8648,
+      "step": 5120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990291035371966e-05,
+      "loss": 2.9342,
+      "step": 5125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990271710707729e-05,
+      "loss": 2.7795,
+      "step": 5130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990252366849485e-05,
+      "loss": 2.8963,
+      "step": 5135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990233003797313e-05,
+      "loss": 2.8869,
+      "step": 5140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990213621551285e-05,
+      "loss": 2.8546,
+      "step": 5145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990194220111477e-05,
+      "loss": 2.7148,
+      "step": 5150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990174799477963e-05,
+      "loss": 2.7522,
+      "step": 5155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990155359650818e-05,
+      "loss": 2.9067,
+      "step": 5160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990135900630116e-05,
+      "loss": 2.8302,
+      "step": 5165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990116422415934e-05,
+      "loss": 2.8406,
+      "step": 5170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990096925008343e-05,
+      "loss": 2.8458,
+      "step": 5175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990077408407421e-05,
+      "loss": 2.893,
+      "step": 5180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990057872613243e-05,
+      "loss": 2.7953,
+      "step": 5185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990038317625883e-05,
+      "loss": 2.7391,
+      "step": 5190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990018743445418e-05,
+      "loss": 2.9778,
+      "step": 5195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98999915007192e-05,
+      "loss": 3.0209,
+      "step": 5200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989979537505468e-05,
+      "loss": 2.8533,
+      "step": 5205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989959905746135e-05,
+      "loss": 2.8606,
+      "step": 5210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989940254793997e-05,
+      "loss": 2.8147,
+      "step": 5215
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98992058464913e-05,
+      "loss": 2.7068,
+      "step": 5220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98990089531161e-05,
+      "loss": 2.8903,
+      "step": 5225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989881186781511e-05,
+      "loss": 2.8929,
+      "step": 5230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989861459058911e-05,
+      "loss": 2.8358,
+      "step": 5235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989841712143885e-05,
+      "loss": 2.8309,
+      "step": 5240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989821946036508e-05,
+      "loss": 2.7455,
+      "step": 5245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989802160736859e-05,
+      "loss": 2.9303,
+      "step": 5250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98978235624501e-05,
+      "loss": 2.9108,
+      "step": 5255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98976253256104e-05,
+      "loss": 2.8467,
+      "step": 5260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989742689685025e-05,
+      "loss": 2.7748,
+      "step": 5265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98972282761704e-05,
+      "loss": 2.9717,
+      "step": 5270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989702946357162e-05,
+      "loss": 2.7965,
+      "step": 5275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989683045905468e-05,
+      "loss": 2.8164,
+      "step": 5280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989663126262035e-05,
+      "loss": 2.9972,
+      "step": 5285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989643187426937e-05,
+      "loss": 2.7716,
+      "step": 5290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989623229400254e-05,
+      "loss": 2.7401,
+      "step": 5295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98960325218206e-05,
+      "loss": 2.8559,
+      "step": 5300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989583255772433e-05,
+      "loss": 2.8383,
+      "step": 5305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989563240171451e-05,
+      "loss": 2.7331,
+      "step": 5310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98954320537919e-05,
+      "loss": 2.6999,
+      "step": 5315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989523151395726e-05,
+      "loss": 3.0156,
+      "step": 5320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989503078221138e-05,
+      "loss": 2.7704,
+      "step": 5325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989482985855503e-05,
+      "loss": 2.8674,
+      "step": 5330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989462874298896e-05,
+      "loss": 2.7731,
+      "step": 5335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989442743551397e-05,
+      "loss": 2.8514,
+      "step": 5340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989422593613082e-05,
+      "loss": 2.7738,
+      "step": 5345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98940242448403e-05,
+      "loss": 2.8348,
+      "step": 5350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989382236164317e-05,
+      "loss": 2.9625,
+      "step": 5355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989362028654022e-05,
+      "loss": 2.8034,
+      "step": 5360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98934180195322e-05,
+      "loss": 2.675,
+      "step": 5365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98932155606199e-05,
+      "loss": 2.7373,
+      "step": 5370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989301290980413e-05,
+      "loss": 2.7883,
+      "step": 5375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989281006708563e-05,
+      "loss": 2.8881,
+      "step": 5380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989260703246521e-05,
+      "loss": 2.7805,
+      "step": 5385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989240380594362e-05,
+      "loss": 2.7529,
+      "step": 5390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989220038752168e-05,
+      "loss": 2.7396,
+      "step": 5395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989199677720012e-05,
+      "loss": 2.9087,
+      "step": 5400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989179297497977e-05,
+      "loss": 2.8232,
+      "step": 5405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989158898086139e-05,
+      "loss": 2.8819,
+      "step": 5410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989138479484579e-05,
+      "loss": 2.9828,
+      "step": 5415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989118041693372e-05,
+      "loss": 2.8742,
+      "step": 5420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989097584712599e-05,
+      "loss": 2.7802,
+      "step": 5425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989077108542338e-05,
+      "loss": 2.8046,
+      "step": 5430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989056613182667e-05,
+      "loss": 2.8288,
+      "step": 5435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989036098633667e-05,
+      "loss": 2.833,
+      "step": 5440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989015564895416e-05,
+      "loss": 2.8506,
+      "step": 5445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98899501196799e-05,
+      "loss": 2.8315,
+      "step": 5450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988974439851473e-05,
+      "loss": 2.6436,
+      "step": 5455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98895384854594e-05,
+      "loss": 2.7784,
+      "step": 5460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988933238051472e-05,
+      "loss": 2.9267,
+      "step": 5465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988912608368148e-05,
+      "loss": 2.8023,
+      "step": 5470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98889195949605e-05,
+      "loss": 2.8426,
+      "step": 5475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988871291435253e-05,
+      "loss": 2.7829,
+      "step": 5480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988850604185838e-05,
+      "loss": 2.9335,
+      "step": 5485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988829897747885e-05,
+      "loss": 2.7431,
+      "step": 5490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988809172121473e-05,
+      "loss": 2.9993,
+      "step": 5495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988788427306684e-05,
+      "loss": 2.9427,
+      "step": 5500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988767663303595e-05,
+      "loss": 2.8004,
+      "step": 5505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988746880112287e-05,
+      "loss": 2.8501,
+      "step": 5510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988726077732842e-05,
+      "loss": 2.7476,
+      "step": 5515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988705256165335e-05,
+      "loss": 2.8735,
+      "step": 5520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98868441540985e-05,
+      "loss": 3.0131,
+      "step": 5525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988663555466467e-05,
+      "loss": 2.7398,
+      "step": 5530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988642676335264e-05,
+      "loss": 2.8133,
+      "step": 5535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988621778016325e-05,
+      "loss": 2.8704,
+      "step": 5540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988600860509727e-05,
+      "loss": 2.914,
+      "step": 5545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988579923815551e-05,
+      "loss": 2.8573,
+      "step": 5550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988558967933879e-05,
+      "loss": 2.7427,
+      "step": 5555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988537992864789e-05,
+      "loss": 2.7723,
+      "step": 5560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988516998608365e-05,
+      "loss": 2.8097,
+      "step": 5565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988495985164685e-05,
+      "loss": 2.8393,
+      "step": 5570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988474952533833e-05,
+      "loss": 2.8884,
+      "step": 5575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988453900715886e-05,
+      "loss": 2.7273,
+      "step": 5580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988432829710926e-05,
+      "loss": 2.8235,
+      "step": 5585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988411739519036e-05,
+      "loss": 2.7715,
+      "step": 5590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988390630140296e-05,
+      "loss": 2.8221,
+      "step": 5595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988369501574788e-05,
+      "loss": 2.8419,
+      "step": 5600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988348353822592e-05,
+      "loss": 2.823,
+      "step": 5605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988327186883789e-05,
+      "loss": 2.7862,
+      "step": 5610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988306000758461e-05,
+      "loss": 2.929,
+      "step": 5615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988284795446691e-05,
+      "loss": 2.8953,
+      "step": 5620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988263570948558e-05,
+      "loss": 2.7204,
+      "step": 5625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988242327264145e-05,
+      "loss": 2.8488,
+      "step": 5630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988221064393535e-05,
+      "loss": 2.7629,
+      "step": 5635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988199782336807e-05,
+      "loss": 2.8198,
+      "step": 5640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988178481094044e-05,
+      "loss": 2.8678,
+      "step": 5645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988157160665329e-05,
+      "loss": 2.7688,
+      "step": 5650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988135821050744e-05,
+      "loss": 2.7876,
+      "step": 5655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988114462250368e-05,
+      "loss": 2.7894,
+      "step": 5660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988093084264287e-05,
+      "loss": 2.8445,
+      "step": 5665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98807168709258e-05,
+      "loss": 2.7887,
+      "step": 5670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988050270735332e-05,
+      "loss": 2.7833,
+      "step": 5675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988028835192624e-05,
+      "loss": 2.7774,
+      "step": 5680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98800738046454e-05,
+      "loss": 2.8728,
+      "step": 5685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987985906551161e-05,
+      "loss": 2.7462,
+      "step": 5690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987964413452569e-05,
+      "loss": 2.956,
+      "step": 5695
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987942901168848e-05,
+      "loss": 2.8273,
+      "step": 5700
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987921369700079e-05,
+      "loss": 2.9638,
+      "step": 5705
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987899819046347e-05,
+      "loss": 2.7584,
+      "step": 5710
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987878249207734e-05,
+      "loss": 2.8056,
+      "step": 5715
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987856660184322e-05,
+      "loss": 2.732,
+      "step": 5720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987835051976197e-05,
+      "loss": 2.7082,
+      "step": 5725
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987813424583439e-05,
+      "loss": 2.7969,
+      "step": 5730
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987791778006132e-05,
+      "loss": 2.711,
+      "step": 5735
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98777011224436e-05,
+      "loss": 2.8703,
+      "step": 5740
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987748427298206e-05,
+      "loss": 2.8266,
+      "step": 5745
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987726723167752e-05,
+      "loss": 2.821,
+      "step": 5750
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987704999853085e-05,
+      "loss": 2.7517,
+      "step": 5755
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987683257354283e-05,
+      "loss": 2.9358,
+      "step": 5760
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987661495671437e-05,
+      "loss": 2.7922,
+      "step": 5765
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987639714804624e-05,
+      "loss": 2.8724,
+      "step": 5770
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987617914753931e-05,
+      "loss": 2.6574,
+      "step": 5775
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987596095519441e-05,
+      "loss": 2.8034,
+      "step": 5780
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987574257101237e-05,
+      "loss": 2.6604,
+      "step": 5785
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987552399499406e-05,
+      "loss": 2.7615,
+      "step": 5790
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98753052271403e-05,
+      "loss": 2.7605,
+      "step": 5795
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987508626745192e-05,
+      "loss": 2.8226,
+      "step": 5800
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98748671159298e-05,
+      "loss": 2.9197,
+      "step": 5805
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987464777257474e-05,
+      "loss": 2.9057,
+      "step": 5810
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987442823738762e-05,
+      "loss": 2.7852,
+      "step": 5815
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987420851036923e-05,
+      "loss": 2.8952,
+      "step": 5820
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987398859152049e-05,
+      "loss": 2.7152,
+      "step": 5825
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98737684808422e-05,
+      "loss": 2.9931,
+      "step": 5830
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987354817833521e-05,
+      "loss": 2.8292,
+      "step": 5835
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987332768400037e-05,
+      "loss": 2.9327,
+      "step": 5840
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987310699783854e-05,
+      "loss": 2.9263,
+      "step": 5845
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987288611985055e-05,
+      "loss": 2.8653,
+      "step": 5850
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987266505003725e-05,
+      "loss": 2.8273,
+      "step": 5855
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987244378839951e-05,
+      "loss": 2.7064,
+      "step": 5860
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987222233493818e-05,
+      "loss": 2.7536,
+      "step": 5865
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987200068965407e-05,
+      "loss": 2.7827,
+      "step": 5870
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987177885254811e-05,
+      "loss": 2.6633,
+      "step": 5875
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987155682362106e-05,
+      "loss": 2.7734,
+      "step": 5880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987133460287386e-05,
+      "loss": 2.8699,
+      "step": 5885
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987111219030729e-05,
+      "loss": 2.8114,
+      "step": 5890
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987088958592226e-05,
+      "loss": 2.8171,
+      "step": 5895
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987066678971961e-05,
+      "loss": 2.8997,
+      "step": 5900
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98704438017002e-05,
+      "loss": 2.6874,
+      "step": 5905
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987022062186487e-05,
+      "loss": 2.8985,
+      "step": 5910
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98699972502145e-05,
+      "loss": 2.8155,
+      "step": 5915
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986977368674994e-05,
+      "loss": 2.7842,
+      "step": 5920
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986954993147204e-05,
+      "loss": 2.7063,
+      "step": 5925
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986932598438168e-05,
+      "loss": 2.8682,
+      "step": 5930
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986910184547971e-05,
+      "loss": 2.7878,
+      "step": 5935
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986887751476697e-05,
+      "loss": 2.8186,
+      "step": 5940
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986865299224437e-05,
+      "loss": 2.7346,
+      "step": 5945
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986842827791277e-05,
+      "loss": 2.8268,
+      "step": 5950
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986820337177298e-05,
+      "loss": 2.892,
+      "step": 5955
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98679782738259e-05,
+      "loss": 2.7434,
+      "step": 5960
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986775298407242e-05,
+      "loss": 2.8549,
+      "step": 5965
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986752750251336e-05,
+      "loss": 2.8694,
+      "step": 5970
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986730182914963e-05,
+      "loss": 2.8681,
+      "step": 5975
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986707596398206e-05,
+      "loss": 2.7144,
+      "step": 5980
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986684990701154e-05,
+      "loss": 2.7741,
+      "step": 5985
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986662365823894e-05,
+      "loss": 2.7203,
+      "step": 5990
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986639721766511e-05,
+      "loss": 2.7426,
+      "step": 5995
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986617058529095e-05,
+      "loss": 2.8361,
+      "step": 6000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986594376111733e-05,
+      "loss": 2.8582,
+      "step": 6005
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98657167451451e-05,
+      "loss": 2.8112,
+      "step": 6010
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986548953737513e-05,
+      "loss": 2.7031,
+      "step": 6015
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986526213780833e-05,
+      "loss": 2.7817,
+      "step": 6020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986503454644553e-05,
+      "loss": 2.8303,
+      "step": 6025
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986480676328765e-05,
+      "loss": 2.8513,
+      "step": 6030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986457878833553e-05,
+      "loss": 2.7668,
+      "step": 6035
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986435062159008e-05,
+      "loss": 2.9309,
+      "step": 6040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986412226305213e-05,
+      "loss": 2.7766,
+      "step": 6045
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98638937127226e-05,
+      "loss": 2.8493,
+      "step": 6050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986366497060236e-05,
+      "loss": 2.8754,
+      "step": 6055
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986343603669228e-05,
+      "loss": 2.7003,
+      "step": 6060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986320691099325e-05,
+      "loss": 2.7519,
+      "step": 6065
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986297759350613e-05,
+      "loss": 2.889,
+      "step": 6070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986274808423185e-05,
+      "loss": 2.8816,
+      "step": 6075
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986251838317124e-05,
+      "loss": 2.8433,
+      "step": 6080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986228849032522e-05,
+      "loss": 2.7282,
+      "step": 6085
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986205840569464e-05,
+      "loss": 2.825,
+      "step": 6090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986182812928043e-05,
+      "loss": 2.8523,
+      "step": 6095
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986159766108342e-05,
+      "loss": 2.7718,
+      "step": 6100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986136700110455e-05,
+      "loss": 2.7181,
+      "step": 6105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986113614934469e-05,
+      "loss": 2.7879,
+      "step": 6110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98609051058047e-05,
+      "loss": 2.8313,
+      "step": 6115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98606738704855e-05,
+      "loss": 2.8279,
+      "step": 6120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.9860442443388e-05,
+      "loss": 2.8139,
+      "step": 6125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986021082451302e-05,
+      "loss": 2.8128,
+      "step": 6130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985997901386151e-05,
+      "loss": 2.7636,
+      "step": 6135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985974701143433e-05,
+      "loss": 2.8263,
+      "step": 6140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985951481723239e-05,
+      "loss": 2.7825,
+      "step": 6145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985928243125657e-05,
+      "loss": 2.8228,
+      "step": 6150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985904985350779e-05,
+      "loss": 2.8734,
+      "step": 6155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985881708398693e-05,
+      "loss": 2.8381,
+      "step": 6160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985858412269487e-05,
+      "loss": 2.7439,
+      "step": 6165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985835096963251e-05,
+      "loss": 2.7752,
+      "step": 6170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985811762480076e-05,
+      "loss": 2.8316,
+      "step": 6175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985788408820052e-05,
+      "loss": 2.8743,
+      "step": 6180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985765035983268e-05,
+      "loss": 2.8061,
+      "step": 6185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985741643969813e-05,
+      "loss": 2.8182,
+      "step": 6190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985718232779779e-05,
+      "loss": 2.8349,
+      "step": 6195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985694802413255e-05,
+      "loss": 2.8048,
+      "step": 6200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98567135287033e-05,
+      "loss": 2.662,
+      "step": 6205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985647884151096e-05,
+      "loss": 2.8308,
+      "step": 6210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985624396255642e-05,
+      "loss": 2.7659,
+      "step": 6215
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985600889184059e-05,
+      "loss": 2.8606,
+      "step": 6220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985577362936439e-05,
+      "loss": 2.8517,
+      "step": 6225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985553817512869e-05,
+      "loss": 2.9058,
+      "step": 6230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985530252913442e-05,
+      "loss": 2.7779,
+      "step": 6235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985506669138248e-05,
+      "loss": 2.7499,
+      "step": 6240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985483066187377e-05,
+      "loss": 2.9396,
+      "step": 6245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98545944406092e-05,
+      "loss": 2.8174,
+      "step": 6250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98543580275897e-05,
+      "loss": 2.8073,
+      "step": 6255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985412142281615e-05,
+      "loss": 2.739,
+      "step": 6260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985388462628948e-05,
+      "loss": 2.5979,
+      "step": 6265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98536476380106e-05,
+      "loss": 2.8452,
+      "step": 6270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98534104579804e-05,
+      "loss": 2.6999,
+      "step": 6275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985317308619981e-05,
+      "loss": 2.8143,
+      "step": 6280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985293552266974e-05,
+      "loss": 2.8288,
+      "step": 6285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985269776739113e-05,
+      "loss": 2.7115,
+      "step": 6290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985245982036484e-05,
+      "loss": 2.7759,
+      "step": 6295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985222168159182e-05,
+      "loss": 2.8901,
+      "step": 6300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985198335107298e-05,
+      "loss": 2.8487,
+      "step": 6305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985174482880923e-05,
+      "loss": 2.8368,
+      "step": 6310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98515061148015e-05,
+      "loss": 2.7925,
+      "step": 6315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985126720905072e-05,
+      "loss": 2.6696,
+      "step": 6320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.985102811155776e-05,
+      "loss": 2.7904,
+      "step": 6325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98507888223236e-05,
+      "loss": 2.8046,
+      "step": 6330
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98505493413491e-05,
+      "loss": 2.8843,
+      "step": 6335
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985030966863523e-05,
+      "loss": 2.8447,
+      "step": 6340
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98500698041829e-05,
+      "loss": 2.7956,
+      "step": 6345
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984982974799302e-05,
+      "loss": 2.8305,
+      "step": 6350
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98495895000665e-05,
+      "loss": 2.6713,
+      "step": 6355
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98493490604043e-05,
+      "loss": 2.8303,
+      "step": 6360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984910842900735e-05,
+      "loss": 2.7845,
+      "step": 6365
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984886760587653e-05,
+      "loss": 2.8152,
+      "step": 6370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98486265910128e-05,
+      "loss": 2.7084,
+      "step": 6375
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984838538441707e-05,
+      "loss": 2.6954,
+      "step": 6380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984814398609027e-05,
+      "loss": 2.8931,
+      "step": 6385
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984790239603334e-05,
+      "loss": 3.0787,
+      "step": 6390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98476606142472e-05,
+      "loss": 2.7968,
+      "step": 6395
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984741864073279e-05,
+      "loss": 2.8593,
+      "step": 6400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984717647549105e-05,
+      "loss": 2.7412,
+      "step": 6405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984693411852285e-05,
+      "loss": 2.8147,
+      "step": 6410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984669156982921e-05,
+      "loss": 2.8581,
+      "step": 6415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9846448829411e-05,
+      "loss": 2.7792,
+      "step": 6420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984620589726918e-05,
+      "loss": 2.7195,
+      "step": 6425
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984596277340468e-05,
+      "loss": 2.8408,
+      "step": 6430
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984571945781843e-05,
+      "loss": 2.6453,
+      "step": 6435
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984547595051136e-05,
+      "loss": 2.8429,
+      "step": 6440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984523225148442e-05,
+      "loss": 2.8216,
+      "step": 6445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984498836073855e-05,
+      "loss": 2.7655,
+      "step": 6450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984474427827469e-05,
+      "loss": 2.758,
+      "step": 6455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984450000409374e-05,
+      "loss": 2.7957,
+      "step": 6460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98442555381967e-05,
+      "loss": 2.8116,
+      "step": 6465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984401088058445e-05,
+      "loss": 2.7318,
+      "step": 6470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984376603125798e-05,
+      "loss": 2.7843,
+      "step": 6475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984352099021822e-05,
+      "loss": 2.7806,
+      "step": 6480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98432757574661e-05,
+      "loss": 2.8854,
+      "step": 6485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984303033300255e-05,
+      "loss": 2.8239,
+      "step": 6490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984278471682856e-05,
+      "loss": 2.8785,
+      "step": 6495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984253890894503e-05,
+      "loss": 2.8698,
+      "step": 6500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984229290935292e-05,
+      "loss": 2.7387,
+      "step": 6505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984204671805318e-05,
+      "loss": 2.7308,
+      "step": 6510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984180033504677e-05,
+      "loss": 2.8854,
+      "step": 6515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984155376033462e-05,
+      "loss": 2.8023,
+      "step": 6520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984130699391765e-05,
+      "loss": 2.701,
+      "step": 6525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984106003579687e-05,
+      "loss": 2.7519,
+      "step": 6530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98408128859732e-05,
+      "loss": 2.7898,
+      "step": 6535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984056554444758e-05,
+      "loss": 2.7744,
+      "step": 6540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984031801122099e-05,
+      "loss": 2.7476,
+      "step": 6545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984007028629437e-05,
+      "loss": 2.6935,
+      "step": 6550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983982236966865e-05,
+      "loss": 2.8136,
+      "step": 6555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983957426134478e-05,
+      "loss": 2.9323,
+      "step": 6560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983932596132377e-05,
+      "loss": 2.8393,
+      "step": 6565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983907746960653e-05,
+      "loss": 2.7665,
+      "step": 6570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983882878619401e-05,
+      "loss": 2.935,
+      "step": 6575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983857991108721e-05,
+      "loss": 2.7636,
+      "step": 6580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983833084428704e-05,
+      "loss": 2.7597,
+      "step": 6585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983808158579449e-05,
+      "loss": 2.8638,
+      "step": 6590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983783213561051e-05,
+      "loss": 2.8441,
+      "step": 6595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983758249373604e-05,
+      "loss": 2.8664,
+      "step": 6600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983733266017206e-05,
+      "loss": 2.6263,
+      "step": 6605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983708263491954e-05,
+      "loss": 2.7744,
+      "step": 6610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98368324179794e-05,
+      "loss": 2.7378,
+      "step": 6615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983658200935267e-05,
+      "loss": 2.7143,
+      "step": 6620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983633140904025e-05,
+      "loss": 2.7948,
+      "step": 6625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983608061704312e-05,
+      "loss": 2.8196,
+      "step": 6630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983582963336229e-05,
+      "loss": 2.7804,
+      "step": 6635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983557845799866e-05,
+      "loss": 2.8283,
+      "step": 6640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983532709095322e-05,
+      "loss": 2.6459,
+      "step": 6645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983507553222695e-05,
+      "loss": 2.6712,
+      "step": 6650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983482378182081e-05,
+      "loss": 2.7955,
+      "step": 6655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983457183973577e-05,
+      "loss": 2.7567,
+      "step": 6660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983431970597279e-05,
+      "loss": 2.7507,
+      "step": 6665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983406738053284e-05,
+      "loss": 2.7352,
+      "step": 6670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983381486341691e-05,
+      "loss": 2.8206,
+      "step": 6675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983356215462595e-05,
+      "loss": 2.7569,
+      "step": 6680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983330925416094e-05,
+      "loss": 2.874,
+      "step": 6685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983305616202286e-05,
+      "loss": 2.9406,
+      "step": 6690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983280287821266e-05,
+      "loss": 2.7569,
+      "step": 6695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983254940273135e-05,
+      "loss": 2.8176,
+      "step": 6700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983229573557987e-05,
+      "loss": 2.7718,
+      "step": 6705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983204187675922e-05,
+      "loss": 2.7796,
+      "step": 6710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983178782627037e-05,
+      "loss": 2.6736,
+      "step": 6715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983153358411428e-05,
+      "loss": 2.8293,
+      "step": 6720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983127915029194e-05,
+      "loss": 2.7381,
+      "step": 6725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983102452480433e-05,
+      "loss": 2.8044,
+      "step": 6730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983076970765244e-05,
+      "loss": 2.8461,
+      "step": 6735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983051469883725e-05,
+      "loss": 2.6839,
+      "step": 6740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983025949835972e-05,
+      "loss": 2.6608,
+      "step": 6745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983000410622084e-05,
+      "loss": 2.8924,
+      "step": 6750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98297485224216e-05,
+      "loss": 2.7382,
+      "step": 6755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982949274696298e-05,
+      "loss": 2.7747,
+      "step": 6760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982923677984594e-05,
+      "loss": 2.8671,
+      "step": 6765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982898062107151e-05,
+      "loss": 2.8684,
+      "step": 6770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982872427064065e-05,
+      "loss": 2.6676,
+      "step": 6775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982846772855434e-05,
+      "loss": 2.8263,
+      "step": 6780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982821099481357e-05,
+      "loss": 2.8551,
+      "step": 6785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982795406941934e-05,
+      "loss": 2.6902,
+      "step": 6790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982769695237264e-05,
+      "loss": 2.7499,
+      "step": 6795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982743964367443e-05,
+      "loss": 2.7783,
+      "step": 6800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982718214332572e-05,
+      "loss": 2.6597,
+      "step": 6805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98269244513275e-05,
+      "loss": 2.948,
+      "step": 6810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982666656768077e-05,
+      "loss": 2.7579,
+      "step": 6815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982640849238651e-05,
+      "loss": 3.0231,
+      "step": 6820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98261502254457e-05,
+      "loss": 2.7033,
+      "step": 6825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982589176685937e-05,
+      "loss": 2.8478,
+      "step": 6830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982563311662849e-05,
+      "loss": 2.7751,
+      "step": 6835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982537427475404e-05,
+      "loss": 2.6541,
+      "step": 6840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982511524123703e-05,
+      "loss": 2.7883,
+      "step": 6845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982485601607847e-05,
+      "loss": 2.8547,
+      "step": 6850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982459659927935e-05,
+      "loss": 2.9133,
+      "step": 6855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982433699084067e-05,
+      "loss": 2.813,
+      "step": 6860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982407719076341e-05,
+      "loss": 2.8166,
+      "step": 6865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982381719904858e-05,
+      "loss": 2.7788,
+      "step": 6870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982355701569718e-05,
+      "loss": 2.7048,
+      "step": 6875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982329664071021e-05,
+      "loss": 2.788,
+      "step": 6880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982303607408869e-05,
+      "loss": 2.7892,
+      "step": 6885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98227753158336e-05,
+      "loss": 2.8298,
+      "step": 6890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982251436594593e-05,
+      "loss": 2.7539,
+      "step": 6895
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982225322442672e-05,
+      "loss": 2.739,
+      "step": 6900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982199189127695e-05,
+      "loss": 2.8282,
+      "step": 6905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982173036649765e-05,
+      "loss": 2.8095,
+      "step": 6910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982146865008978e-05,
+      "loss": 2.7452,
+      "step": 6915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982120674205439e-05,
+      "loss": 2.7682,
+      "step": 6920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982094464239247e-05,
+      "loss": 2.9218,
+      "step": 6925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982068235110504e-05,
+      "loss": 2.8556,
+      "step": 6930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982041986819309e-05,
+      "loss": 2.9368,
+      "step": 6935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982015719365763e-05,
+      "loss": 2.764,
+      "step": 6940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98198943274997e-05,
+      "loss": 2.7065,
+      "step": 6945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981963126972027e-05,
+      "loss": 2.8342,
+      "step": 6950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981936802032037e-05,
+      "loss": 2.7976,
+      "step": 6955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981910457930103e-05,
+      "loss": 2.8591,
+      "step": 6960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981884094666324e-05,
+      "loss": 2.8735,
+      "step": 6965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981857712240802e-05,
+      "loss": 2.7686,
+      "step": 6970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981831310653638e-05,
+      "loss": 2.6974,
+      "step": 6975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981804889904936e-05,
+      "loss": 2.7238,
+      "step": 6980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981778449994794e-05,
+      "loss": 2.8937,
+      "step": 6985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981751990923317e-05,
+      "loss": 2.7149,
+      "step": 6990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981725512690605e-05,
+      "loss": 2.8241,
+      "step": 6995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981699015296759e-05,
+      "loss": 2.7486,
+      "step": 7000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981672498741883e-05,
+      "loss": 2.7287,
+      "step": 7005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981645963026078e-05,
+      "loss": 2.8614,
+      "step": 7010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981619408149445e-05,
+      "loss": 2.7739,
+      "step": 7015
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981592834112087e-05,
+      "loss": 2.6813,
+      "step": 7020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981566240914108e-05,
+      "loss": 2.8132,
+      "step": 7025
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981539628555609e-05,
+      "loss": 2.8269,
+      "step": 7030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981512997036691e-05,
+      "loss": 2.6931,
+      "step": 7035
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981486346357458e-05,
+      "loss": 2.8821,
+      "step": 7040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981459676518012e-05,
+      "loss": 2.7956,
+      "step": 7045
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981432987518456e-05,
+      "loss": 2.7834,
+      "step": 7050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981406279358891e-05,
+      "loss": 2.6263,
+      "step": 7055
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981379552039422e-05,
+      "loss": 2.9494,
+      "step": 7060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981352805560148e-05,
+      "loss": 2.66,
+      "step": 7065
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981326039921178e-05,
+      "loss": 2.7055,
+      "step": 7070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981299255122612e-05,
+      "loss": 2.9361,
+      "step": 7075
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98127245116455e-05,
+      "loss": 2.7641,
+      "step": 7080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981245628047099e-05,
+      "loss": 2.8269,
+      "step": 7085
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981218785770361e-05,
+      "loss": 2.7234,
+      "step": 7090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981191924334439e-05,
+      "loss": 2.695,
+      "step": 7095
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981165043739436e-05,
+      "loss": 2.8599,
+      "step": 7100
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981138143985455e-05,
+      "loss": 2.8147,
+      "step": 7105
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981111225072602e-05,
+      "loss": 2.8655,
+      "step": 7110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981084287000978e-05,
+      "loss": 2.826,
+      "step": 7115
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981057329770688e-05,
+      "loss": 2.5779,
+      "step": 7120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981030353381835e-05,
+      "loss": 2.7884,
+      "step": 7125
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981003357834523e-05,
+      "loss": 2.8299,
+      "step": 7130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980976343128856e-05,
+      "loss": 2.8251,
+      "step": 7135
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980949309264936e-05,
+      "loss": 2.8149,
+      "step": 7140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980922256242871e-05,
+      "loss": 2.7033,
+      "step": 7145
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98089518406276e-05,
+      "loss": 2.6581,
+      "step": 7150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980868092724711e-05,
+      "loss": 2.7978,
+      "step": 7155
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980840982228828e-05,
+      "loss": 2.8518,
+      "step": 7160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980813852575213e-05,
+      "loss": 2.8727,
+      "step": 7165
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980786703763971e-05,
+      "loss": 2.7414,
+      "step": 7170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980759535795208e-05,
+      "loss": 2.8972,
+      "step": 7175
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980732348669029e-05,
+      "loss": 2.7416,
+      "step": 7180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980705142385534e-05,
+      "loss": 2.7288,
+      "step": 7185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980677916944832e-05,
+      "loss": 2.7928,
+      "step": 7190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980650672347027e-05,
+      "loss": 2.6515,
+      "step": 7195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980623408592225e-05,
+      "loss": 2.6703,
+      "step": 7200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980596125680526e-05,
+      "loss": 2.767,
+      "step": 7205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98056882361204e-05,
+      "loss": 2.7611,
+      "step": 7210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980541502386869e-05,
+      "loss": 2.7812,
+      "step": 7215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980514162005119e-05,
+      "loss": 2.7904,
+      "step": 7220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980486802466896e-05,
+      "loss": 2.8194,
+      "step": 7225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980459423772304e-05,
+      "loss": 2.6679,
+      "step": 7230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980432025921449e-05,
+      "loss": 2.8364,
+      "step": 7235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980404608914435e-05,
+      "loss": 2.7747,
+      "step": 7240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980377172751371e-05,
+      "loss": 2.7329,
+      "step": 7245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98034971743236e-05,
+      "loss": 2.8904,
+      "step": 7250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980322242957508e-05,
+      "loss": 2.6765,
+      "step": 7255
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980294749326918e-05,
+      "loss": 2.8286,
+      "step": 7260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9802672365407e-05,
+      "loss": 2.6789,
+      "step": 7265
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980239704598957e-05,
+      "loss": 2.8003,
+      "step": 7270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980212153501798e-05,
+      "loss": 2.6655,
+      "step": 7275
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980184583249324e-05,
+      "loss": 2.7789,
+      "step": 7280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980156993841646e-05,
+      "loss": 2.8328,
+      "step": 7285
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980129385278868e-05,
+      "loss": 2.8467,
+      "step": 7290
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980101757561096e-05,
+      "loss": 2.6602,
+      "step": 7295
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980074110688437e-05,
+      "loss": 2.7692,
+      "step": 7300
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980046444660996e-05,
+      "loss": 2.9474,
+      "step": 7305
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98001875947888e-05,
+      "loss": 2.7529,
+      "step": 7310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979991055142197e-05,
+      "loss": 2.8317,
+      "step": 7315
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97996333165105e-05,
+      "loss": 2.9058,
+      "step": 7320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979935589005551e-05,
+      "loss": 2.8072,
+      "step": 7325
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979907827205802e-05,
+      "loss": 2.6613,
+      "step": 7330
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979880046251913e-05,
+      "loss": 2.7307,
+      "step": 7335
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979852246143987e-05,
+      "loss": 2.7397,
+      "step": 7340
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979824426882134e-05,
+      "loss": 2.7741,
+      "step": 7345
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979796588466462e-05,
+      "loss": 2.78,
+      "step": 7350
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979768730897073e-05,
+      "loss": 2.7167,
+      "step": 7355
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97974085417408e-05,
+      "loss": 2.775,
+      "step": 7360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979712958297587e-05,
+      "loss": 2.6419,
+      "step": 7365
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979685043267702e-05,
+      "loss": 2.7784,
+      "step": 7370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979657109084531e-05,
+      "loss": 2.7199,
+      "step": 7375
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979629155748184e-05,
+      "loss": 2.7147,
+      "step": 7380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979601183258768e-05,
+      "loss": 2.7705,
+      "step": 7385
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97957319161639e-05,
+      "loss": 2.7141,
+      "step": 7390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979545180821155e-05,
+      "loss": 2.7257,
+      "step": 7395
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979517150873175e-05,
+      "loss": 2.7595,
+      "step": 7400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979489101772557e-05,
+      "loss": 2.7603,
+      "step": 7405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979461033519406e-05,
+      "loss": 2.722,
+      "step": 7410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979432946113833e-05,
+      "loss": 2.877,
+      "step": 7415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979404839555947e-05,
+      "loss": 2.6394,
+      "step": 7420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979376713845851e-05,
+      "loss": 2.8442,
+      "step": 7425
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979348568983657e-05,
+      "loss": 2.7678,
+      "step": 7430
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979320404969473e-05,
+      "loss": 2.7037,
+      "step": 7435
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979292221803407e-05,
+      "loss": 2.8303,
+      "step": 7440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979264019485566e-05,
+      "loss": 2.8523,
+      "step": 7445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97923579801606e-05,
+      "loss": 2.6888,
+      "step": 7450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979207557394998e-05,
+      "loss": 2.6967,
+      "step": 7455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979179297622488e-05,
+      "loss": 2.7883,
+      "step": 7460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979151018698637e-05,
+      "loss": 2.6354,
+      "step": 7465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979122720623558e-05,
+      "loss": 2.6013,
+      "step": 7470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979094403397356e-05,
+      "loss": 2.8434,
+      "step": 7475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97906606702014e-05,
+      "loss": 2.6819,
+      "step": 7480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97903771149202e-05,
+      "loss": 2.8499,
+      "step": 7485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979009336813108e-05,
+      "loss": 2.6964,
+      "step": 7490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978980942983508e-05,
+      "loss": 2.8571,
+      "step": 7495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978952530003332e-05,
+      "loss": 2.7745,
+      "step": 7500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978924097872688e-05,
+      "loss": 2.7701,
+      "step": 7505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978895646591687e-05,
+      "loss": 2.7107,
+      "step": 7510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978867176160438e-05,
+      "loss": 2.9252,
+      "step": 7515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97883868657905e-05,
+      "loss": 2.7195,
+      "step": 7520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978810177847631e-05,
+      "loss": 2.7281,
+      "step": 7525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978781649966294e-05,
+      "loss": 2.7156,
+      "step": 7530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978753102935147e-05,
+      "loss": 2.7344,
+      "step": 7535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978724536754299e-05,
+      "loss": 2.8822,
+      "step": 7540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978695951423861e-05,
+      "loss": 2.6729,
+      "step": 7545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978667346943942e-05,
+      "loss": 2.7944,
+      "step": 7550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978638723314655e-05,
+      "loss": 2.7827,
+      "step": 7555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978610080536104e-05,
+      "loss": 2.7547,
+      "step": 7560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978581418608405e-05,
+      "loss": 2.7192,
+      "step": 7565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978552737531668e-05,
+      "loss": 2.7494,
+      "step": 7570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978524037305999e-05,
+      "loss": 2.7936,
+      "step": 7575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978495317931512e-05,
+      "loss": 2.7457,
+      "step": 7580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978466579408316e-05,
+      "loss": 2.7867,
+      "step": 7585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978437821736521e-05,
+      "loss": 2.7819,
+      "step": 7590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978409044916241e-05,
+      "loss": 2.7272,
+      "step": 7595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978380248947582e-05,
+      "loss": 2.6838,
+      "step": 7600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978351433830657e-05,
+      "loss": 2.7736,
+      "step": 7605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978322599565577e-05,
+      "loss": 2.6159,
+      "step": 7610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978293746152453e-05,
+      "loss": 2.7379,
+      "step": 7615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978264873591396e-05,
+      "loss": 2.7041,
+      "step": 7620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978235981882518e-05,
+      "loss": 2.7915,
+      "step": 7625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978207071025926e-05,
+      "loss": 2.7681,
+      "step": 7630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978178141021736e-05,
+      "loss": 2.8786,
+      "step": 7635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978149191870057e-05,
+      "loss": 2.6966,
+      "step": 7640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978120223570999e-05,
+      "loss": 2.5281,
+      "step": 7645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978091236124676e-05,
+      "loss": 2.747,
+      "step": 7650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9780622295312e-05,
+      "loss": 2.8441,
+      "step": 7655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97803320379068e-05,
+      "loss": 2.7288,
+      "step": 7660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978004158903228e-05,
+      "loss": 2.7002,
+      "step": 7665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977975094868957e-05,
+      "loss": 2.9818,
+      "step": 7670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977946011687979e-05,
+      "loss": 2.8064,
+      "step": 7675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977916909360404e-05,
+      "loss": 2.7275,
+      "step": 7680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977887787886347e-05,
+      "loss": 2.7253,
+      "step": 7685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977858647265917e-05,
+      "loss": 2.8026,
+      "step": 7690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977829487499227e-05,
+      "loss": 2.6897,
+      "step": 7695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977800308586389e-05,
+      "loss": 2.649,
+      "step": 7700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977771110527515e-05,
+      "loss": 2.9185,
+      "step": 7705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977741893322718e-05,
+      "loss": 2.6929,
+      "step": 7710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977712656972111e-05,
+      "loss": 2.8429,
+      "step": 7715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977683401475805e-05,
+      "loss": 2.8118,
+      "step": 7720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977654126833915e-05,
+      "loss": 2.9548,
+      "step": 7725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97762483304655e-05,
+      "loss": 2.7734,
+      "step": 7730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977595520113826e-05,
+      "loss": 2.8143,
+      "step": 7735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977566188035854e-05,
+      "loss": 2.7883,
+      "step": 7740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977536836812746e-05,
+      "loss": 2.6673,
+      "step": 7745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977507466444616e-05,
+      "loss": 2.7961,
+      "step": 7750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977478076931577e-05,
+      "loss": 2.7406,
+      "step": 7755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977448668273743e-05,
+      "loss": 2.6792,
+      "step": 7760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977419240471225e-05,
+      "loss": 2.8359,
+      "step": 7765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977389793524137e-05,
+      "loss": 2.6225,
+      "step": 7770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977360327432594e-05,
+      "loss": 2.8644,
+      "step": 7775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977330842196706e-05,
+      "loss": 2.7265,
+      "step": 7780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97730133781659e-05,
+      "loss": 2.5313,
+      "step": 7785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977271814292357e-05,
+      "loss": 2.7859,
+      "step": 7790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977242271624121e-05,
+      "loss": 2.8997,
+      "step": 7795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977212709811996e-05,
+      "loss": 2.6659,
+      "step": 7800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977183128856095e-05,
+      "loss": 2.6089,
+      "step": 7805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977153528756535e-05,
+      "loss": 2.7646,
+      "step": 7810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977123909513425e-05,
+      "loss": 2.7669,
+      "step": 7815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97709427112688e-05,
+      "loss": 2.76,
+      "step": 7820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977064613597018e-05,
+      "loss": 2.8244,
+      "step": 7825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97703493692395e-05,
+      "loss": 2.7524,
+      "step": 7830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977005241107788e-05,
+      "loss": 2.7324,
+      "step": 7835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97697552614865e-05,
+      "loss": 2.7447,
+      "step": 7840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97694579204665e-05,
+      "loss": 2.8164,
+      "step": 7845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9769160388019e-05,
+      "loss": 2.773,
+      "step": 7850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976886266414515e-05,
+      "loss": 2.7966,
+      "step": 7855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976856474884613e-05,
+      "loss": 2.6705,
+      "step": 7860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976826664212304e-05,
+      "loss": 2.7633,
+      "step": 7865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976796834397703e-05,
+      "loss": 2.8182,
+      "step": 7870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976766985440928e-05,
+      "loss": 2.6807,
+      "step": 7875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976737117342092e-05,
+      "loss": 2.7505,
+      "step": 7880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97670723010131e-05,
+      "loss": 2.8728,
+      "step": 7885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976677323718696e-05,
+      "loss": 2.7521,
+      "step": 7890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976647398194367e-05,
+      "loss": 2.851,
+      "step": 7895
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976617453528436e-05,
+      "loss": 2.7504,
+      "step": 7900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976587489721018e-05,
+      "loss": 2.7438,
+      "step": 7905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976557506772233e-05,
+      "loss": 2.6225,
+      "step": 7910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97652750468219e-05,
+      "loss": 2.7815,
+      "step": 7915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976497483451007e-05,
+      "loss": 2.8966,
+      "step": 7920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976467443078801e-05,
+      "loss": 2.7337,
+      "step": 7925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976437383565685e-05,
+      "loss": 2.8188,
+      "step": 7930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976407304911777e-05,
+      "loss": 2.851,
+      "step": 7935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976377207117189e-05,
+      "loss": 2.8286,
+      "step": 7940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976347090182041e-05,
+      "loss": 2.716,
+      "step": 7945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976316954106447e-05,
+      "loss": 2.8107,
+      "step": 7950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976286798890522e-05,
+      "loss": 2.6096,
+      "step": 7955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976256624534385e-05,
+      "loss": 2.7938,
+      "step": 7960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97622643103815e-05,
+      "loss": 2.759,
+      "step": 7965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976196218401929e-05,
+      "loss": 2.6423,
+      "step": 7970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976165986625847e-05,
+      "loss": 2.7999,
+      "step": 7975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976135735710014e-05,
+      "loss": 2.6729,
+      "step": 7980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976105465654548e-05,
+      "loss": 2.7843,
+      "step": 7985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976075176459565e-05,
+      "loss": 2.8157,
+      "step": 7990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976044868125181e-05,
+      "loss": 2.7957,
+      "step": 7995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976014540651515e-05,
+      "loss": 2.6724,
+      "step": 8000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975984194038683e-05,
+      "loss": 2.7199,
+      "step": 8005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975953828286799e-05,
+      "loss": 2.702,
+      "step": 8010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975923443395983e-05,
+      "loss": 2.6947,
+      "step": 8015
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975893039366348e-05,
+      "loss": 2.7252,
+      "step": 8020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975862616198017e-05,
+      "loss": 2.7863,
+      "step": 8025
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975832173891102e-05,
+      "loss": 2.8088,
+      "step": 8030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975801712445721e-05,
+      "loss": 2.6873,
+      "step": 8035
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975771231861991e-05,
+      "loss": 2.6176,
+      "step": 8040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975740732140031e-05,
+      "loss": 2.6905,
+      "step": 8045
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975710213279957e-05,
+      "loss": 2.773,
+      "step": 8050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975679675281887e-05,
+      "loss": 2.7945,
+      "step": 8055
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975649118145938e-05,
+      "loss": 2.8039,
+      "step": 8060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975618541872229e-05,
+      "loss": 2.7062,
+      "step": 8065
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975587946460873e-05,
+      "loss": 2.8862,
+      "step": 8070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975557331911993e-05,
+      "loss": 2.6891,
+      "step": 8075
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975526698225706e-05,
+      "loss": 2.7986,
+      "step": 8080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975496045402127e-05,
+      "loss": 2.8028,
+      "step": 8085
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975465373441375e-05,
+      "loss": 2.7511,
+      "step": 8090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975434682343569e-05,
+      "loss": 2.6998,
+      "step": 8095
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975403972108826e-05,
+      "loss": 2.6534,
+      "step": 8100
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975373242737265e-05,
+      "loss": 2.6847,
+      "step": 8105
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975342494229003e-05,
+      "loss": 2.7878,
+      "step": 8110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97531172658416e-05,
+      "loss": 2.739,
+      "step": 8115
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975280939802853e-05,
+      "loss": 2.6376,
+      "step": 8120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9752501338852e-05,
+      "loss": 2.7675,
+      "step": 8125
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97521930883132e-05,
+      "loss": 2.8076,
+      "step": 8130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975188464641334e-05,
+      "loss": 2.7626,
+      "step": 8135
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975157601315357e-05,
+      "loss": 2.7383,
+      "step": 8140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975126718853509e-05,
+      "loss": 2.7438,
+      "step": 8145
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975095817255909e-05,
+      "loss": 2.7598,
+      "step": 8150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975064896522677e-05,
+      "loss": 2.8791,
+      "step": 8155
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97503395665393e-05,
+      "loss": 2.5041,
+      "step": 8160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975002997649787e-05,
+      "loss": 2.7076,
+      "step": 8165
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974972019510369e-05,
+      "loss": 2.8557,
+      "step": 8170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974941022235794e-05,
+      "loss": 2.7808,
+      "step": 8175
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974910005826179e-05,
+      "loss": 2.768,
+      "step": 8180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974878970281647e-05,
+      "loss": 2.7465,
+      "step": 8185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974847915602316e-05,
+      "loss": 2.8433,
+      "step": 8190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974816841788304e-05,
+      "loss": 2.7752,
+      "step": 8195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974785748839735e-05,
+      "loss": 2.8901,
+      "step": 8200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974754636756723e-05,
+      "loss": 2.8333,
+      "step": 8205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97472350553939e-05,
+      "loss": 2.8102,
+      "step": 8210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974692355187856e-05,
+      "loss": 2.9493,
+      "step": 8215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974661185702241e-05,
+      "loss": 2.9019,
+      "step": 8220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974629997082665e-05,
+      "loss": 2.6015,
+      "step": 8225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974598789329245e-05,
+      "loss": 2.6117,
+      "step": 8230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974567562442106e-05,
+      "loss": 2.8456,
+      "step": 8235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974536316421366e-05,
+      "loss": 2.8703,
+      "step": 8240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974505051267144e-05,
+      "loss": 2.7245,
+      "step": 8245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97447376697956e-05,
+      "loss": 2.7939,
+      "step": 8250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974442463558737e-05,
+      "loss": 2.7399,
+      "step": 8255
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974411141004792e-05,
+      "loss": 2.7543,
+      "step": 8260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974379799317849e-05,
+      "loss": 2.6182,
+      "step": 8265
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974348438498026e-05,
+      "loss": 2.6448,
+      "step": 8270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974317058545444e-05,
+      "loss": 2.5776,
+      "step": 8275
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974285659460225e-05,
+      "loss": 2.7298,
+      "step": 8280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974254241242489e-05,
+      "loss": 2.7083,
+      "step": 8285
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974222803892355e-05,
+      "loss": 2.7343,
+      "step": 8290
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974191347409948e-05,
+      "loss": 2.6576,
+      "step": 8295
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974159871795386e-05,
+      "loss": 2.7388,
+      "step": 8300
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974128377048791e-05,
+      "loss": 2.7423,
+      "step": 8305
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974096863170282e-05,
+      "loss": 2.6652,
+      "step": 8310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974065330159983e-05,
+      "loss": 2.7201,
+      "step": 8315
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974033778018015e-05,
+      "loss": 2.816,
+      "step": 8320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974002206744498e-05,
+      "loss": 2.6702,
+      "step": 8325
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973970616339554e-05,
+      "loss": 2.861,
+      "step": 8330
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973939006803305e-05,
+      "loss": 2.7562,
+      "step": 8335
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973907378135872e-05,
+      "loss": 2.7119,
+      "step": 8340
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973875730337378e-05,
+      "loss": 2.6695,
+      "step": 8345
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973844063407942e-05,
+      "loss": 2.7468,
+      "step": 8350
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973812377347689e-05,
+      "loss": 2.7735,
+      "step": 8355
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973780672156738e-05,
+      "loss": 2.9197,
+      "step": 8360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973748947835213e-05,
+      "loss": 2.6462,
+      "step": 8365
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973717204383234e-05,
+      "loss": 2.7048,
+      "step": 8370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973685441800927e-05,
+      "loss": 2.7262,
+      "step": 8375
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973653660088408e-05,
+      "loss": 2.7421,
+      "step": 8380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973621859245805e-05,
+      "loss": 2.7305,
+      "step": 8385
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973590039273237e-05,
+      "loss": 2.71,
+      "step": 8390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973558200170828e-05,
+      "loss": 2.7642,
+      "step": 8395
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9735263419387e-05,
+      "loss": 2.6818,
+      "step": 8400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973494464576976e-05,
+      "loss": 2.7206,
+      "step": 8405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973462568085777e-05,
+      "loss": 2.7294,
+      "step": 8410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973430652465228e-05,
+      "loss": 2.8318,
+      "step": 8415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973398717715449e-05,
+      "loss": 2.7476,
+      "step": 8420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973366763836564e-05,
+      "loss": 2.6693,
+      "step": 8425
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973334790828698e-05,
+      "loss": 2.7042,
+      "step": 8430
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973302798691971e-05,
+      "loss": 2.8042,
+      "step": 8435
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973270787426508e-05,
+      "loss": 2.6981,
+      "step": 8440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97323875703243e-05,
+      "loss": 2.689,
+      "step": 8445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973206707509864e-05,
+      "loss": 2.7409,
+      "step": 8450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973174638858929e-05,
+      "loss": 2.8471,
+      "step": 8455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973142551079751e-05,
+      "loss": 2.7601,
+      "step": 8460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973110444172453e-05,
+      "loss": 2.6501,
+      "step": 8465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973078318137157e-05,
+      "loss": 2.889,
+      "step": 8470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973046172973987e-05,
+      "loss": 2.6667,
+      "step": 8475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97301400868307e-05,
+      "loss": 2.7757,
+      "step": 8480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972981825264524e-05,
+      "loss": 2.7179,
+      "step": 8485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972949622718477e-05,
+      "loss": 2.7449,
+      "step": 8490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972917401045052e-05,
+      "loss": 2.7027,
+      "step": 8495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972885160244373e-05,
+      "loss": 2.8957,
+      "step": 8500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972852900316563e-05,
+      "loss": 2.8922,
+      "step": 8505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972820621261748e-05,
+      "loss": 2.8798,
+      "step": 8510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972788323080048e-05,
+      "loss": 2.6278,
+      "step": 8515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972756005771593e-05,
+      "loss": 2.7819,
+      "step": 8520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972723669336502e-05,
+      "loss": 2.727,
+      "step": 8525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972691313774904e-05,
+      "loss": 2.7114,
+      "step": 8530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97265893908692e-05,
+      "loss": 2.7442,
+      "step": 8535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972626545272675e-05,
+      "loss": 2.8394,
+      "step": 8540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972594132332296e-05,
+      "loss": 2.7417,
+      "step": 8545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972561700265905e-05,
+      "loss": 2.7215,
+      "step": 8550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972529249073627e-05,
+      "loss": 2.8063,
+      "step": 8555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972496778755589e-05,
+      "loss": 2.797,
+      "step": 8560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972464289311913e-05,
+      "loss": 2.7027,
+      "step": 8565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972431780742728e-05,
+      "loss": 2.7375,
+      "step": 8570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972399253048153e-05,
+      "loss": 2.6725,
+      "step": 8575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972366706228318e-05,
+      "loss": 2.8414,
+      "step": 8580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972334140283349e-05,
+      "loss": 2.7553,
+      "step": 8585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972301555213366e-05,
+      "loss": 2.6955,
+      "step": 8590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972268951018497e-05,
+      "loss": 2.815,
+      "step": 8595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97223632769887e-05,
+      "loss": 2.8269,
+      "step": 8600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972203685254607e-05,
+      "loss": 2.6215,
+      "step": 8605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972171023685834e-05,
+      "loss": 2.7601,
+      "step": 8610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97213834299268e-05,
+      "loss": 2.7627,
+      "step": 8615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972105643175267e-05,
+      "loss": 2.7009,
+      "step": 8620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972072924233721e-05,
+      "loss": 2.7485,
+      "step": 8625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972040186168168e-05,
+      "loss": 2.7013,
+      "step": 8630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972007428978736e-05,
+      "loss": 2.7858,
+      "step": 8635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97197465266555e-05,
+      "loss": 2.7283,
+      "step": 8640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971941857228734e-05,
+      "loss": 2.8285,
+      "step": 8645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971909042668416e-05,
+      "loss": 2.6372,
+      "step": 8650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971876208984724e-05,
+      "loss": 2.833,
+      "step": 8655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971843356177781e-05,
+      "loss": 2.7983,
+      "step": 8660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971810484247714e-05,
+      "loss": 2.7107,
+      "step": 8665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971777593194652e-05,
+      "loss": 2.7681,
+      "step": 8670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971744683018718e-05,
+      "loss": 2.7518,
+      "step": 8675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971711753720041e-05,
+      "loss": 2.7327,
+      "step": 8680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971678805298748e-05,
+      "loss": 2.744,
+      "step": 8685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971645837754963e-05,
+      "loss": 2.8235,
+      "step": 8690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971612851088813e-05,
+      "loss": 2.7804,
+      "step": 8695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971579845300429e-05,
+      "loss": 2.6345,
+      "step": 8700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971546820389935e-05,
+      "loss": 2.701,
+      "step": 8705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971513776357457e-05,
+      "loss": 2.7576,
+      "step": 8710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971480713203123e-05,
+      "loss": 2.6059,
+      "step": 8715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971447630927062e-05,
+      "loss": 2.7702,
+      "step": 8720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971414529529399e-05,
+      "loss": 2.6823,
+      "step": 8725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971381409010262e-05,
+      "loss": 2.866,
+      "step": 8730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97134826936978e-05,
+      "loss": 2.7563,
+      "step": 8735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971315110608077e-05,
+      "loss": 2.7169,
+      "step": 8740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971281932725282e-05,
+      "loss": 2.8211,
+      "step": 8745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971248735721525e-05,
+      "loss": 2.7615,
+      "step": 8750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97121551959693e-05,
+      "loss": 2.5767,
+      "step": 8755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971182284351627e-05,
+      "loss": 2.8347,
+      "step": 8760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971149029985744e-05,
+      "loss": 2.6051,
+      "step": 8765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971115756499409e-05,
+      "loss": 2.5635,
+      "step": 8770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971082463892748e-05,
+      "loss": 2.7491,
+      "step": 8775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97104915216589e-05,
+      "loss": 2.8358,
+      "step": 8780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971015821318965e-05,
+      "loss": 2.7544,
+      "step": 8785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970982471352099e-05,
+      "loss": 2.7324,
+      "step": 8790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97094910226542e-05,
+      "loss": 2.817,
+      "step": 8795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970915714059057e-05,
+      "loss": 2.7277,
+      "step": 8800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97088230673314e-05,
+      "loss": 2.7821,
+      "step": 8805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970848880287795e-05,
+      "loss": 2.7922,
+      "step": 8810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970815434723153e-05,
+      "loss": 2.8458,
+      "step": 8815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97078197003934e-05,
+      "loss": 2.7961,
+      "step": 8820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970748486236487e-05,
+      "loss": 2.8634,
+      "step": 8825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970714983314722e-05,
+      "loss": 2.8369,
+      "step": 8830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970681461274173e-05,
+      "loss": 2.7676,
+      "step": 8835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97064792011497e-05,
+      "loss": 2.6619,
+      "step": 8840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970614359837242e-05,
+      "loss": 2.6789,
+      "step": 8845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970580780441118e-05,
+      "loss": 2.773,
+      "step": 8850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970547181926726e-05,
+      "loss": 2.7638,
+      "step": 8855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970513564294197e-05,
+      "loss": 2.6253,
+      "step": 8860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970479927543659e-05,
+      "loss": 2.7437,
+      "step": 8865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970446271675241e-05,
+      "loss": 2.7738,
+      "step": 8870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970412596689073e-05,
+      "loss": 2.7674,
+      "step": 8875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970378902585286e-05,
+      "loss": 2.7013,
+      "step": 8880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970345189364009e-05,
+      "loss": 2.883,
+      "step": 8885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.97031145702537e-05,
+      "loss": 2.6271,
+      "step": 8890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970277705569502e-05,
+      "loss": 2.6555,
+      "step": 8895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.97024393499653e-05,
+      "loss": 2.7843,
+      "step": 8900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970210145306587e-05,
+      "loss": 2.7857,
+      "step": 8905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970176336499803e-05,
+      "loss": 2.7543,
+      "step": 8910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970142508576308e-05,
+      "loss": 2.6471,
+      "step": 8915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.97010866153623e-05,
+      "loss": 2.7137,
+      "step": 8920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970074795379703e-05,
+      "loss": 2.5899,
+      "step": 8925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970040910106853e-05,
+      "loss": 2.7889,
+      "step": 8930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970007005717813e-05,
+      "loss": 2.6622,
+      "step": 8935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969973082212714e-05,
+      "loss": 2.8376,
+      "step": 8940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969939139591684e-05,
+      "loss": 2.7964,
+      "step": 8945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969905177854855e-05,
+      "loss": 2.7314,
+      "step": 8950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969871197002358e-05,
+      "loss": 2.7939,
+      "step": 8955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969837197034322e-05,
+      "loss": 2.6925,
+      "step": 8960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96980317795088e-05,
+      "loss": 2.7552,
+      "step": 8965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969769139752161e-05,
+      "loss": 2.7248,
+      "step": 8970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969735082438299e-05,
+      "loss": 2.9024,
+      "step": 8975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96970100600942e-05,
+      "loss": 2.7979,
+      "step": 8980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969666910465658e-05,
+      "loss": 2.8161,
+      "step": 8985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969632795807146e-05,
+      "loss": 2.963,
+      "step": 8990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969598662034011e-05,
+      "loss": 2.6274,
+      "step": 8995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969564509146388e-05,
+      "loss": 2.6592,
+      "step": 9000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969530337144405e-05,
+      "loss": 2.708,
+      "step": 9005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969496146028197e-05,
+      "loss": 2.7075,
+      "step": 9010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969461935797892e-05,
+      "loss": 2.7051,
+      "step": 9015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969427706453624e-05,
+      "loss": 2.7163,
+      "step": 9020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969393457995526e-05,
+      "loss": 2.7222,
+      "step": 9025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969359190423726e-05,
+      "loss": 2.6826,
+      "step": 9030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969324903738358e-05,
+      "loss": 2.7686,
+      "step": 9035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969290597939552e-05,
+      "loss": 2.8519,
+      "step": 9040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969256273027441e-05,
+      "loss": 2.7432,
+      "step": 9045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969221929002159e-05,
+      "loss": 2.7851,
+      "step": 9050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969187565863835e-05,
+      "loss": 2.6111,
+      "step": 9055
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969153183612605e-05,
+      "loss": 2.8532,
+      "step": 9060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969118782248596e-05,
+      "loss": 2.7831,
+      "step": 9065
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969084361771945e-05,
+      "loss": 2.8052,
+      "step": 9070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969049922182783e-05,
+      "loss": 2.6997,
+      "step": 9075
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969015463481242e-05,
+      "loss": 2.5938,
+      "step": 9080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968980985667453e-05,
+      "loss": 2.6756,
+      "step": 9085
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968946488741551e-05,
+      "loss": 2.8168,
+      "step": 9090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968911972703668e-05,
+      "loss": 2.7356,
+      "step": 9095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968877437553936e-05,
+      "loss": 2.9008,
+      "step": 9100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968842883292488e-05,
+      "loss": 2.6981,
+      "step": 9105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96880830991946e-05,
+      "loss": 2.7956,
+      "step": 9110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96877371743498e-05,
+      "loss": 2.8175,
+      "step": 9115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968739105839184e-05,
+      "loss": 2.7666,
+      "step": 9120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968704475132204e-05,
+      "loss": 2.6787,
+      "step": 9125
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968669825314174e-05,
+      "loss": 2.72,
+      "step": 9130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968635156385228e-05,
+      "loss": 2.819,
+      "step": 9135
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968600468345497e-05,
+      "loss": 2.7238,
+      "step": 9140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968565761195117e-05,
+      "loss": 2.6663,
+      "step": 9145
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968531034934219e-05,
+      "loss": 2.6077,
+      "step": 9150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968496289562939e-05,
+      "loss": 2.8301,
+      "step": 9155
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968461525081408e-05,
+      "loss": 2.8211,
+      "step": 9160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968426741489763e-05,
+      "loss": 2.7014,
+      "step": 9165
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968391938788135e-05,
+      "loss": 2.7231,
+      "step": 9170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968357116976658e-05,
+      "loss": 2.6316,
+      "step": 9175
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968322276055469e-05,
+      "loss": 2.7295,
+      "step": 9180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968287416024698e-05,
+      "loss": 2.5682,
+      "step": 9185
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968252536884481e-05,
+      "loss": 2.7883,
+      "step": 9190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968217638634951e-05,
+      "loss": 2.8621,
+      "step": 9195
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968182721276245e-05,
+      "loss": 2.814,
+      "step": 9200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968147784808495e-05,
+      "loss": 2.5237,
+      "step": 9205
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968112829231836e-05,
+      "loss": 2.5476,
+      "step": 9210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968077854546403e-05,
+      "loss": 2.7912,
+      "step": 9215
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968042860752327e-05,
+      "loss": 2.6696,
+      "step": 9220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968007847849748e-05,
+      "loss": 2.728,
+      "step": 9225
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967972815838799e-05,
+      "loss": 2.7157,
+      "step": 9230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967937764719613e-05,
+      "loss": 2.6091,
+      "step": 9235
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967902694492324e-05,
+      "loss": 2.7123,
+      "step": 9240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967867605157071e-05,
+      "loss": 2.7467,
+      "step": 9245
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967832496713984e-05,
+      "loss": 2.7226,
+      "step": 9250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967797369163203e-05,
+      "loss": 2.8311,
+      "step": 9255
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96776222250486e-05,
+      "loss": 2.6409,
+      "step": 9260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96772705673909e-05,
+      "loss": 2.567,
+      "step": 9265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96769187186603e-05,
+      "loss": 2.622,
+      "step": 9270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967656667885813e-05,
+      "loss": 2.6299,
+      "step": 9275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967621444798579e-05,
+      "loss": 2.8023,
+      "step": 9280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967586202604458e-05,
+      "loss": 2.783,
+      "step": 9285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967550941303588e-05,
+      "loss": 2.8256,
+      "step": 9290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967515660896105e-05,
+      "loss": 2.6807,
+      "step": 9295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967480361382144e-05,
+      "loss": 2.6069,
+      "step": 9300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967445042761842e-05,
+      "loss": 2.6059,
+      "step": 9305
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967409705035333e-05,
+      "loss": 2.7283,
+      "step": 9310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967374348202754e-05,
+      "loss": 2.6578,
+      "step": 9315
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96733897226424e-05,
+      "loss": 2.7013,
+      "step": 9320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967303577219928e-05,
+      "loss": 2.6136,
+      "step": 9325
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967268163069955e-05,
+      "loss": 2.6504,
+      "step": 9330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967232729814455e-05,
+      "loss": 2.7655,
+      "step": 9335
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967197277453566e-05,
+      "loss": 2.753,
+      "step": 9340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967161805987424e-05,
+      "loss": 2.8895,
+      "step": 9345
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967126315416166e-05,
+      "loss": 2.7153,
+      "step": 9350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967090805739927e-05,
+      "loss": 2.7616,
+      "step": 9355
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967055276958843e-05,
+      "loss": 2.7381,
+      "step": 9360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967019729073053e-05,
+      "loss": 2.8133,
+      "step": 9365
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966984162082693e-05,
+      "loss": 2.7486,
+      "step": 9370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9669485759879e-05,
+      "loss": 2.6635,
+      "step": 9375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96691297078881e-05,
+      "loss": 2.6954,
+      "step": 9380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966877346485561e-05,
+      "loss": 2.7702,
+      "step": 9385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96684170307829e-05,
+      "loss": 2.8112,
+      "step": 9390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96680604056713e-05,
+      "loss": 2.717,
+      "step": 9395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966770358952225e-05,
+      "loss": 2.5509,
+      "step": 9400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966734658233707e-05,
+      "loss": 2.7365,
+      "step": 9405
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966698938411716e-05,
+      "loss": 2.8325,
+      "step": 9410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966663199486388e-05,
+      "loss": 2.6549,
+      "step": 9415
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96662744145786e-05,
+      "loss": 2.8812,
+      "step": 9420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966591664326273e-05,
+      "loss": 2.8024,
+      "step": 9425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966555868091762e-05,
+      "loss": 2.5811,
+      "step": 9430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966520052754464e-05,
+      "loss": 2.6442,
+      "step": 9435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966484218314518e-05,
+      "loss": 2.7737,
+      "step": 9440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966448364772061e-05,
+      "loss": 2.6186,
+      "step": 9445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966412492127231e-05,
+      "loss": 2.8526,
+      "step": 9450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966376600380167e-05,
+      "loss": 2.8554,
+      "step": 9455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966340689531006e-05,
+      "loss": 2.6386,
+      "step": 9460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966304759579888e-05,
+      "loss": 2.8365,
+      "step": 9465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966268810526948e-05,
+      "loss": 2.6676,
+      "step": 9470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966232842372327e-05,
+      "loss": 2.6523,
+      "step": 9475
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966196855116162e-05,
+      "loss": 2.6994,
+      "step": 9480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966160848758591e-05,
+      "loss": 2.6937,
+      "step": 9485
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966124823299755e-05,
+      "loss": 2.6733,
+      "step": 9490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966088778739788e-05,
+      "loss": 2.8202,
+      "step": 9495
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966052715078833e-05,
+      "loss": 2.6985,
+      "step": 9500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966016632317027e-05,
+      "loss": 2.7335,
+      "step": 9505
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96598053045451e-05,
+      "loss": 2.6137,
+      "step": 9510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965944409491419e-05,
+      "loss": 2.6508,
+      "step": 9515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965908269427893e-05,
+      "loss": 2.7897,
+      "step": 9520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965872110264072e-05,
+      "loss": 2.6543,
+      "step": 9525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965835932000096e-05,
+      "loss": 2.6145,
+      "step": 9530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965799734636101e-05,
+      "loss": 2.7802,
+      "step": 9535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96576351817223e-05,
+      "loss": 2.7002,
+      "step": 9540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965727282608618e-05,
+      "loss": 2.6114,
+      "step": 9545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965691027945409e-05,
+      "loss": 2.7739,
+      "step": 9550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96565475418274e-05,
+      "loss": 2.6899,
+      "step": 9555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965618461320752e-05,
+      "loss": 2.754,
+      "step": 9560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965582149359581e-05,
+      "loss": 2.6879,
+      "step": 9565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965545818299371e-05,
+      "loss": 2.5683,
+      "step": 9570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965509468140258e-05,
+      "loss": 2.6002,
+      "step": 9575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965473098882385e-05,
+      "loss": 2.8101,
+      "step": 9580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96543671052589e-05,
+      "loss": 2.8493,
+      "step": 9585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965400303070913e-05,
+      "loss": 2.7861,
+      "step": 9590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965363876517596e-05,
+      "loss": 2.8174,
+      "step": 9595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965327430866076e-05,
+      "loss": 2.713,
+      "step": 9600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965290966116496e-05,
+      "loss": 2.7594,
+      "step": 9605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965254482268995e-05,
+      "loss": 2.7182,
+      "step": 9610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965217979323712e-05,
+      "loss": 2.7907,
+      "step": 9615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96518145728079e-05,
+      "loss": 2.6688,
+      "step": 9620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965144916140368e-05,
+      "loss": 2.6594,
+      "step": 9625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965108355902585e-05,
+      "loss": 2.7953,
+      "step": 9630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965071776567586e-05,
+      "loss": 2.6958,
+      "step": 9635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965035178135509e-05,
+      "loss": 2.6201,
+      "step": 9640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964998560606494e-05,
+      "loss": 2.8499,
+      "step": 9645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964961923980682e-05,
+      "loss": 2.7356,
+      "step": 9650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964925268258217e-05,
+      "loss": 2.6643,
+      "step": 9655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964888593439234e-05,
+      "loss": 2.6507,
+      "step": 9660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96485189952388e-05,
+      "loss": 2.7244,
+      "step": 9665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964815186512294e-05,
+      "loss": 2.7826,
+      "step": 9670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964778454404617e-05,
+      "loss": 2.7458,
+      "step": 9675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96474170320099e-05,
+      "loss": 2.6944,
+      "step": 9680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964704932901554e-05,
+      "loss": 2.6276,
+      "step": 9685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964668143506452e-05,
+      "loss": 2.8949,
+      "step": 9690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964631335015824e-05,
+      "loss": 2.7763,
+      "step": 9695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964594507429812e-05,
+      "loss": 2.7839,
+      "step": 9700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96455766074856e-05,
+      "loss": 2.773,
+      "step": 9705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964520794972206e-05,
+      "loss": 2.692,
+      "step": 9710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964483910100893e-05,
+      "loss": 2.7594,
+      "step": 9715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964447006134763e-05,
+      "loss": 2.7045,
+      "step": 9720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964410083073957e-05,
+      "loss": 2.6579,
+      "step": 9725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964373140918621e-05,
+      "loss": 2.8329,
+      "step": 9730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964336179668892e-05,
+      "loss": 2.6432,
+      "step": 9735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964299199324916e-05,
+      "loss": 2.6067,
+      "step": 9740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964262199886832e-05,
+      "loss": 2.8002,
+      "step": 9745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964225181354784e-05,
+      "loss": 2.7034,
+      "step": 9750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964188143728917e-05,
+      "loss": 2.6698,
+      "step": 9755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96415108700937e-05,
+      "loss": 2.5826,
+      "step": 9760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964114011196283e-05,
+      "loss": 2.6812,
+      "step": 9765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964076916289806e-05,
+      "loss": 2.7138,
+      "step": 9770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964039802290076e-05,
+      "loss": 2.7348,
+      "step": 9775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964002669197238e-05,
+      "loss": 2.7483,
+      "step": 9780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963965517011433e-05,
+      "loss": 2.677,
+      "step": 9785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963928345732806e-05,
+      "loss": 2.7186,
+      "step": 9790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963891155361499e-05,
+      "loss": 2.7572,
+      "step": 9795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963853945897654e-05,
+      "loss": 2.7483,
+      "step": 9800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963816717341416e-05,
+      "loss": 2.6169,
+      "step": 9805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963779469692928e-05,
+      "loss": 2.7635,
+      "step": 9810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963742202952333e-05,
+      "loss": 2.8046,
+      "step": 9815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963704917119773e-05,
+      "loss": 2.6634,
+      "step": 9820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963667612195394e-05,
+      "loss": 2.7571,
+      "step": 9825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963630288179336e-05,
+      "loss": 2.7344,
+      "step": 9830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963592945071744e-05,
+      "loss": 2.7327,
+      "step": 9835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963555582872764e-05,
+      "loss": 2.8704,
+      "step": 9840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963518201582538e-05,
+      "loss": 2.7132,
+      "step": 9845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963480801201208e-05,
+      "loss": 2.6211,
+      "step": 9850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963443381728921e-05,
+      "loss": 2.6429,
+      "step": 9855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963405943165818e-05,
+      "loss": 2.7973,
+      "step": 9860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963368485512045e-05,
+      "loss": 2.6595,
+      "step": 9865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963331008767745e-05,
+      "loss": 2.7968,
+      "step": 9870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963293512933063e-05,
+      "loss": 2.6661,
+      "step": 9875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963255998008143e-05,
+      "loss": 2.7361,
+      "step": 9880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96321846399313e-05,
+      "loss": 2.7985,
+      "step": 9885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963180910888166e-05,
+      "loss": 2.5134,
+      "step": 9890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963143338693397e-05,
+      "loss": 2.7024,
+      "step": 9895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963105747408968e-05,
+      "loss": 2.7876,
+      "step": 9900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963068137035022e-05,
+      "loss": 2.7551,
+      "step": 9905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963030507571706e-05,
+      "loss": 2.7203,
+      "step": 9910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962992859019163e-05,
+      "loss": 2.7637,
+      "step": 9915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962955191377536e-05,
+      "loss": 2.7966,
+      "step": 9920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962917504646976e-05,
+      "loss": 2.7417,
+      "step": 9925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962879798827622e-05,
+      "loss": 2.8136,
+      "step": 9930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96284207391962e-05,
+      "loss": 2.7484,
+      "step": 9935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962804329923119e-05,
+      "loss": 2.7679,
+      "step": 9940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96276656683826e-05,
+      "loss": 2.7065,
+      "step": 9945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96272878466519e-05,
+      "loss": 2.7535,
+      "step": 9950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962690983404053e-05,
+      "loss": 2.7052,
+      "step": 9955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962653163054995e-05,
+      "loss": 2.8289,
+      "step": 9960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962615323618164e-05,
+      "loss": 2.6596,
+      "step": 9965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962577465093699e-05,
+      "loss": 2.8246,
+      "step": 9970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962539587481754e-05,
+      "loss": 2.7395,
+      "step": 9975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962501690782469e-05,
+      "loss": 2.7416,
+      "step": 9980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962463774995992e-05,
+      "loss": 2.8,
+      "step": 9985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962425840122467e-05,
+      "loss": 2.5172,
+      "step": 9990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962387886162043e-05,
+      "loss": 2.7187,
+      "step": 9995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962349913114864e-05,
+      "loss": 2.7694,
+      "step": 10000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962311920981074e-05,
+      "loss": 2.7254,
+      "step": 10005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962273909760823e-05,
+      "loss": 2.6156,
+      "step": 10010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962235879454255e-05,
+      "loss": 2.8003,
+      "step": 10015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962197830061517e-05,
+      "loss": 2.7094,
+      "step": 10020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962159761582755e-05,
+      "loss": 2.71,
+      "step": 10025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962121674018117e-05,
+      "loss": 2.8343,
+      "step": 10030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962083567367746e-05,
+      "loss": 2.7253,
+      "step": 10035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962045441631791e-05,
+      "loss": 2.7858,
+      "step": 10040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962007296810398e-05,
+      "loss": 2.751,
+      "step": 10045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961969132903714e-05,
+      "loss": 2.7643,
+      "step": 10050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961930949911888e-05,
+      "loss": 2.7662,
+      "step": 10055
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961892747835063e-05,
+      "loss": 2.6343,
+      "step": 10060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961854526673386e-05,
+      "loss": 2.6235,
+      "step": 10065
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961816286427008e-05,
+      "loss": 2.6902,
+      "step": 10070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961778027096072e-05,
+      "loss": 2.6966,
+      "step": 10075
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961739748680728e-05,
+      "loss": 2.7333,
+      "step": 10080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961701451181122e-05,
+      "loss": 2.7466,
+      "step": 10085
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961663134597401e-05,
+      "loss": 2.7174,
+      "step": 10090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961624798929712e-05,
+      "loss": 2.8556,
+      "step": 10095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961586444178203e-05,
+      "loss": 2.7753,
+      "step": 10100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961548070343023e-05,
+      "loss": 2.7819,
+      "step": 10105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961509677424319e-05,
+      "loss": 2.5846,
+      "step": 10110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961471265422235e-05,
+      "loss": 2.7659,
+      "step": 10115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961432834336924e-05,
+      "loss": 2.8435,
+      "step": 10120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96139438416853e-05,
+      "loss": 2.6372,
+      "step": 10125
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961355914917204e-05,
+      "loss": 2.6435,
+      "step": 10130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961317426583092e-05,
+      "loss": 2.6644,
+      "step": 10135
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961278919166343e-05,
+      "loss": 2.7995,
+      "step": 10140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961240392667102e-05,
+      "loss": 2.585,
+      "step": 10145
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961201847085521e-05,
+      "loss": 2.6852,
+      "step": 10150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961163282421749e-05,
+      "loss": 2.6331,
+      "step": 10155
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96112469867593e-05,
+      "loss": 2.7298,
+      "step": 10160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961086095848215e-05,
+      "loss": 2.6567,
+      "step": 10165
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961047473938752e-05,
+      "loss": 2.7747,
+      "step": 10170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961008832947691e-05,
+      "loss": 2.7411,
+      "step": 10175
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96097017287518e-05,
+      "loss": 2.6885,
+      "step": 10180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960931493721364e-05,
+      "loss": 2.7462,
+      "step": 10185
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960892795486396e-05,
+      "loss": 2.7715,
+      "step": 10190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960854078170425e-05,
+      "loss": 2.6887,
+      "step": 10195
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960815341773597e-05,
+      "loss": 2.6017,
+      "step": 10200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960776586296064e-05,
+      "loss": 2.6591,
+      "step": 10205
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960737811737973e-05,
+      "loss": 2.8073,
+      "step": 10210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960699018099473e-05,
+      "loss": 2.8129,
+      "step": 10215
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960660205380714e-05,
+      "loss": 2.6008,
+      "step": 10220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960621373581847e-05,
+      "loss": 2.8633,
+      "step": 10225
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960582522703018e-05,
+      "loss": 2.6927,
+      "step": 10230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960543652744378e-05,
+      "loss": 2.8031,
+      "step": 10235
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960504763706077e-05,
+      "loss": 2.6502,
+      "step": 10240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960465855588264e-05,
+      "loss": 2.646,
+      "step": 10245
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960426928391089e-05,
+      "loss": 2.6262,
+      "step": 10250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960387982114702e-05,
+      "loss": 2.6174,
+      "step": 10255
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960349016759251e-05,
+      "loss": 2.663,
+      "step": 10260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960310032324888e-05,
+      "loss": 2.6853,
+      "step": 10265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960271028811763e-05,
+      "loss": 2.5715,
+      "step": 10270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960232006220024e-05,
+      "loss": 2.8747,
+      "step": 10275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960192964549822e-05,
+      "loss": 2.7363,
+      "step": 10280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960153903801308e-05,
+      "loss": 2.7104,
+      "step": 10285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960114823974632e-05,
+      "loss": 2.8362,
+      "step": 10290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960075725069944e-05,
+      "loss": 2.7343,
+      "step": 10295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960036607087393e-05,
+      "loss": 2.8091,
+      "step": 10300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959997470027132e-05,
+      "loss": 2.7727,
+      "step": 10305
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959958313889311e-05,
+      "loss": 2.7707,
+      "step": 10310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959919138674079e-05,
+      "loss": 2.5223,
+      "step": 10315
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959879944381587e-05,
+      "loss": 2.7793,
+      "step": 10320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959840731011987e-05,
+      "loss": 2.7793,
+      "step": 10325
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959801498565428e-05,
+      "loss": 2.7256,
+      "step": 10330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959762247042064e-05,
+      "loss": 2.751,
+      "step": 10335
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959722976442044e-05,
+      "loss": 2.5726,
+      "step": 10340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959683686765517e-05,
+      "loss": 2.7712,
+      "step": 10345
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959644378012638e-05,
+      "loss": 2.7963,
+      "step": 10350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959605050183557e-05,
+      "loss": 2.7502,
+      "step": 10355
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959565703278424e-05,
+      "loss": 2.6824,
+      "step": 10360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959526337297391e-05,
+      "loss": 2.69,
+      "step": 10365
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959486952240609e-05,
+      "loss": 2.6111,
+      "step": 10370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959447548108229e-05,
+      "loss": 2.8184,
+      "step": 10375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959408124900406e-05,
+      "loss": 2.6289,
+      "step": 10380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959368682617287e-05,
+      "loss": 2.6331,
+      "step": 10385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959329221259027e-05,
+      "loss": 2.8042,
+      "step": 10390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959289740825776e-05,
+      "loss": 2.7804,
+      "step": 10395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959250241317688e-05,
+      "loss": 2.6794,
+      "step": 10400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959210722734912e-05,
+      "loss": 2.6291,
+      "step": 10405
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9591711850776e-05,
+      "loss": 2.7145,
+      "step": 10410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959131628345907e-05,
+      "loss": 2.7032,
+      "step": 10415
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959092052539984e-05,
+      "loss": 2.7018,
+      "step": 10420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959052457659984e-05,
+      "loss": 2.7232,
+      "step": 10425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959012843706056e-05,
+      "loss": 2.5876,
+      "step": 10430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958973210678355e-05,
+      "loss": 2.7693,
+      "step": 10435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958933558577033e-05,
+      "loss": 2.8495,
+      "step": 10440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958893887402243e-05,
+      "loss": 2.6839,
+      "step": 10445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958854197154136e-05,
+      "loss": 2.7599,
+      "step": 10450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958814487832866e-05,
+      "loss": 2.7271,
+      "step": 10455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958774759438585e-05,
+      "loss": 2.6412,
+      "step": 10460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958735011971448e-05,
+      "loss": 2.7114,
+      "step": 10465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958695245431604e-05,
+      "loss": 2.5917,
+      "step": 10470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958655459819208e-05,
+      "loss": 2.6402,
+      "step": 10475
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958615655134416e-05,
+      "loss": 2.7427,
+      "step": 10480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958575831377374e-05,
+      "loss": 2.7007,
+      "step": 10485
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958535988548243e-05,
+      "loss": 2.7265,
+      "step": 10490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958496126647171e-05,
+      "loss": 2.6157,
+      "step": 10495
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958456245674314e-05,
+      "loss": 2.6695,
+      "step": 10500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958416345629823e-05,
+      "loss": 2.5999,
+      "step": 10505
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958376426513853e-05,
+      "loss": 2.7925,
+      "step": 10510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958336488326558e-05,
+      "loss": 2.803,
+      "step": 10515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95829653106809e-05,
+      "loss": 2.6006,
+      "step": 10520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958256554738605e-05,
+      "loss": 2.7516,
+      "step": 10525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958216559338254e-05,
+      "loss": 2.7101,
+      "step": 10530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958176544867193e-05,
+      "loss": 2.6398,
+      "step": 10535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958136511325576e-05,
+      "loss": 2.664,
+      "step": 10540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958096458713554e-05,
+      "loss": 2.7875,
+      "step": 10545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958056387031283e-05,
+      "loss": 2.702,
+      "step": 10550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95801629627892e-05,
+      "loss": 2.5811,
+      "step": 10555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957976186456615e-05,
+      "loss": 2.6946,
+      "step": 10560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957936057564523e-05,
+      "loss": 2.5471,
+      "step": 10565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957895909602801e-05,
+      "loss": 2.6019,
+      "step": 10570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9578557425716e-05,
+      "loss": 2.8385,
+      "step": 10575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957815556471075e-05,
+      "loss": 2.7571,
+      "step": 10580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957775351301384e-05,
+      "loss": 2.7782,
+      "step": 10585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957735127062677e-05,
+      "loss": 2.6613,
+      "step": 10590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957694883755113e-05,
+      "loss": 2.6948,
+      "step": 10595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957654621378844e-05,
+      "loss": 2.6005,
+      "step": 10600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957614339934026e-05,
+      "loss": 2.5888,
+      "step": 10605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957574039420812e-05,
+      "loss": 2.7108,
+      "step": 10610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95753371983936e-05,
+      "loss": 2.7481,
+      "step": 10615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957493381189824e-05,
+      "loss": 2.7615,
+      "step": 10620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957453023472358e-05,
+      "loss": 2.7158,
+      "step": 10625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957412646687117e-05,
+      "loss": 2.6212,
+      "step": 10630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957372250834259e-05,
+      "loss": 2.749,
+      "step": 10635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957331835913936e-05,
+      "loss": 2.6478,
+      "step": 10640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957291401926307e-05,
+      "loss": 2.6969,
+      "step": 10645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957250948871524e-05,
+      "loss": 2.5709,
+      "step": 10650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957210476749745e-05,
+      "loss": 2.4934,
+      "step": 10655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957169985561126e-05,
+      "loss": 2.6874,
+      "step": 10660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95712947530582e-05,
+      "loss": 2.6344,
+      "step": 10665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957088945983986e-05,
+      "loss": 2.6831,
+      "step": 10670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957048397595776e-05,
+      "loss": 2.7289,
+      "step": 10675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95700783014135e-05,
+      "loss": 2.6642,
+      "step": 10680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956967243620863e-05,
+      "loss": 2.6578,
+      "step": 10685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95692663803447e-05,
+      "loss": 2.6551,
+      "step": 10690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956886013382326e-05,
+      "loss": 2.7511,
+      "step": 10695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956845369664589e-05,
+      "loss": 2.804,
+      "step": 10700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956804706881417e-05,
+      "loss": 2.7526,
+      "step": 10705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956764025032965e-05,
+      "loss": 2.7007,
+      "step": 10710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956723324119387e-05,
+      "loss": 2.7137,
+      "step": 10715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956682604140843e-05,
+      "loss": 2.6988,
+      "step": 10720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956641865097486e-05,
+      "loss": 2.6465,
+      "step": 10725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956601106989479e-05,
+      "loss": 2.7288,
+      "step": 10730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956560329816971e-05,
+      "loss": 2.625,
+      "step": 10735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956519533580124e-05,
+      "loss": 2.6679,
+      "step": 10740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956478718279094e-05,
+      "loss": 2.8538,
+      "step": 10745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956437883914035e-05,
+      "loss": 2.6751,
+      "step": 10750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95639703048511e-05,
+      "loss": 2.8109,
+      "step": 10755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95635615799247e-05,
+      "loss": 2.6862,
+      "step": 10760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956315266436277e-05,
+      "loss": 2.6816,
+      "step": 10765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956274355816685e-05,
+      "loss": 2.775,
+      "step": 10770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956233426133852e-05,
+      "loss": 2.7247,
+      "step": 10775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956192477387937e-05,
+      "loss": 2.6755,
+      "step": 10780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956151509579096e-05,
+      "loss": 2.6072,
+      "step": 10785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956110522707487e-05,
+      "loss": 2.6677,
+      "step": 10790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956069516773268e-05,
+      "loss": 2.6956,
+      "step": 10795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956028491776595e-05,
+      "loss": 2.7192,
+      "step": 10800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955987447717627e-05,
+      "loss": 2.7108,
+      "step": 10805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955946384596525e-05,
+      "loss": 2.735,
+      "step": 10810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955905302413441e-05,
+      "loss": 2.7306,
+      "step": 10815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955864201168536e-05,
+      "loss": 2.7768,
+      "step": 10820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95582308086197e-05,
+      "loss": 2.6778,
+      "step": 10825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955781941493897e-05,
+      "loss": 2.7115,
+      "step": 10830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95574078306448e-05,
+      "loss": 2.7683,
+      "step": 10835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955699605573874e-05,
+      "loss": 2.7789,
+      "step": 10840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955658409022238e-05,
+      "loss": 2.9012,
+      "step": 10845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95561719340973e-05,
+      "loss": 2.7594,
+      "step": 10850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95557595873651e-05,
+      "loss": 2.6964,
+      "step": 10855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955534705002734e-05,
+      "loss": 2.6493,
+      "step": 10860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955493432208563e-05,
+      "loss": 2.6211,
+      "step": 10865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955452140354156e-05,
+      "loss": 2.7221,
+      "step": 10870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955410829439671e-05,
+      "loss": 2.7547,
+      "step": 10875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955369499465266e-05,
+      "loss": 2.6757,
+      "step": 10880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955328150431102e-05,
+      "loss": 2.7327,
+      "step": 10885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955286782337337e-05,
+      "loss": 2.5945,
+      "step": 10890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955245395184127e-05,
+      "loss": 2.4973,
+      "step": 10895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955203988971638e-05,
+      "loss": 2.6176,
+      "step": 10900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955162563700022e-05,
+      "loss": 2.638,
+      "step": 10905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955121119369443e-05,
+      "loss": 2.6765,
+      "step": 10910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955079655980062e-05,
+      "loss": 2.694,
+      "step": 10915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955038173532032e-05,
+      "loss": 2.6654,
+      "step": 10920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954996672025519e-05,
+      "loss": 2.6317,
+      "step": 10925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954955151460678e-05,
+      "loss": 2.7048,
+      "step": 10930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954913611837671e-05,
+      "loss": 2.7532,
+      "step": 10935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954872053156657e-05,
+      "loss": 2.6814,
+      "step": 10940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954830475417795e-05,
+      "loss": 2.6846,
+      "step": 10945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954788878621249e-05,
+      "loss": 2.6493,
+      "step": 10950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954747262767173e-05,
+      "loss": 2.6394,
+      "step": 10955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954705627855732e-05,
+      "loss": 2.7398,
+      "step": 10960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954663973887084e-05,
+      "loss": 2.6437,
+      "step": 10965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95462230086139e-05,
+      "loss": 2.6437,
+      "step": 10970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954580608778808e-05,
+      "loss": 2.7469,
+      "step": 10975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954538897639502e-05,
+      "loss": 2.832,
+      "step": 10980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95449716744363e-05,
+      "loss": 2.5275,
+      "step": 10985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954455418191353e-05,
+      "loss": 2.605,
+      "step": 10990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95441364988283e-05,
+      "loss": 2.7518,
+      "step": 10995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954371862518226e-05,
+      "loss": 2.6245,
+      "step": 11000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954330056097696e-05,
+      "loss": 2.6896,
+      "step": 11005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954288230621407e-05,
+      "loss": 2.7605,
+      "step": 11010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954246386089515e-05,
+      "loss": 2.7081,
+      "step": 11015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954204522502184e-05,
+      "loss": 2.6191,
+      "step": 11020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954162639859573e-05,
+      "loss": 2.7833,
+      "step": 11025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954120738161843e-05,
+      "loss": 2.8027,
+      "step": 11030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954078817409156e-05,
+      "loss": 2.7128,
+      "step": 11035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954036877601673e-05,
+      "loss": 2.7711,
+      "step": 11040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953994918739555e-05,
+      "loss": 2.8401,
+      "step": 11045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953952940822965e-05,
+      "loss": 2.6653,
+      "step": 11050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953910943852063e-05,
+      "loss": 2.8356,
+      "step": 11055
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953868927827012e-05,
+      "loss": 2.5758,
+      "step": 11060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95382689274797e-05,
+      "loss": 2.6402,
+      "step": 11065
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953784838615104e-05,
+      "loss": 2.7311,
+      "step": 11070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953742765428572e-05,
+      "loss": 2.7564,
+      "step": 11075
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953700673188535e-05,
+      "loss": 2.6585,
+      "step": 11080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953658561895158e-05,
+      "loss": 2.7809,
+      "step": 11085
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9536164315486e-05,
+      "loss": 2.7693,
+      "step": 11090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953574282149027e-05,
+      "loss": 2.8486,
+      "step": 11095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953532113696598e-05,
+      "loss": 2.6701,
+      "step": 11100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953489926191474e-05,
+      "loss": 2.8188,
+      "step": 11105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953447719633822e-05,
+      "loss": 2.7591,
+      "step": 11110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9534054940238e-05,
+      "loss": 2.6512,
+      "step": 11115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953363249361572e-05,
+      "loss": 2.8563,
+      "step": 11120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9533209856473e-05,
+      "loss": 2.6474,
+      "step": 11125
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953278702881149e-05,
+      "loss": 2.7064,
+      "step": 11130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953236401063279e-05,
+      "loss": 2.7286,
+      "step": 11135
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953194080193852e-05,
+      "loss": 2.7431,
+      "step": 11140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953151740273033e-05,
+      "loss": 2.8136,
+      "step": 11145
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953109381300983e-05,
+      "loss": 2.5507,
+      "step": 11150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953067003277867e-05,
+      "loss": 2.581,
+      "step": 11155
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953024606203847e-05,
+      "loss": 2.6907,
+      "step": 11160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952982190079085e-05,
+      "loss": 2.6366,
+      "step": 11165
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952939754903745e-05,
+      "loss": 2.813,
+      "step": 11170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952897300677991e-05,
+      "loss": 2.662,
+      "step": 11175
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952854827401985e-05,
+      "loss": 2.7404,
+      "step": 11180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952812335075891e-05,
+      "loss": 2.7746,
+      "step": 11185
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952769823699875e-05,
+      "loss": 2.7492,
+      "step": 11190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952727293274095e-05,
+      "loss": 2.4037,
+      "step": 11195
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952684743798718e-05,
+      "loss": 2.8084,
+      "step": 11200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952642175273908e-05,
+      "loss": 2.7445,
+      "step": 11205
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952599587699826e-05,
+      "loss": 2.6732,
+      "step": 11210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95255698107664e-05,
+      "loss": 2.6952,
+      "step": 11215
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95251435540451e-05,
+      "loss": 2.646,
+      "step": 11220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952471710683602e-05,
+      "loss": 2.6483,
+      "step": 11225
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95242904691408e-05,
+      "loss": 2.5366,
+      "step": 11230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952386364096106e-05,
+      "loss": 2.6513,
+      "step": 11235
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952343662229849e-05,
+      "loss": 2.712,
+      "step": 11240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952300941315468e-05,
+      "loss": 2.864,
+      "step": 11245
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95225820135313e-05,
+      "loss": 2.6852,
+      "step": 11250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952215442342998e-05,
+      "loss": 2.6755,
+      "step": 11255
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952172664285237e-05,
+      "loss": 2.6552,
+      "step": 11260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952129867180014e-05,
+      "loss": 2.7225,
+      "step": 11265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952087051027488e-05,
+      "loss": 2.7095,
+      "step": 11270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.952044215827829e-05,
+      "loss": 2.7419,
+      "step": 11275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9520013615812e-05,
+      "loss": 2.7888,
+      "step": 11280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951958488287765e-05,
+      "loss": 2.6888,
+      "step": 11285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95191559594769e-05,
+      "loss": 2.6569,
+      "step": 11290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951872684561139e-05,
+      "loss": 2.8294,
+      "step": 11295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951829754128277e-05,
+      "loss": 2.7169,
+      "step": 11300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951786804649271e-05,
+      "loss": 2.6645,
+      "step": 11305
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951743836124284e-05,
+      "loss": 2.7351,
+      "step": 11310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951700848553483e-05,
+      "loss": 2.589,
+      "step": 11315
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951657841937032e-05,
+      "loss": 2.7119,
+      "step": 11320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951614816275096e-05,
+      "loss": 2.6931,
+      "step": 11325
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951571771567843e-05,
+      "loss": 2.6264,
+      "step": 11330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951528707815436e-05,
+      "loss": 2.7897,
+      "step": 11335
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951485625018041e-05,
+      "loss": 2.6396,
+      "step": 11340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951442523175826e-05,
+      "loss": 2.7249,
+      "step": 11345
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951399402288956e-05,
+      "loss": 2.6724,
+      "step": 11350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951356262357594e-05,
+      "loss": 2.7072,
+      "step": 11355
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951313103381907e-05,
+      "loss": 2.5483,
+      "step": 11360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951269925362063e-05,
+      "loss": 2.652,
+      "step": 11365
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951226728298229e-05,
+      "loss": 2.5671,
+      "step": 11370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951183512190566e-05,
+      "loss": 2.5961,
+      "step": 11375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951140277039245e-05,
+      "loss": 2.6053,
+      "step": 11380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95109702284443e-05,
+      "loss": 2.7375,
+      "step": 11385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951053749606288e-05,
+      "loss": 2.6697,
+      "step": 11390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.951010457324985e-05,
+      "loss": 2.5998,
+      "step": 11395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95096714600069e-05,
+      "loss": 2.79,
+      "step": 11400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950923815633565e-05,
+      "loss": 2.7667,
+      "step": 11405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.95088046622378e-05,
+      "loss": 2.7221,
+      "step": 11410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950837097771502e-05,
+      "loss": 2.8603,
+      "step": 11415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950793710276895e-05,
+      "loss": 2.6151,
+      "step": 11420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950750303740128e-05,
+      "loss": 2.6686,
+      "step": 11425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950706878161369e-05,
+      "loss": 2.5767,
+      "step": 11430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950663433540782e-05,
+      "loss": 2.7045,
+      "step": 11435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950619969878536e-05,
+      "loss": 2.7005,
+      "step": 11440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950576487174799e-05,
+      "loss": 2.5823,
+      "step": 11445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950532985429736e-05,
+      "loss": 2.7496,
+      "step": 11450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950489464643516e-05,
+      "loss": 2.7422,
+      "step": 11455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950445924816305e-05,
+      "loss": 2.6206,
+      "step": 11460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950402365948271e-05,
+      "loss": 2.7077,
+      "step": 11465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950358788039584e-05,
+      "loss": 2.6965,
+      "step": 11470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950315191090408e-05,
+      "loss": 2.8215,
+      "step": 11475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950271575100912e-05,
+      "loss": 2.7084,
+      "step": 11480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950227940071263e-05,
+      "loss": 2.5532,
+      "step": 11485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950184286001631e-05,
+      "loss": 2.7514,
+      "step": 11490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950140612892183e-05,
+      "loss": 2.7468,
+      "step": 11495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950096920743086e-05,
+      "loss": 2.6553,
+      "step": 11500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950053209554508e-05,
+      "loss": 2.7271,
+      "step": 11505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950009479326619e-05,
+      "loss": 2.688,
+      "step": 11510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949965730059586e-05,
+      "loss": 2.6153,
+      "step": 11515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949921961753576e-05,
+      "loss": 2.7374,
+      "step": 11520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94987817440876e-05,
+      "loss": 2.7621,
+      "step": 11525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949834368025305e-05,
+      "loss": 2.7129,
+      "step": 11530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949790542603379e-05,
+      "loss": 2.5378,
+      "step": 11535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949746698143151e-05,
+      "loss": 2.7843,
+      "step": 11540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94970283464479e-05,
+      "loss": 2.6721,
+      "step": 11545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949658952108464e-05,
+      "loss": 2.7227,
+      "step": 11550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949615050534342e-05,
+      "loss": 2.7138,
+      "step": 11555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949571129922594e-05,
+      "loss": 2.7146,
+      "step": 11560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949527190273387e-05,
+      "loss": 2.5922,
+      "step": 11565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949483231586892e-05,
+      "loss": 2.5842,
+      "step": 11570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949439253863276e-05,
+      "loss": 2.7955,
+      "step": 11575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949395257102711e-05,
+      "loss": 2.655,
+      "step": 11580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949351241305363e-05,
+      "loss": 2.6026,
+      "step": 11585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949307206471404e-05,
+      "loss": 2.6523,
+      "step": 11590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949263152601001e-05,
+      "loss": 2.7278,
+      "step": 11595
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949219079694323e-05,
+      "loss": 2.4757,
+      "step": 11600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949174987751544e-05,
+      "loss": 2.6346,
+      "step": 11605
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949130876772829e-05,
+      "loss": 2.6002,
+      "step": 11610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94908674675835e-05,
+      "loss": 2.7429,
+      "step": 11615
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949042597708276e-05,
+      "loss": 2.8779,
+      "step": 11620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948998429622776e-05,
+      "loss": 2.7336,
+      "step": 11625
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948954242502022e-05,
+      "loss": 2.5609,
+      "step": 11630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948910036346181e-05,
+      "loss": 2.6752,
+      "step": 11635
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948865811155427e-05,
+      "loss": 2.7277,
+      "step": 11640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948821566929927e-05,
+      "loss": 2.7214,
+      "step": 11645
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948777303669851e-05,
+      "loss": 2.7205,
+      "step": 11650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948733021375373e-05,
+      "loss": 2.535,
+      "step": 11655
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948688720046659e-05,
+      "loss": 2.6735,
+      "step": 11660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948644399683881e-05,
+      "loss": 2.5678,
+      "step": 11665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948600060287208e-05,
+      "loss": 2.7546,
+      "step": 11670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948555701856814e-05,
+      "loss": 2.6819,
+      "step": 11675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948511324392868e-05,
+      "loss": 2.619,
+      "step": 11680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948466927895539e-05,
+      "loss": 2.7387,
+      "step": 11685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948422512364999e-05,
+      "loss": 2.7197,
+      "step": 11690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948378077801419e-05,
+      "loss": 2.6609,
+      "step": 11695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94833362420497e-05,
+      "loss": 2.6399,
+      "step": 11700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948289151575824e-05,
+      "loss": 2.7195,
+      "step": 11705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94824465991415e-05,
+      "loss": 2.6482,
+      "step": 11710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94820014922012e-05,
+      "loss": 2.5596,
+      "step": 11715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948155619493904e-05,
+      "loss": 2.6879,
+      "step": 11720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948111070735675e-05,
+      "loss": 2.628,
+      "step": 11725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948066502945604e-05,
+      "loss": 2.788,
+      "step": 11730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948021916123863e-05,
+      "loss": 2.7166,
+      "step": 11735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947977310270622e-05,
+      "loss": 2.7395,
+      "step": 11740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947932685386055e-05,
+      "loss": 2.7307,
+      "step": 11745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947888041470328e-05,
+      "loss": 2.6634,
+      "step": 11750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94784337852362e-05,
+      "loss": 2.6926,
+      "step": 11755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947798696546099e-05,
+      "loss": 2.6949,
+      "step": 11760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947753995537937e-05,
+      "loss": 2.6962,
+      "step": 11765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947709275499307e-05,
+      "loss": 2.7219,
+      "step": 11770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947664536430378e-05,
+      "loss": 2.7349,
+      "step": 11775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947619778331325e-05,
+      "loss": 2.73,
+      "step": 11780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947575001202321e-05,
+      "loss": 2.6559,
+      "step": 11785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947530205043534e-05,
+      "loss": 2.5135,
+      "step": 11790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947485389855143e-05,
+      "loss": 2.6013,
+      "step": 11795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947440555637313e-05,
+      "loss": 2.7269,
+      "step": 11800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947395702390221e-05,
+      "loss": 2.6422,
+      "step": 11805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947350830114037e-05,
+      "loss": 2.6968,
+      "step": 11810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947305938808938e-05,
+      "loss": 2.7369,
+      "step": 11815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94726102847509e-05,
+      "loss": 2.6314,
+      "step": 11820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94721609911267e-05,
+      "loss": 2.5987,
+      "step": 11825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947171150721852e-05,
+      "loss": 2.7333,
+      "step": 11830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947126183302806e-05,
+      "loss": 2.6067,
+      "step": 11835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947081196855706e-05,
+      "loss": 2.5878,
+      "step": 11840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947036191380725e-05,
+      "loss": 2.6019,
+      "step": 11845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946991166878036e-05,
+      "loss": 2.763,
+      "step": 11850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946946123347813e-05,
+      "loss": 2.6414,
+      "step": 11855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946901060790228e-05,
+      "loss": 2.6727,
+      "step": 11860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946855979205454e-05,
+      "loss": 2.7798,
+      "step": 11865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946810878593667e-05,
+      "loss": 2.6372,
+      "step": 11870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946765758955037e-05,
+      "loss": 2.6274,
+      "step": 11875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94672062028974e-05,
+      "loss": 2.7058,
+      "step": 11880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94667546259795e-05,
+      "loss": 2.5773,
+      "step": 11885
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946630285879837e-05,
+      "loss": 2.645,
+      "step": 11890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946585090135579e-05,
+      "loss": 2.7373,
+      "step": 11895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946539875365348e-05,
+      "loss": 2.525,
+      "step": 11900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946494641569318e-05,
+      "loss": 2.6115,
+      "step": 11905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946449388747663e-05,
+      "loss": 2.6117,
+      "step": 11910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946404116900558e-05,
+      "loss": 2.7365,
+      "step": 11915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946358826028175e-05,
+      "loss": 2.5769,
+      "step": 11920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94631351613069e-05,
+      "loss": 2.6156,
+      "step": 11925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946268187208277e-05,
+      "loss": 2.6587,
+      "step": 11930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946222839261108e-05,
+      "loss": 2.7388,
+      "step": 11935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946177472289361e-05,
+      "loss": 2.5931,
+      "step": 11940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946132086293209e-05,
+      "loss": 2.5995,
+      "step": 11945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946086681272828e-05,
+      "loss": 2.7304,
+      "step": 11950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946041257228389e-05,
+      "loss": 2.7152,
+      "step": 11955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94599581416007e-05,
+      "loss": 2.6167,
+      "step": 11960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945950352068044e-05,
+      "loss": 2.6642,
+      "step": 11965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945904870952487e-05,
+      "loss": 2.7824,
+      "step": 11970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945859370813572e-05,
+      "loss": 2.7261,
+      "step": 11975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945813851651478e-05,
+      "loss": 2.8387,
+      "step": 11980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945768313466377e-05,
+      "loss": 2.6631,
+      "step": 11985
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945722756258444e-05,
+      "loss": 2.6542,
+      "step": 11990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945677180027854e-05,
+      "loss": 2.7225,
+      "step": 11995
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945631584774786e-05,
+      "loss": 2.783,
+      "step": 12000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94558597049941e-05,
+      "loss": 2.7227,
+      "step": 12005
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945540337201906e-05,
+      "loss": 2.6211,
+      "step": 12010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945494684882446e-05,
+      "loss": 2.6507,
+      "step": 12015
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945449013541209e-05,
+      "loss": 2.5528,
+      "step": 12020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945403323178366e-05,
+      "loss": 2.8117,
+      "step": 12025
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945357613794097e-05,
+      "loss": 2.6431,
+      "step": 12030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945311885388577e-05,
+      "loss": 2.6327,
+      "step": 12035
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945266137961982e-05,
+      "loss": 2.6701,
+      "step": 12040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945220371514486e-05,
+      "loss": 2.5563,
+      "step": 12045
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945174586046266e-05,
+      "loss": 2.6008,
+      "step": 12050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945128781557498e-05,
+      "loss": 2.6957,
+      "step": 12055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945082958048359e-05,
+      "loss": 2.7684,
+      "step": 12060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945037115519026e-05,
+      "loss": 2.6359,
+      "step": 12065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944991253969674e-05,
+      "loss": 2.7777,
+      "step": 12070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944945373400478e-05,
+      "loss": 2.6606,
+      "step": 12075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944899473811617e-05,
+      "loss": 2.5789,
+      "step": 12080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944853555203265e-05,
+      "loss": 2.6406,
+      "step": 12085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944807617575602e-05,
+      "loss": 2.9032,
+      "step": 12090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944761660928801e-05,
+      "loss": 2.7225,
+      "step": 12095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944715685263042e-05,
+      "loss": 2.6566,
+      "step": 12100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9446696905785e-05,
+      "loss": 2.4678,
+      "step": 12105
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944623676875353e-05,
+      "loss": 2.7247,
+      "step": 12110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944577644153777e-05,
+      "loss": 2.6277,
+      "step": 12115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944531592413949e-05,
+      "loss": 2.705,
+      "step": 12120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944485521656046e-05,
+      "loss": 2.6013,
+      "step": 12125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944439431880247e-05,
+      "loss": 2.7102,
+      "step": 12130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944393323086728e-05,
+      "loss": 2.7037,
+      "step": 12135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944347195275665e-05,
+      "loss": 2.533,
+      "step": 12140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944301048447237e-05,
+      "loss": 2.7894,
+      "step": 12145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944254882601622e-05,
+      "loss": 2.7373,
+      "step": 12150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944208697738997e-05,
+      "loss": 2.7086,
+      "step": 12155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944162493859539e-05,
+      "loss": 2.6017,
+      "step": 12160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944116270963426e-05,
+      "loss": 2.7383,
+      "step": 12165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944070029050836e-05,
+      "loss": 2.7468,
+      "step": 12170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944023768121946e-05,
+      "loss": 2.6786,
+      "step": 12175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943977488176937e-05,
+      "loss": 2.7108,
+      "step": 12180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943931189215983e-05,
+      "loss": 2.7298,
+      "step": 12185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943884871239263e-05,
+      "loss": 2.6418,
+      "step": 12190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943838534246958e-05,
+      "loss": 2.6179,
+      "step": 12195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943792178239242e-05,
+      "loss": 2.7351,
+      "step": 12200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943745803216298e-05,
+      "loss": 2.6276,
+      "step": 12205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9436994091783e-05,
+      "loss": 2.6842,
+      "step": 12210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943652996125428e-05,
+      "loss": 2.7542,
+      "step": 12215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943606564057864e-05,
+      "loss": 2.7165,
+      "step": 12220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94356011297578e-05,
+      "loss": 2.612,
+      "step": 12225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943513642879358e-05,
+      "loss": 2.844,
+      "step": 12230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94346715376878e-05,
+      "loss": 2.6479,
+      "step": 12235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943420645644219e-05,
+      "loss": 2.7573,
+      "step": 12240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943374118505857e-05,
+      "loss": 2.7265,
+      "step": 12245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943327572353874e-05,
+      "loss": 2.6947,
+      "step": 12250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943281007188446e-05,
+      "loss": 2.6687,
+      "step": 12255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943234423009753e-05,
+      "loss": 2.7296,
+      "step": 12260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943187819817976e-05,
+      "loss": 2.6721,
+      "step": 12265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943141197613293e-05,
+      "loss": 2.574,
+      "step": 12270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943094556395883e-05,
+      "loss": 2.7609,
+      "step": 12275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943047896165925e-05,
+      "loss": 2.8064,
+      "step": 12280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9430012169236e-05,
+      "loss": 2.6804,
+      "step": 12285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942954518669088e-05,
+      "loss": 2.7078,
+      "step": 12290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942907801402566e-05,
+      "loss": 2.5352,
+      "step": 12295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942861065124213e-05,
+      "loss": 2.6959,
+      "step": 12300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942814309834215e-05,
+      "loss": 2.6283,
+      "step": 12305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942767535532744e-05,
+      "loss": 2.6656,
+      "step": 12310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942720742219987e-05,
+      "loss": 2.6746,
+      "step": 12315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942673929896118e-05,
+      "loss": 2.8461,
+      "step": 12320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94262709856132e-05,
+      "loss": 2.5045,
+      "step": 12325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942580248215774e-05,
+      "loss": 2.6845,
+      "step": 12330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942533378859658e-05,
+      "loss": 2.5506,
+      "step": 12335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942486490493153e-05,
+      "loss": 2.6132,
+      "step": 12340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942439583116438e-05,
+      "loss": 2.6678,
+      "step": 12345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942392656729698e-05,
+      "loss": 2.7833,
+      "step": 12350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942345711333109e-05,
+      "loss": 2.7035,
+      "step": 12355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942298746926853e-05,
+      "loss": 2.7319,
+      "step": 12360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94225176351111e-05,
+      "loss": 2.7626,
+      "step": 12365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942204761086062e-05,
+      "loss": 2.6635,
+      "step": 12370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942157739651891e-05,
+      "loss": 2.7163,
+      "step": 12375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942110699208772e-05,
+      "loss": 2.7053,
+      "step": 12380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942063639756893e-05,
+      "loss": 2.5813,
+      "step": 12385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942016561296432e-05,
+      "loss": 2.6646,
+      "step": 12390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94196946382757e-05,
+      "loss": 2.7433,
+      "step": 12395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941922347350487e-05,
+      "loss": 2.6984,
+      "step": 12400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941875211865366e-05,
+      "loss": 2.6136,
+      "step": 12405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941828057372387e-05,
+      "loss": 2.6431,
+      "step": 12410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941780883871733e-05,
+      "loss": 2.5829,
+      "step": 12415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941733691363584e-05,
+      "loss": 2.6434,
+      "step": 12420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941686479848122e-05,
+      "loss": 2.7427,
+      "step": 12425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941639249325529e-05,
+      "loss": 2.546,
+      "step": 12430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941591999795986e-05,
+      "loss": 2.6347,
+      "step": 12435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941544731259675e-05,
+      "loss": 2.7595,
+      "step": 12440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941497443716779e-05,
+      "loss": 2.7805,
+      "step": 12445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941450137167476e-05,
+      "loss": 2.5443,
+      "step": 12450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941402811611953e-05,
+      "loss": 2.6555,
+      "step": 12455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941355467050388e-05,
+      "loss": 2.7425,
+      "step": 12460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941308103482966e-05,
+      "loss": 2.7267,
+      "step": 12465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941260720909868e-05,
+      "loss": 2.6644,
+      "step": 12470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941213319331275e-05,
+      "loss": 2.7065,
+      "step": 12475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941165898747371e-05,
+      "loss": 2.7525,
+      "step": 12480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941118459158338e-05,
+      "loss": 2.6111,
+      "step": 12485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941071000564359e-05,
+      "loss": 2.5688,
+      "step": 12490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941023522965616e-05,
+      "loss": 2.7819,
+      "step": 12495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940976026362291e-05,
+      "loss": 2.6025,
+      "step": 12500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940928510754568e-05,
+      "loss": 2.6122,
+      "step": 12505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940880976142628e-05,
+      "loss": 2.6996,
+      "step": 12510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940833422526654e-05,
+      "loss": 2.7325,
+      "step": 12515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940785849906832e-05,
+      "loss": 2.6779,
+      "step": 12520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940738258283342e-05,
+      "loss": 2.7181,
+      "step": 12525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940690647656367e-05,
+      "loss": 2.7322,
+      "step": 12530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94064301802609e-05,
+      "loss": 2.7441,
+      "step": 12535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940595369392697e-05,
+      "loss": 2.706,
+      "step": 12540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940547701756369e-05,
+      "loss": 2.6882,
+      "step": 12545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940500015117289e-05,
+      "loss": 2.6852,
+      "step": 12550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94045230947564e-05,
+      "loss": 2.7048,
+      "step": 12555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940404584831608e-05,
+      "loss": 2.6992,
+      "step": 12560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940356841185376e-05,
+      "loss": 2.7731,
+      "step": 12565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940309078537124e-05,
+      "loss": 2.624,
+      "step": 12570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940261296887041e-05,
+      "loss": 2.6205,
+      "step": 12575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940213496235308e-05,
+      "loss": 2.7958,
+      "step": 12580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940165676582108e-05,
+      "loss": 2.5259,
+      "step": 12585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940117837927627e-05,
+      "loss": 2.6875,
+      "step": 12590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940069980272048e-05,
+      "loss": 2.6602,
+      "step": 12595
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940022103615556e-05,
+      "loss": 2.6909,
+      "step": 12600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939974207958332e-05,
+      "loss": 2.6657,
+      "step": 12605
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939926293300564e-05,
+      "loss": 2.6885,
+      "step": 12610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939878359642433e-05,
+      "loss": 2.5971,
+      "step": 12615
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939830406984127e-05,
+      "loss": 2.8563,
+      "step": 12620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939782435325829e-05,
+      "loss": 2.7012,
+      "step": 12625
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939734444667722e-05,
+      "loss": 2.5799,
+      "step": 12630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939686435009991e-05,
+      "loss": 2.6507,
+      "step": 12635
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939638406352822e-05,
+      "loss": 2.7749,
+      "step": 12640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9395903586964e-05,
+      "loss": 2.6659,
+      "step": 12645
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939542292040908e-05,
+      "loss": 2.618,
+      "step": 12650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939494206386533e-05,
+      "loss": 2.663,
+      "step": 12655
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939446101733458e-05,
+      "loss": 2.7191,
+      "step": 12660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939397978081868e-05,
+      "loss": 2.6542,
+      "step": 12665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93934983543195e-05,
+      "loss": 2.6322,
+      "step": 12670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939301673783887e-05,
+      "loss": 2.6827,
+      "step": 12675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939253493137865e-05,
+      "loss": 2.6831,
+      "step": 12680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939205293494072e-05,
+      "loss": 2.6615,
+      "step": 12685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939157074852687e-05,
+      "loss": 2.7603,
+      "step": 12690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939108837213903e-05,
+      "loss": 2.7285,
+      "step": 12695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9390605805779e-05,
+      "loss": 2.6593,
+      "step": 12700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939012304944866e-05,
+      "loss": 2.5928,
+      "step": 12705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938964010314987e-05,
+      "loss": 2.8517,
+      "step": 12710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938915696688447e-05,
+      "loss": 2.7205,
+      "step": 12715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938867364065434e-05,
+      "loss": 2.6601,
+      "step": 12720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938819012446132e-05,
+      "loss": 2.4997,
+      "step": 12725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938770641830728e-05,
+      "loss": 2.6947,
+      "step": 12730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938722252219406e-05,
+      "loss": 2.7803,
+      "step": 12735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938673843612356e-05,
+      "loss": 2.6216,
+      "step": 12740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938625416009762e-05,
+      "loss": 2.7869,
+      "step": 12745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938576969411808e-05,
+      "loss": 2.8174,
+      "step": 12750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938528503818684e-05,
+      "loss": 2.6197,
+      "step": 12755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938480019230576e-05,
+      "loss": 2.8074,
+      "step": 12760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938431515647669e-05,
+      "loss": 2.6056,
+      "step": 12765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93838299307015e-05,
+      "loss": 2.6404,
+      "step": 12770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938334451498206e-05,
+      "loss": 2.6959,
+      "step": 12775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938285890932025e-05,
+      "loss": 2.5328,
+      "step": 12780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938237311371789e-05,
+      "loss": 2.8233,
+      "step": 12785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93818871281769e-05,
+      "loss": 2.7038,
+      "step": 12790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938140095269912e-05,
+      "loss": 2.6482,
+      "step": 12795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938091458728644e-05,
+      "loss": 2.7361,
+      "step": 12800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938042803194072e-05,
+      "loss": 2.6439,
+      "step": 12805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937994128666383e-05,
+      "loss": 2.5036,
+      "step": 12810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937945435145765e-05,
+      "loss": 2.6565,
+      "step": 12815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937896722632404e-05,
+      "loss": 2.8607,
+      "step": 12820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937847991126487e-05,
+      "loss": 2.6299,
+      "step": 12825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937799240628205e-05,
+      "loss": 2.744,
+      "step": 12830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937750471137741e-05,
+      "loss": 2.6753,
+      "step": 12835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937701682655287e-05,
+      "loss": 2.6346,
+      "step": 12840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937652875181026e-05,
+      "loss": 2.7294,
+      "step": 12845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937604048715149e-05,
+      "loss": 2.605,
+      "step": 12850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937555203257843e-05,
+      "loss": 2.7102,
+      "step": 12855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937506338809295e-05,
+      "loss": 2.6314,
+      "step": 12860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937457455369694e-05,
+      "loss": 2.6528,
+      "step": 12865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937408552939227e-05,
+      "loss": 2.7609,
+      "step": 12870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937359631518084e-05,
+      "loss": 2.594,
+      "step": 12875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93731069110645e-05,
+      "loss": 2.6912,
+      "step": 12880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937261731704518e-05,
+      "loss": 2.5463,
+      "step": 12885
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937212753312471e-05,
+      "loss": 2.5609,
+      "step": 12890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9371637559305e-05,
+      "loss": 2.6811,
+      "step": 12895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937114739558794e-05,
+      "loss": 2.7091,
+      "step": 12900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937065704197541e-05,
+      "loss": 2.7711,
+      "step": 12905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937016649846929e-05,
+      "loss": 2.6271,
+      "step": 12910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936967576507147e-05,
+      "loss": 2.6947,
+      "step": 12915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936918484178384e-05,
+      "loss": 2.6588,
+      "step": 12920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936869372860829e-05,
+      "loss": 2.5525,
+      "step": 12925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93682024255467e-05,
+      "loss": 2.7225,
+      "step": 12930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936771093260098e-05,
+      "loss": 2.8134,
+      "step": 12935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936721924977299e-05,
+      "loss": 2.7284,
+      "step": 12940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936672737706462e-05,
+      "loss": 2.7936,
+      "step": 12945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936623531447782e-05,
+      "loss": 2.7503,
+      "step": 12950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936574306201442e-05,
+      "loss": 2.7604,
+      "step": 12955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936525061967635e-05,
+      "loss": 2.5926,
+      "step": 12960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936475798746546e-05,
+      "loss": 2.6455,
+      "step": 12965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936426516538369e-05,
+      "loss": 2.7689,
+      "step": 12970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936377215343291e-05,
+      "loss": 2.5628,
+      "step": 12975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936327895161504e-05,
+      "loss": 2.7569,
+      "step": 12980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936278555993197e-05,
+      "loss": 2.8317,
+      "step": 12985
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936229197838556e-05,
+      "loss": 2.6758,
+      "step": 12990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936179820697777e-05,
+      "loss": 2.6637,
+      "step": 12995
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936130424571044e-05,
+      "loss": 2.818,
+      "step": 13000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936081009458551e-05,
+      "loss": 2.675,
+      "step": 13005
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936031575360487e-05,
+      "loss": 2.4786,
+      "step": 13010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935982122277043e-05,
+      "loss": 2.6461,
+      "step": 13015
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935932650208406e-05,
+      "loss": 2.7156,
+      "step": 13020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93588315915477e-05,
+      "loss": 2.6423,
+      "step": 13025
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935833649116324e-05,
+      "loss": 2.7376,
+      "step": 13030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935784120093258e-05,
+      "loss": 2.6991,
+      "step": 13035
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935734572085761e-05,
+      "loss": 2.7683,
+      "step": 13040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935685005094028e-05,
+      "loss": 2.5844,
+      "step": 13045
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935635419118245e-05,
+      "loss": 2.6901,
+      "step": 13050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935585814158606e-05,
+      "loss": 2.7171,
+      "step": 13055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9355361902153e-05,
+      "loss": 2.5886,
+      "step": 13060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93548654728852e-05,
+      "loss": 2.8203,
+      "step": 13065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935436885378454e-05,
+      "loss": 2.7473,
+      "step": 13070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935387204485292e-05,
+      "loss": 2.7261,
+      "step": 13075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93533750460923e-05,
+      "loss": 2.6231,
+      "step": 13080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935287785750455e-05,
+      "loss": 2.6246,
+      "step": 13085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935238047909161e-05,
+      "loss": 2.7234,
+      "step": 13090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935188291085539e-05,
+      "loss": 2.7038,
+      "step": 13095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935138515279779e-05,
+      "loss": 2.6567,
+      "step": 13100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935088720492071e-05,
+      "loss": 2.6848,
+      "step": 13105
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935038906722611e-05,
+      "loss": 2.6825,
+      "step": 13110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934989073971587e-05,
+      "loss": 2.6045,
+      "step": 13115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934939222239191e-05,
+      "loss": 2.4965,
+      "step": 13120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934889351525615e-05,
+      "loss": 2.6751,
+      "step": 13125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934839461831053e-05,
+      "loss": 2.6259,
+      "step": 13130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934789553155696e-05,
+      "loss": 2.6294,
+      "step": 13135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934739625499734e-05,
+      "loss": 2.6626,
+      "step": 13140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93468967886336e-05,
+      "loss": 2.6795,
+      "step": 13145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934639713246765e-05,
+      "loss": 2.7588,
+      "step": 13150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934589728650144e-05,
+      "loss": 2.7584,
+      "step": 13155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934539725073688e-05,
+      "loss": 2.582,
+      "step": 13160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934489702517589e-05,
+      "loss": 2.6836,
+      "step": 13165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934439660982037e-05,
+      "loss": 2.7238,
+      "step": 13170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934389600467231e-05,
+      "loss": 2.6964,
+      "step": 13175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934339520973357e-05,
+      "loss": 2.6425,
+      "step": 13180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934289422500611e-05,
+      "loss": 2.6348,
+      "step": 13185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934239305049184e-05,
+      "loss": 2.7247,
+      "step": 13190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934189168619271e-05,
+      "loss": 2.6803,
+      "step": 13195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934139013211061e-05,
+      "loss": 2.7536,
+      "step": 13200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934088838824751e-05,
+      "loss": 2.6556,
+      "step": 13205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934038645460532e-05,
+      "loss": 2.6761,
+      "step": 13210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933988433118598e-05,
+      "loss": 2.5556,
+      "step": 13215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93393820179914e-05,
+      "loss": 2.6266,
+      "step": 13220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933887951502353e-05,
+      "loss": 2.5656,
+      "step": 13225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93383768222843e-05,
+      "loss": 2.7276,
+      "step": 13230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933787393977565e-05,
+      "loss": 2.703,
+      "step": 13235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93373708674995e-05,
+      "loss": 2.8341,
+      "step": 13240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93368676054578e-05,
+      "loss": 2.5851,
+      "step": 13245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933636415365249e-05,
+      "loss": 2.7174,
+      "step": 13250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933586051208548e-05,
+      "loss": 2.6694,
+      "step": 13255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933535668075871e-05,
+      "loss": 2.7261,
+      "step": 13260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933485265967414e-05,
+      "loss": 2.6228,
+      "step": 13265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93343484488337e-05,
+      "loss": 2.7122,
+      "step": 13270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933384404823933e-05,
+      "loss": 2.6745,
+      "step": 13275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933333945789295e-05,
+      "loss": 2.5244,
+      "step": 13280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933283467779656e-05,
+      "loss": 2.7004,
+      "step": 13285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933232970795201e-05,
+      "loss": 2.771,
+      "step": 13290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933182454836131e-05,
+      "loss": 2.6342,
+      "step": 13295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933131919902641e-05,
+      "loss": 2.733,
+      "step": 13300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93308136599492e-05,
+      "loss": 2.7083,
+      "step": 13305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933030793113166e-05,
+      "loss": 2.6784,
+      "step": 13310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932980201257573e-05,
+      "loss": 2.469,
+      "step": 13315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932929590428335e-05,
+      "loss": 2.6256,
+      "step": 13320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932878960625649e-05,
+      "loss": 2.6089,
+      "step": 13325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932828311849705e-05,
+      "loss": 2.6765,
+      "step": 13330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932777644100702e-05,
+      "loss": 2.5398,
+      "step": 13335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932726957378833e-05,
+      "loss": 2.6196,
+      "step": 13340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932676251684295e-05,
+      "loss": 2.7022,
+      "step": 13345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93262552701728e-05,
+      "loss": 2.6236,
+      "step": 13350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932574783377986e-05,
+      "loss": 2.7033,
+      "step": 13355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932524020766606e-05,
+      "loss": 2.7751,
+      "step": 13360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932473239183335e-05,
+      "loss": 2.8311,
+      "step": 13365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932422438628371e-05,
+      "loss": 2.7263,
+      "step": 13370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932371619101907e-05,
+      "loss": 2.7888,
+      "step": 13375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93232078060414e-05,
+      "loss": 2.6467,
+      "step": 13380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932269923135264e-05,
+      "loss": 2.5551,
+      "step": 13385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932219046695474e-05,
+      "loss": 2.6878,
+      "step": 13390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932168151284969e-05,
+      "loss": 2.6844,
+      "step": 13395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932117236903943e-05,
+      "loss": 2.6198,
+      "step": 13400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93206630355259e-05,
+      "loss": 2.5805,
+      "step": 13405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932015351231109e-05,
+      "loss": 2.6655,
+      "step": 13410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931964379939693e-05,
+      "loss": 2.5875,
+      "step": 13415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931913389678543e-05,
+      "loss": 2.6225,
+      "step": 13420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931862380447848e-05,
+      "loss": 2.5464,
+      "step": 13425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931811352247808e-05,
+      "loss": 2.6761,
+      "step": 13430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931760305078621e-05,
+      "loss": 2.6232,
+      "step": 13435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931709238940481e-05,
+      "loss": 2.718,
+      "step": 13440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931658153833585e-05,
+      "loss": 2.6219,
+      "step": 13445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931607049758129e-05,
+      "loss": 2.6225,
+      "step": 13450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93155592671431e-05,
+      "loss": 2.6728,
+      "step": 13455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931504784702325e-05,
+      "loss": 2.587,
+      "step": 13460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931453623722368e-05,
+      "loss": 2.5687,
+      "step": 13465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931402443774643e-05,
+      "loss": 2.6843,
+      "step": 13470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931351244859337e-05,
+      "loss": 2.6146,
+      "step": 13475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931300026976654e-05,
+      "loss": 2.5839,
+      "step": 13480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931248790126789e-05,
+      "loss": 2.7056,
+      "step": 13485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93119753430994e-05,
+      "loss": 2.5289,
+      "step": 13490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931146259526301e-05,
+      "loss": 2.8506,
+      "step": 13495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931094965776071e-05,
+      "loss": 2.68,
+      "step": 13500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931043653059449e-05,
+      "loss": 2.5907,
+      "step": 13505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930992321376631e-05,
+      "loss": 2.5203,
+      "step": 13510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930940970727813e-05,
+      "loss": 2.6995,
+      "step": 13515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930889601113195e-05,
+      "loss": 2.6999,
+      "step": 13520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930838212532973e-05,
+      "loss": 2.548,
+      "step": 13525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930786804987347e-05,
+      "loss": 2.5856,
+      "step": 13530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93073537847651e-05,
+      "loss": 2.5809,
+      "step": 13535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930683933000663e-05,
+      "loss": 2.5937,
+      "step": 13540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930632468560006e-05,
+      "loss": 2.645,
+      "step": 13545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930580985154734e-05,
+      "loss": 2.4719,
+      "step": 13550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930529482785043e-05,
+      "loss": 2.6439,
+      "step": 13555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930477961451136e-05,
+      "loss": 2.6779,
+      "step": 13560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930426421153207e-05,
+      "loss": 2.7066,
+      "step": 13565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930374861891459e-05,
+      "loss": 2.6663,
+      "step": 13570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930323283666083e-05,
+      "loss": 2.7404,
+      "step": 13575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930271686477284e-05,
+      "loss": 2.6448,
+      "step": 13580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930220070325258e-05,
+      "loss": 2.802,
+      "step": 13585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930168435210204e-05,
+      "loss": 2.8057,
+      "step": 13590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93011678113232e-05,
+      "loss": 2.7506,
+      "step": 13595
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930065108091805e-05,
+      "loss": 2.5829,
+      "step": 13600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930013416088858e-05,
+      "loss": 2.6685,
+      "step": 13605
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929961705123677e-05,
+      "loss": 2.6231,
+      "step": 13610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929909975196462e-05,
+      "loss": 2.6104,
+      "step": 13615
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929858226307411e-05,
+      "loss": 2.8485,
+      "step": 13620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929806458456722e-05,
+      "loss": 2.6779,
+      "step": 13625
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929754671644598e-05,
+      "loss": 2.6479,
+      "step": 13630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929702865871236e-05,
+      "loss": 2.7071,
+      "step": 13635
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929651041136833e-05,
+      "loss": 2.8153,
+      "step": 13640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.92959919744159e-05,
+      "loss": 2.6954,
+      "step": 13645
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929547334785709e-05,
+      "loss": 2.6769,
+      "step": 13650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929495453169387e-05,
+      "loss": 2.7933,
+      "step": 13655
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.92944355259282e-05,
+      "loss": 2.7934,
+      "step": 13660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929391633056215e-05,
+      "loss": 2.5961,
+      "step": 13665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929339694559767e-05,
+      "loss": 2.7079,
+      "step": 13670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929287737103678e-05,
+      "loss": 2.7237,
+      "step": 13675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929235760688146e-05,
+      "loss": 2.511,
+      "step": 13680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929183765313371e-05,
+      "loss": 2.6833,
+      "step": 13685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929131750979555e-05,
+      "loss": 2.8404,
+      "step": 13690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929079717686896e-05,
+      "loss": 2.7074,
+      "step": 13695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929027665435592e-05,
+      "loss": 2.6223,
+      "step": 13700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.92897559422585e-05,
+      "loss": 2.5785,
+      "step": 13705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928923504057865e-05,
+      "loss": 2.6754,
+      "step": 13710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928871394931838e-05,
+      "loss": 2.7416,
+      "step": 13715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928819266847971e-05,
+      "loss": 2.6928,
+      "step": 13720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928767119806463e-05,
+      "loss": 2.5937,
+      "step": 13725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928714953807515e-05,
+      "loss": 2.5571,
+      "step": 13730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928662768851329e-05,
+      "loss": 2.5965,
+      "step": 13735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928610564938103e-05,
+      "loss": 2.6175,
+      "step": 13740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.92855834206804e-05,
+      "loss": 2.5534,
+      "step": 13745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928506100241339e-05,
+      "loss": 2.5784,
+      "step": 13750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928453839458204e-05,
+      "loss": 2.6905,
+      "step": 13755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928401559718833e-05,
+      "loss": 2.7548,
+      "step": 13760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928349261023428e-05,
+      "loss": 2.551,
+      "step": 13765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928296943372191e-05,
+      "loss": 2.74,
+      "step": 13770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928244606765322e-05,
+      "loss": 2.6955,
+      "step": 13775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928192251203024e-05,
+      "loss": 2.6086,
+      "step": 13780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928139876685496e-05,
+      "loss": 2.6608,
+      "step": 13785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.928087483212943e-05,
+      "loss": 2.5559,
+      "step": 13790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.92803507078556e-05,
+      "loss": 2.6936,
+      "step": 13795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927982639403557e-05,
+      "loss": 2.7316,
+      "step": 13800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.92793018906713e-05,
+      "loss": 2.6559,
+      "step": 13805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927877719776481e-05,
+      "loss": 2.5901,
+      "step": 13810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927825231531813e-05,
+      "loss": 2.6201,
+      "step": 13815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.92777272433333e-05,
+      "loss": 2.6304,
+      "step": 13820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927720198181229e-05,
+      "loss": 2.5713,
+      "step": 13825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927667653075719e-05,
+      "loss": 2.7219,
+      "step": 13830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927615089016995e-05,
+      "loss": 2.5935,
+      "step": 13835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927562506005263e-05,
+      "loss": 2.6841,
+      "step": 13840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927509904040724e-05,
+      "loss": 2.7701,
+      "step": 13845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.92745728312358e-05,
+      "loss": 2.6558,
+      "step": 13850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927404643254035e-05,
+      "loss": 2.7093,
+      "step": 13855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927351984432291e-05,
+      "loss": 2.5098,
+      "step": 13860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.92729930665855e-05,
+      "loss": 2.6781,
+      "step": 13865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927246609933015e-05,
+      "loss": 2.6344,
+      "step": 13870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927193894255888e-05,
+      "loss": 2.7722,
+      "step": 13875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927141159627373e-05,
+      "loss": 2.6638,
+      "step": 13880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927088406047671e-05,
+      "loss": 2.6002,
+      "step": 13885
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.927035633516986e-05,
+      "loss": 2.6903,
+      "step": 13890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.926982842035521e-05,
+      "loss": 2.6318,
+      "step": 13895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.926930031603479e-05,
+      "loss": 2.5438,
+      "step": 13900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.926877202221063e-05,
+      "loss": 2.6905,
+      "step": 13905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.926824353888478e-05,
+      "loss": 2.6947,
+      "step": 13910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.926771486605924e-05,
+      "loss": 2.6785,
+      "step": 13915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.926718600373605e-05,
+      "loss": 2.6796,
+      "step": 13920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.926665695191729e-05,
+      "loss": 2.6176,
+      "step": 13925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.926612771060492e-05,
+      "loss": 2.6069,
+      "step": 13930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.926559827980102e-05,
+      "loss": 2.5718,
+      "step": 13935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926506865950764e-05,
+      "loss": 2.5942,
+      "step": 13940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926453884972678e-05,
+      "loss": 2.6079,
+      "step": 13945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92640088504605e-05,
+      "loss": 2.5535,
+      "step": 13950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926347866171083e-05,
+      "loss": 2.6111,
+      "step": 13955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926294828347982e-05,
+      "loss": 2.5597,
+      "step": 13960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92624177157695e-05,
+      "loss": 2.7208,
+      "step": 13965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926188695858191e-05,
+      "loss": 2.6263,
+      "step": 13970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926135601191909e-05,
+      "loss": 2.6374,
+      "step": 13975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92608248757831e-05,
+      "loss": 2.7126,
+      "step": 13980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926029355017597e-05,
+      "loss": 2.6215,
+      "step": 13985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925976203509973e-05,
+      "loss": 2.6384,
+      "step": 13990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925923033055645e-05,
+      "loss": 2.6253,
+      "step": 13995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925869843654816e-05,
+      "loss": 2.7006,
+      "step": 14000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92581663530769e-05,
+      "loss": 2.5817,
+      "step": 14005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925763408014473e-05,
+      "loss": 2.7941,
+      "step": 14010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92571016177537e-05,
+      "loss": 2.6655,
+      "step": 14015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925656896590586e-05,
+      "loss": 2.7142,
+      "step": 14020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925603612460323e-05,
+      "loss": 2.6462,
+      "step": 14025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925550309384788e-05,
+      "loss": 2.6354,
+      "step": 14030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925496987364187e-05,
+      "loss": 2.5984,
+      "step": 14035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925443646398722e-05,
+      "loss": 2.7303,
+      "step": 14040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925390286488602e-05,
+      "loss": 2.7498,
+      "step": 14045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925336907634031e-05,
+      "loss": 2.6622,
+      "step": 14050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925283509835213e-05,
+      "loss": 2.6118,
+      "step": 14055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925230093092352e-05,
+      "loss": 2.759,
+      "step": 14060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925176657405658e-05,
+      "loss": 2.6475,
+      "step": 14065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925123202775334e-05,
+      "loss": 2.7414,
+      "step": 14070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925069729201585e-05,
+      "loss": 2.6294,
+      "step": 14075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925016236684616e-05,
+      "loss": 2.7048,
+      "step": 14080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924962725224637e-05,
+      "loss": 2.5628,
+      "step": 14085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92490919482185e-05,
+      "loss": 2.7212,
+      "step": 14090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924855645476461e-05,
+      "loss": 2.6865,
+      "step": 14095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924802077188677e-05,
+      "loss": 2.6341,
+      "step": 14100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924748489958704e-05,
+      "loss": 2.5578,
+      "step": 14105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924694883786747e-05,
+      "loss": 2.7665,
+      "step": 14110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924641258673015e-05,
+      "loss": 2.6782,
+      "step": 14115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924587614617711e-05,
+      "loss": 2.577,
+      "step": 14120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924533951621044e-05,
+      "loss": 2.565,
+      "step": 14125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924480269683217e-05,
+      "loss": 2.7286,
+      "step": 14130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92442656880444e-05,
+      "loss": 2.7068,
+      "step": 14135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924372848984917e-05,
+      "loss": 2.7101,
+      "step": 14140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924319110224858e-05,
+      "loss": 2.62,
+      "step": 14145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924265352524467e-05,
+      "loss": 2.5702,
+      "step": 14150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92421157588395e-05,
+      "loss": 2.6974,
+      "step": 14155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924157780303516e-05,
+      "loss": 2.5488,
+      "step": 14160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924103965783369e-05,
+      "loss": 2.7557,
+      "step": 14165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924050132323719e-05,
+      "loss": 2.7,
+      "step": 14170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923996279924773e-05,
+      "loss": 2.5807,
+      "step": 14175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923942408586738e-05,
+      "loss": 2.5726,
+      "step": 14180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923888518309817e-05,
+      "loss": 2.6279,
+      "step": 14185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923834609094223e-05,
+      "loss": 2.5916,
+      "step": 14190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92378068094016e-05,
+      "loss": 2.6829,
+      "step": 14195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923726733847838e-05,
+      "loss": 2.5234,
+      "step": 14200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923672767817462e-05,
+      "loss": 2.6226,
+      "step": 14205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92361878284924e-05,
+      "loss": 2.6568,
+      "step": 14210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92356477894338e-05,
+      "loss": 2.641,
+      "step": 14215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92351075610009e-05,
+      "loss": 2.5903,
+      "step": 14220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923456714319577e-05,
+      "loss": 2.6315,
+      "step": 14225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923402653602052e-05,
+      "loss": 2.6545,
+      "step": 14230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923348573947718e-05,
+      "loss": 2.6698,
+      "step": 14235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923294475356786e-05,
+      "loss": 2.551,
+      "step": 14240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923240357829465e-05,
+      "loss": 2.625,
+      "step": 14245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923186221365959e-05,
+      "loss": 2.5225,
+      "step": 14250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923132065966481e-05,
+      "loss": 2.6523,
+      "step": 14255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923077891631236e-05,
+      "loss": 2.4874,
+      "step": 14260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923023698360434e-05,
+      "loss": 2.6717,
+      "step": 14265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922969486154282e-05,
+      "loss": 2.5446,
+      "step": 14270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92291525501299e-05,
+      "loss": 2.6414,
+      "step": 14275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922861004936766e-05,
+      "loss": 2.6354,
+      "step": 14280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92280673592582e-05,
+      "loss": 2.5385,
+      "step": 14285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922752447980359e-05,
+      "loss": 2.5941,
+      "step": 14290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922698141100592e-05,
+      "loss": 2.5526,
+      "step": 14295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922643815286727e-05,
+      "loss": 2.7106,
+      "step": 14300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922589470538976e-05,
+      "loss": 2.6714,
+      "step": 14305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922535106857544e-05,
+      "loss": 2.8272,
+      "step": 14310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922480724242644e-05,
+      "loss": 2.6177,
+      "step": 14315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922426322694482e-05,
+      "loss": 2.6274,
+      "step": 14320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92237190221327e-05,
+      "loss": 2.5952,
+      "step": 14325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922317462799216e-05,
+      "loss": 2.7771,
+      "step": 14330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92226300445253e-05,
+      "loss": 2.7469,
+      "step": 14335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922208527173419e-05,
+      "loss": 2.6412,
+      "step": 14340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922154030962096e-05,
+      "loss": 2.6842,
+      "step": 14345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922099515818767e-05,
+      "loss": 2.6964,
+      "step": 14350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922044981743645e-05,
+      "loss": 2.7432,
+      "step": 14355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921990428736938e-05,
+      "loss": 2.6005,
+      "step": 14360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921935856798856e-05,
+      "loss": 2.5219,
+      "step": 14365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92188126592961e-05,
+      "loss": 2.6623,
+      "step": 14370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921826656129409e-05,
+      "loss": 2.8099,
+      "step": 14375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921772027398462e-05,
+      "loss": 2.6168,
+      "step": 14380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921717379736982e-05,
+      "loss": 2.7138,
+      "step": 14385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921662713145174e-05,
+      "loss": 2.6926,
+      "step": 14390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921608027623255e-05,
+      "loss": 2.5378,
+      "step": 14395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921553323171431e-05,
+      "loss": 2.6707,
+      "step": 14400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921498599789913e-05,
+      "loss": 2.6213,
+      "step": 14405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921443857478913e-05,
+      "loss": 2.625,
+      "step": 14410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92138909623864e-05,
+      "loss": 2.6631,
+      "step": 14415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921334316069306e-05,
+      "loss": 2.6235,
+      "step": 14420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92127951697112e-05,
+      "loss": 2.62,
+      "step": 14425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921224698944293e-05,
+      "loss": 2.7185,
+      "step": 14430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921169861989037e-05,
+      "loss": 2.4737,
+      "step": 14435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921115006105562e-05,
+      "loss": 2.7115,
+      "step": 14440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92106013129408e-05,
+      "loss": 2.5701,
+      "step": 14445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921005237554801e-05,
+      "loss": 2.7485,
+      "step": 14450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920950324887937e-05,
+      "loss": 2.632,
+      "step": 14455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920895393293699e-05,
+      "loss": 2.723,
+      "step": 14460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920840442772296e-05,
+      "loss": 2.6064,
+      "step": 14465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920785473323944e-05,
+      "loss": 2.6804,
+      "step": 14470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920730484948851e-05,
+      "loss": 2.6257,
+      "step": 14475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920675477647228e-05,
+      "loss": 2.6224,
+      "step": 14480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920620451419289e-05,
+      "loss": 2.5844,
+      "step": 14485
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920565406265245e-05,
+      "loss": 2.59,
+      "step": 14490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920510342185307e-05,
+      "loss": 2.6407,
+      "step": 14495
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920455259179686e-05,
+      "loss": 2.7454,
+      "step": 14500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920400157248596e-05,
+      "loss": 2.61,
+      "step": 14505
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920345036392248e-05,
+      "loss": 2.6761,
+      "step": 14510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920289896610852e-05,
+      "loss": 2.552,
+      "step": 14515
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920234737904624e-05,
+      "loss": 2.6653,
+      "step": 14520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920179560273774e-05,
+      "loss": 2.7182,
+      "step": 14525
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920124363718513e-05,
+      "loss": 2.7322,
+      "step": 14530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920069148239057e-05,
+      "loss": 2.6566,
+      "step": 14535
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920013913835614e-05,
+      "loss": 2.6407,
+      "step": 14540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919958660508399e-05,
+      "loss": 2.6216,
+      "step": 14545
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919903388257623e-05,
+      "loss": 2.4707,
+      "step": 14550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919848097083501e-05,
+      "loss": 2.6063,
+      "step": 14555
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919792786986244e-05,
+      "loss": 2.6447,
+      "step": 14560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919737457966066e-05,
+      "loss": 2.6466,
+      "step": 14565
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919682110023176e-05,
+      "loss": 2.629,
+      "step": 14570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919626743157792e-05,
+      "loss": 2.7072,
+      "step": 14575
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919571357370123e-05,
+      "loss": 2.7481,
+      "step": 14580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919515952660386e-05,
+      "loss": 2.6306,
+      "step": 14585
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919460529028789e-05,
+      "loss": 2.6097,
+      "step": 14590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91940508647555e-05,
+      "loss": 2.6612,
+      "step": 14595
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919349625000879e-05,
+      "loss": 2.6533,
+      "step": 14600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919294144604991e-05,
+      "loss": 2.5757,
+      "step": 14605
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919238645288098e-05,
+      "loss": 2.6507,
+      "step": 14610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919183127050415e-05,
+      "loss": 2.7005,
+      "step": 14615
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919127589892157e-05,
+      "loss": 2.6021,
+      "step": 14620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919072033813532e-05,
+      "loss": 2.7171,
+      "step": 14625
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919016458814759e-05,
+      "loss": 2.6517,
+      "step": 14630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91896086489605e-05,
+      "loss": 2.7487,
+      "step": 14635
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918905252057619e-05,
+      "loss": 2.673,
+      "step": 14640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918849620299678e-05,
+      "loss": 2.6495,
+      "step": 14645
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918793969622444e-05,
+      "loss": 2.7583,
+      "step": 14650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91873830002613e-05,
+      "loss": 2.7354,
+      "step": 14655
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91868261151095e-05,
+      "loss": 2.5927,
+      "step": 14660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918626904077117e-05,
+      "loss": 2.6415,
+      "step": 14665
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918571177724846e-05,
+      "loss": 2.8059,
+      "step": 14670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918515432454353e-05,
+      "loss": 2.5839,
+      "step": 14675
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918459668265851e-05,
+      "loss": 2.6169,
+      "step": 14680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918403885159553e-05,
+      "loss": 2.6265,
+      "step": 14685
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918348083135676e-05,
+      "loss": 2.7174,
+      "step": 14690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918292262194435e-05,
+      "loss": 2.6814,
+      "step": 14695
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918236422336043e-05,
+      "loss": 2.6653,
+      "step": 14700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918180563560713e-05,
+      "loss": 2.5181,
+      "step": 14705
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918124685868663e-05,
+      "loss": 2.5985,
+      "step": 14710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918068789260109e-05,
+      "loss": 2.6898,
+      "step": 14715
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918012873735262e-05,
+      "loss": 2.7273,
+      "step": 14720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917956939294339e-05,
+      "loss": 2.544,
+      "step": 14725
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917900985937556e-05,
+      "loss": 2.5235,
+      "step": 14730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917845013665125e-05,
+      "loss": 2.583,
+      "step": 14735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917789022477267e-05,
+      "loss": 2.5045,
+      "step": 14740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917733012374193e-05,
+      "loss": 2.6604,
+      "step": 14745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917676983356118e-05,
+      "loss": 2.4764,
+      "step": 14750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917620935423261e-05,
+      "loss": 2.706,
+      "step": 14755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917564868575835e-05,
+      "loss": 2.747,
+      "step": 14760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917508782814055e-05,
+      "loss": 2.6106,
+      "step": 14765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91745267813814e-05,
+      "loss": 2.6008,
+      "step": 14770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917396554548301e-05,
+      "loss": 2.6666,
+      "step": 14775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917340412044759e-05,
+      "loss": 2.5578,
+      "step": 14780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917284250627726e-05,
+      "loss": 2.6008,
+      "step": 14785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91722807029742e-05,
+      "loss": 2.6635,
+      "step": 14790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917171871054057e-05,
+      "loss": 2.7524,
+      "step": 14795
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917115652897853e-05,
+      "loss": 2.6458,
+      "step": 14800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917059415829024e-05,
+      "loss": 2.6836,
+      "step": 14805
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917003159847784e-05,
+      "loss": 2.6572,
+      "step": 14810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916946884954355e-05,
+      "loss": 2.5695,
+      "step": 14815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916890591148948e-05,
+      "loss": 2.7283,
+      "step": 14820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916834278431781e-05,
+      "loss": 2.6159,
+      "step": 14825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916777946803074e-05,
+      "loss": 2.6423,
+      "step": 14830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916721596263038e-05,
+      "loss": 2.5733,
+      "step": 14835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916665226811894e-05,
+      "loss": 2.7503,
+      "step": 14840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916608838449859e-05,
+      "loss": 2.665,
+      "step": 14845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916552431177147e-05,
+      "loss": 2.5689,
+      "step": 14850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916496004993976e-05,
+      "loss": 2.639,
+      "step": 14855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916439559900564e-05,
+      "loss": 2.5941,
+      "step": 14860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916383095897125e-05,
+      "loss": 2.7724,
+      "step": 14865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916326612983881e-05,
+      "loss": 2.6695,
+      "step": 14870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916270111161046e-05,
+      "loss": 2.7742,
+      "step": 14875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916213590428839e-05,
+      "loss": 2.7664,
+      "step": 14880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916157050787476e-05,
+      "loss": 2.722,
+      "step": 14885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916100492237175e-05,
+      "loss": 2.6636,
+      "step": 14890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916043914778155e-05,
+      "loss": 2.6017,
+      "step": 14895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915987318410631e-05,
+      "loss": 2.5791,
+      "step": 14900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915930703134821e-05,
+      "loss": 2.5967,
+      "step": 14905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915874068950946e-05,
+      "loss": 2.5629,
+      "step": 14910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915817415859219e-05,
+      "loss": 2.6235,
+      "step": 14915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91576074385986e-05,
+      "loss": 2.5862,
+      "step": 14920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91570405295309e-05,
+      "loss": 2.7688,
+      "step": 14925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915647343139123e-05,
+      "loss": 2.61,
+      "step": 14930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915590614418179e-05,
+      "loss": 2.5485,
+      "step": 14935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915533866790476e-05,
+      "loss": 2.6223,
+      "step": 14940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915477100256232e-05,
+      "loss": 2.6446,
+      "step": 14945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915420314815664e-05,
+      "loss": 2.7004,
+      "step": 14950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915363510468993e-05,
+      "loss": 2.7414,
+      "step": 14955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915306687216435e-05,
+      "loss": 2.6462,
+      "step": 14960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915249845058211e-05,
+      "loss": 2.7031,
+      "step": 14965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915192983994538e-05,
+      "loss": 2.6175,
+      "step": 14970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915136104025635e-05,
+      "loss": 2.7152,
+      "step": 14975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915079205151721e-05,
+      "loss": 2.5215,
+      "step": 14980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915022287373014e-05,
+      "loss": 2.5679,
+      "step": 14985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914965350689736e-05,
+      "loss": 2.6447,
+      "step": 14990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914908395102103e-05,
+      "loss": 2.6886,
+      "step": 14995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914851420610332e-05,
+      "loss": 2.5582,
+      "step": 15000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914794427214648e-05,
+      "loss": 2.674,
+      "step": 15005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914737414915265e-05,
+      "loss": 2.6427,
+      "step": 15010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914680383712406e-05,
+      "loss": 2.6626,
+      "step": 15015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914623333606288e-05,
+      "loss": 2.6139,
+      "step": 15020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914566264597131e-05,
+      "loss": 2.5373,
+      "step": 15025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914509176685155e-05,
+      "loss": 2.6659,
+      "step": 15030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914452069870579e-05,
+      "loss": 2.7313,
+      "step": 15035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914394944153622e-05,
+      "loss": 2.5578,
+      "step": 15040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914337799534506e-05,
+      "loss": 2.6329,
+      "step": 15045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914280636013449e-05,
+      "loss": 2.573,
+      "step": 15050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914223453590672e-05,
+      "loss": 2.631,
+      "step": 15055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914166252266393e-05,
+      "loss": 2.6606,
+      "step": 15060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914109032040834e-05,
+      "loss": 2.6323,
+      "step": 15065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914051792914214e-05,
+      "loss": 2.598,
+      "step": 15070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913994534886753e-05,
+      "loss": 2.5562,
+      "step": 15075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913937257958673e-05,
+      "loss": 2.6943,
+      "step": 15080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913879962130192e-05,
+      "loss": 2.5281,
+      "step": 15085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913822647401532e-05,
+      "loss": 2.6877,
+      "step": 15090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913765313772912e-05,
+      "loss": 2.5295,
+      "step": 15095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913707961244554e-05,
+      "loss": 2.7313,
+      "step": 15100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913650589816679e-05,
+      "loss": 2.6137,
+      "step": 15105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913593199489504e-05,
+      "loss": 2.5627,
+      "step": 15110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913535790263254e-05,
+      "loss": 2.7933,
+      "step": 15115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913478362138149e-05,
+      "loss": 2.6441,
+      "step": 15120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913420915114407e-05,
+      "loss": 2.6995,
+      "step": 15125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913363449192252e-05,
+      "loss": 2.6712,
+      "step": 15130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913305964371904e-05,
+      "loss": 2.716,
+      "step": 15135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913248460653583e-05,
+      "loss": 2.5221,
+      "step": 15140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913190938037513e-05,
+      "loss": 2.6447,
+      "step": 15145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913133396523914e-05,
+      "loss": 2.5066,
+      "step": 15150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913075836113006e-05,
+      "loss": 2.6011,
+      "step": 15155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913018256805011e-05,
+      "loss": 2.6104,
+      "step": 15160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912960658600151e-05,
+      "loss": 2.6537,
+      "step": 15165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912903041498647e-05,
+      "loss": 2.6387,
+      "step": 15170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912845405500722e-05,
+      "loss": 2.7024,
+      "step": 15175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912787750606595e-05,
+      "loss": 2.7162,
+      "step": 15180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91273007681649e-05,
+      "loss": 2.6271,
+      "step": 15185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912672384130628e-05,
+      "loss": 2.6903,
+      "step": 15190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912614672549232e-05,
+      "loss": 2.7285,
+      "step": 15195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912556942072522e-05,
+      "loss": 2.5853,
+      "step": 15200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912499192700722e-05,
+      "loss": 2.6096,
+      "step": 15205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912441424434053e-05,
+      "loss": 2.6641,
+      "step": 15210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912383637272736e-05,
+      "loss": 2.6041,
+      "step": 15215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912325831216996e-05,
+      "loss": 2.685,
+      "step": 15220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912268006267054e-05,
+      "loss": 2.3992,
+      "step": 15225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912210162423132e-05,
+      "loss": 2.6198,
+      "step": 15230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912152299685454e-05,
+      "loss": 2.6605,
+      "step": 15235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912094418054241e-05,
+      "loss": 2.5482,
+      "step": 15240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912036517529715e-05,
+      "loss": 2.676,
+      "step": 15245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9119785981121e-05,
+      "loss": 2.67,
+      "step": 15250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911920659801619e-05,
+      "loss": 2.6766,
+      "step": 15255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911862702598496e-05,
+      "loss": 2.6157,
+      "step": 15260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911804726502949e-05,
+      "loss": 2.6439,
+      "step": 15265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911746731515206e-05,
+      "loss": 2.7885,
+      "step": 15270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91168871763549e-05,
+      "loss": 2.6647,
+      "step": 15275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911630684864021e-05,
+      "loss": 2.7618,
+      "step": 15280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911572633201023e-05,
+      "loss": 2.4864,
+      "step": 15285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911514562646721e-05,
+      "loss": 2.6856,
+      "step": 15290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911456473201338e-05,
+      "loss": 2.6675,
+      "step": 15295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911398364865096e-05,
+      "loss": 2.62,
+      "step": 15300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911340237638218e-05,
+      "loss": 2.6719,
+      "step": 15305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91128209152093e-05,
+      "loss": 2.6086,
+      "step": 15310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911223926513456e-05,
+      "loss": 2.6127,
+      "step": 15315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911165742616018e-05,
+      "loss": 2.5163,
+      "step": 15320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911107539828838e-05,
+      "loss": 2.7521,
+      "step": 15325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911049318152144e-05,
+      "loss": 2.6633,
+      "step": 15330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910991077586156e-05,
+      "loss": 2.6236,
+      "step": 15335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9109328181311e-05,
+      "loss": 2.6153,
+      "step": 15340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910874539787201e-05,
+      "loss": 2.6149,
+      "step": 15345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910816242554681e-05,
+      "loss": 2.7496,
+      "step": 15350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910757926433767e-05,
+      "loss": 2.6849,
+      "step": 15355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910699591424682e-05,
+      "loss": 2.6343,
+      "step": 15360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910641237527647e-05,
+      "loss": 2.6857,
+      "step": 15365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910582864742892e-05,
+      "loss": 2.7095,
+      "step": 15370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910524473070637e-05,
+      "loss": 2.7781,
+      "step": 15375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91046606251111e-05,
+      "loss": 2.6091,
+      "step": 15380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910407633064533e-05,
+      "loss": 2.7123,
+      "step": 15385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910349184731134e-05,
+      "loss": 2.7454,
+      "step": 15390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910290717511134e-05,
+      "loss": 2.725,
+      "step": 15395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910232231404762e-05,
+      "loss": 2.6049,
+      "step": 15400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910173726412238e-05,
+      "loss": 2.5209,
+      "step": 15405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91011520253379e-05,
+      "loss": 2.6212,
+      "step": 15410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910056659769644e-05,
+      "loss": 2.4975,
+      "step": 15415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909998098120021e-05,
+      "loss": 2.6569,
+      "step": 15420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909939517585152e-05,
+      "loss": 2.7829,
+      "step": 15425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909880918165258e-05,
+      "loss": 2.6189,
+      "step": 15430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909822299860567e-05,
+      "loss": 2.7171,
+      "step": 15435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909763662671303e-05,
+      "loss": 2.6601,
+      "step": 15440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909705006597692e-05,
+      "loss": 2.7111,
+      "step": 15445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909646331639958e-05,
+      "loss": 2.591,
+      "step": 15450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90958763779833e-05,
+      "loss": 2.5098,
+      "step": 15455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90952892507303e-05,
+      "loss": 2.7473,
+      "step": 15460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909470193464288e-05,
+      "loss": 2.6208,
+      "step": 15465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909411442972326e-05,
+      "loss": 2.7275,
+      "step": 15470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909352673597372e-05,
+      "loss": 2.5667,
+      "step": 15475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909293885339652e-05,
+      "loss": 2.6622,
+      "step": 15480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90923507819939e-05,
+      "loss": 2.717,
+      "step": 15485
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909176252176817e-05,
+      "loss": 2.6017,
+      "step": 15490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909117407272153e-05,
+      "loss": 2.5592,
+      "step": 15495
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90905854348563e-05,
+      "loss": 2.6342,
+      "step": 15500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90899966081747e-05,
+      "loss": 2.6576,
+      "step": 15505
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908940759267904e-05,
+      "loss": 2.644,
+      "step": 15510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908881838837154e-05,
+      "loss": 2.7942,
+      "step": 15515
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908822899525448e-05,
+      "loss": 2.5466,
+      "step": 15520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908763941333014e-05,
+      "loss": 2.6029,
+      "step": 15525
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908704964260078e-05,
+      "loss": 2.7017,
+      "step": 15530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908645968306868e-05,
+      "loss": 2.7402,
+      "step": 15535
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908586953473609e-05,
+      "loss": 2.7692,
+      "step": 15540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908527919760528e-05,
+      "loss": 2.5899,
+      "step": 15545
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908468867167855e-05,
+      "loss": 2.6605,
+      "step": 15550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908409795695813e-05,
+      "loss": 2.664,
+      "step": 15555
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908350705344631e-05,
+      "loss": 2.6306,
+      "step": 15560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908291596114539e-05,
+      "loss": 2.6187,
+      "step": 15565
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908232468005759e-05,
+      "loss": 2.6823,
+      "step": 15570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908173321018523e-05,
+      "loss": 2.574,
+      "step": 15575
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908114155153056e-05,
+      "loss": 2.5748,
+      "step": 15580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908054970409586e-05,
+      "loss": 2.6998,
+      "step": 15585
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907995766788341e-05,
+      "loss": 2.5849,
+      "step": 15590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907936544289549e-05,
+      "loss": 2.6372,
+      "step": 15595
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907877302913437e-05,
+      "loss": 2.6139,
+      "step": 15600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907818042660236e-05,
+      "loss": 2.6708,
+      "step": 15605
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907758763530168e-05,
+      "loss": 2.5355,
+      "step": 15610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907699465523465e-05,
+      "loss": 2.5547,
+      "step": 15615
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907640148640354e-05,
+      "loss": 2.5619,
+      "step": 15620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907580812881065e-05,
+      "loss": 2.589,
+      "step": 15625
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907521458245824e-05,
+      "loss": 2.6441,
+      "step": 15630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90746208473486e-05,
+      "loss": 2.6212,
+      "step": 15635
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907402692348401e-05,
+      "loss": 2.6214,
+      "step": 15640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907343281086677e-05,
+      "loss": 2.5301,
+      "step": 15645
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907283850949913e-05,
+      "loss": 2.5917,
+      "step": 15650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907224401938342e-05,
+      "loss": 2.8077,
+      "step": 15655
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90716493405219e-05,
+      "loss": 2.651,
+      "step": 15660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907105447291686e-05,
+      "loss": 2.6139,
+      "step": 15665
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907045941657058e-05,
+      "loss": 2.5969,
+      "step": 15670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906986417148538e-05,
+      "loss": 2.5635,
+      "step": 15675
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906926873766352e-05,
+      "loss": 2.5573,
+      "step": 15680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906867311510731e-05,
+      "loss": 2.5911,
+      "step": 15685
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906807730381902e-05,
+      "loss": 2.7282,
+      "step": 15690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906748130380095e-05,
+      "loss": 2.5488,
+      "step": 15695
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90668851150554e-05,
+      "loss": 2.5966,
+      "step": 15700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906628873758465e-05,
+      "loss": 2.7248,
+      "step": 15705
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9065692171391e-05,
+      "loss": 2.5884,
+      "step": 15710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906509541647675e-05,
+      "loss": 2.7367,
+      "step": 15715
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90644984728442e-05,
+      "loss": 2.6073,
+      "step": 15720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906390134049563e-05,
+      "loss": 2.592,
+      "step": 15725
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906330401943335e-05,
+      "loss": 2.7562,
+      "step": 15730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906270650965964e-05,
+      "loss": 2.6438,
+      "step": 15735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90621088111768e-05,
+      "loss": 2.7581,
+      "step": 15740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906151092398717e-05,
+      "loss": 2.525,
+      "step": 15745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9060912848093e-05,
+      "loss": 2.7119,
+      "step": 15750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90603145834966e-05,
+      "loss": 2.6232,
+      "step": 15755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905971613020031e-05,
+      "loss": 2.6564,
+      "step": 15760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905911748820638e-05,
+      "loss": 2.7438,
+      "step": 15765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905851865751713e-05,
+      "loss": 2.5448,
+      "step": 15770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905791963813486e-05,
+      "loss": 2.5827,
+      "step": 15775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90573204300619e-05,
+      "loss": 2.5441,
+      "step": 15780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905672103330054e-05,
+      "loss": 2.826,
+      "step": 15785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905612144785307e-05,
+      "loss": 2.6225,
+      "step": 15790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90555216737218e-05,
+      "loss": 2.5454,
+      "step": 15795
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905492171090907e-05,
+      "loss": 2.5553,
+      "step": 15800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905432155941715e-05,
+      "loss": 2.6807,
+      "step": 15805
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905372121924835e-05,
+      "loss": 2.6826,
+      "step": 15810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905312069040499e-05,
+      "loss": 2.4932,
+      "step": 15815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905251997288938e-05,
+      "loss": 2.6356,
+      "step": 15820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905191906670383e-05,
+      "loss": 2.6217,
+      "step": 15825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905131797185067e-05,
+      "loss": 2.7673,
+      "step": 15830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905071668833217e-05,
+      "loss": 2.6327,
+      "step": 15835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905011521615069e-05,
+      "loss": 2.6099,
+      "step": 15840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904951355530849e-05,
+      "loss": 2.5683,
+      "step": 15845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904891170580794e-05,
+      "loss": 2.6933,
+      "step": 15850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904830966765131e-05,
+      "loss": 2.5742,
+      "step": 15855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904770744084094e-05,
+      "loss": 2.7296,
+      "step": 15860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904710502537914e-05,
+      "loss": 2.6652,
+      "step": 15865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904650242126823e-05,
+      "loss": 2.5894,
+      "step": 15870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904589962851052e-05,
+      "loss": 2.7711,
+      "step": 15875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904529664710834e-05,
+      "loss": 2.6583,
+      "step": 15880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9044693477064e-05,
+      "loss": 2.6208,
+      "step": 15885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904409011837982e-05,
+      "loss": 2.7253,
+      "step": 15890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904348657105814e-05,
+      "loss": 2.6136,
+      "step": 15895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904288283510124e-05,
+      "loss": 2.4199,
+      "step": 15900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904227891051147e-05,
+      "loss": 2.5575,
+      "step": 15905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904167479729117e-05,
+      "loss": 2.803,
+      "step": 15910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904107049544263e-05,
+      "loss": 2.4564,
+      "step": 15915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904046600496819e-05,
+      "loss": 2.68,
+      "step": 15920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903986132587018e-05,
+      "loss": 2.7728,
+      "step": 15925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90392564581509e-05,
+      "loss": 2.6884,
+      "step": 15930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90386514018127e-05,
+      "loss": 2.5573,
+      "step": 15935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903804615685793e-05,
+      "loss": 2.7688,
+      "step": 15940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903744072328886e-05,
+      "loss": 2.7146,
+      "step": 15945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903683510110785e-05,
+      "loss": 2.4802,
+      "step": 15950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903622929031723e-05,
+      "loss": 2.7737,
+      "step": 15955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903562329091934e-05,
+      "loss": 2.4767,
+      "step": 15960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90350171029165e-05,
+      "loss": 2.644,
+      "step": 15965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903441072631104e-05,
+      "loss": 2.6285,
+      "step": 15970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903380416110529e-05,
+      "loss": 2.6251,
+      "step": 15975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903319740730156e-05,
+      "loss": 2.8247,
+      "step": 15980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903259046490226e-05,
+      "loss": 2.6727,
+      "step": 15985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903198333390964e-05,
+      "loss": 2.5835,
+      "step": 15990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903137601432609e-05,
+      "loss": 2.6169,
+      "step": 15995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903076850615389e-05,
+      "loss": 2.6824,
+      "step": 16000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903016080939544e-05,
+      "loss": 2.6284,
+      "step": 16005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902955292405305e-05,
+      "loss": 2.595,
+      "step": 16010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902894485012906e-05,
+      "loss": 2.7581,
+      "step": 16015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90283365876258e-05,
+      "loss": 2.8192,
+      "step": 16020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902772813654562e-05,
+      "loss": 2.6915,
+      "step": 16025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902711949689086e-05,
+      "loss": 2.7401,
+      "step": 16030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902651066866385e-05,
+      "loss": 2.7078,
+      "step": 16035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902590165186692e-05,
+      "loss": 2.6534,
+      "step": 16040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902529244650245e-05,
+      "loss": 2.6298,
+      "step": 16045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902468305257277e-05,
+      "loss": 2.6882,
+      "step": 16050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902407347008022e-05,
+      "loss": 2.7229,
+      "step": 16055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902346369902714e-05,
+      "loss": 2.6316,
+      "step": 16060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902285373941586e-05,
+      "loss": 2.5656,
+      "step": 16065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902224359124876e-05,
+      "loss": 2.62,
+      "step": 16070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902163325452817e-05,
+      "loss": 2.7298,
+      "step": 16075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902102272925645e-05,
+      "loss": 2.5623,
+      "step": 16080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902041201543593e-05,
+      "loss": 2.587,
+      "step": 16085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901980111306897e-05,
+      "loss": 2.6415,
+      "step": 16090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901919002215792e-05,
+      "loss": 2.6827,
+      "step": 16095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90185787427051e-05,
+      "loss": 2.4588,
+      "step": 16100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901796727471291e-05,
+      "loss": 2.5994,
+      "step": 16105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901735561818367e-05,
+      "loss": 2.6155,
+      "step": 16110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901674377311976e-05,
+      "loss": 2.8359,
+      "step": 16115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90161317395235e-05,
+      "loss": 2.5918,
+      "step": 16120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901551951739725e-05,
+      "loss": 2.6736,
+      "step": 16125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90149071067434e-05,
+      "loss": 2.7025,
+      "step": 16130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901429450756425e-05,
+      "loss": 2.6345,
+      "step": 16135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901368171986221e-05,
+      "loss": 2.819,
+      "step": 16140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90130687436396e-05,
+      "loss": 2.7681,
+      "step": 16145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901245557889879e-05,
+      "loss": 2.403,
+      "step": 16150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901184222564215e-05,
+      "loss": 2.6758,
+      "step": 16155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901122868387201e-05,
+      "loss": 2.7074,
+      "step": 16160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901061495359076e-05,
+      "loss": 2.8027,
+      "step": 16165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901000103480075e-05,
+      "loss": 2.582,
+      "step": 16170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900938692750435e-05,
+      "loss": 2.6607,
+      "step": 16175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90087726317039e-05,
+      "loss": 2.5569,
+      "step": 16180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900815814740176e-05,
+      "loss": 2.5689,
+      "step": 16185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900754347460033e-05,
+      "loss": 2.5962,
+      "step": 16190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900692861330194e-05,
+      "loss": 2.6156,
+      "step": 16195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900631356350897e-05,
+      "loss": 2.6277,
+      "step": 16200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900569832522378e-05,
+      "loss": 2.5475,
+      "step": 16205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900508289844874e-05,
+      "loss": 2.569,
+      "step": 16210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900446728318621e-05,
+      "loss": 2.6315,
+      "step": 16215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90038514794386e-05,
+      "loss": 2.6212,
+      "step": 16220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90032354872082e-05,
+      "loss": 2.6515,
+      "step": 16225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900261930649744e-05,
+      "loss": 2.6009,
+      "step": 16230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900200293730868e-05,
+      "loss": 2.6948,
+      "step": 16235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900138637964427e-05,
+      "loss": 2.5501,
+      "step": 16240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900076963350659e-05,
+      "loss": 2.6428,
+      "step": 16245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900015269889804e-05,
+      "loss": 2.6322,
+      "step": 16250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899953557582096e-05,
+      "loss": 2.6976,
+      "step": 16255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899891826427772e-05,
+      "loss": 2.7834,
+      "step": 16260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899830076427071e-05,
+      "loss": 2.5312,
+      "step": 16265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.89976830758023e-05,
+      "loss": 2.4941,
+      "step": 16270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899706519887488e-05,
+      "loss": 2.7024,
+      "step": 16275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899644713349081e-05,
+      "loss": 2.7224,
+      "step": 16280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899582887965248e-05,
+      "loss": 2.616,
+      "step": 16285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899521043736223e-05,
+      "loss": 2.5691,
+      "step": 16290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899459180662251e-05,
+      "loss": 2.7066,
+      "step": 16295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899397298743564e-05,
+      "loss": 2.5149,
+      "step": 16300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899335397980401e-05,
+      "loss": 2.574,
+      "step": 16305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899273478373001e-05,
+      "loss": 2.6699,
+      "step": 16310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899211539921604e-05,
+      "loss": 2.6595,
+      "step": 16315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899149582626445e-05,
+      "loss": 2.7357,
+      "step": 16320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899087606487762e-05,
+      "loss": 2.6762,
+      "step": 16325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.899025611505798e-05,
+      "loss": 2.7373,
+      "step": 16330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898963597680788e-05,
+      "loss": 2.6859,
+      "step": 16335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.89890156501297e-05,
+      "loss": 2.6056,
+      "step": 16340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898839513502582e-05,
+      "loss": 2.6582,
+      "step": 16345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898777443149867e-05,
+      "loss": 2.767,
+      "step": 16350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898715353955059e-05,
+      "loss": 2.7019,
+      "step": 16355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.8986532459184e-05,
+      "loss": 2.6833,
+      "step": 16360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898591119040126e-05,
+      "loss": 2.6258,
+      "step": 16365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.89852897332048e-05,
+      "loss": 2.6149,
+      "step": 16370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898466808759696e-05,
+      "loss": 2.6764,
+      "step": 16375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898404625358018e-05,
+      "loss": 2.5442,
+      "step": 16380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898342423115683e-05,
+      "loss": 2.684,
+      "step": 16385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898280202032928e-05,
+      "loss": 2.6238,
+      "step": 16390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898217962109997e-05,
+      "loss": 2.6007,
+      "step": 16395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898155703347125e-05,
+      "loss": 2.6688,
+      "step": 16400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898093425744554e-05,
+      "loss": 2.737,
+      "step": 16405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.898031129302523e-05,
+      "loss": 2.9155,
+      "step": 16410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.897968814021273e-05,
+      "loss": 2.6064,
+      "step": 16415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.89790647990104e-05,
+      "loss": 2.6074,
+      "step": 16420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.897844126942066e-05,
+      "loss": 2.6672,
+      "step": 16425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.89778175514459e-05,
+      "loss": 2.7111,
+      "step": 16430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.897719364508856e-05,
+      "loss": 2.6638,
+      "step": 16435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.897656955035097e-05,
+      "loss": 2.5807,
+      "step": 16440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.897594526723559e-05,
+      "loss": 2.6771,
+      "step": 16445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.897532079574479e-05,
+      "loss": 2.6482,
+      "step": 16450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.8974696135881e-05,
+      "loss": 2.5246,
+      "step": 16455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.897407128764658e-05,
+      "loss": 2.6151,
+      "step": 16460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.897344625104395e-05,
+      "loss": 2.5583,
+      "step": 16465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.897282102607552e-05,
+      "loss": 2.4372,
+      "step": 16470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.897219561274371e-05,
+      "loss": 2.7006,
+      "step": 16475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89715700110509e-05,
+      "loss": 2.5745,
+      "step": 16480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.897094422099953e-05,
+      "loss": 2.7066,
+      "step": 16485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.897031824259196e-05,
+      "loss": 2.5734,
+      "step": 16490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896969207583062e-05,
+      "loss": 2.5535,
+      "step": 16495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896906572071794e-05,
+      "loss": 2.5091,
+      "step": 16500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89684391772563e-05,
+      "loss": 2.5563,
+      "step": 16505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896781244544809e-05,
+      "loss": 2.6243,
+      "step": 16510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896718552529579e-05,
+      "loss": 2.6053,
+      "step": 16515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896655841680173e-05,
+      "loss": 2.5814,
+      "step": 16520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896593111996839e-05,
+      "loss": 2.6244,
+      "step": 16525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896530363479814e-05,
+      "loss": 2.5756,
+      "step": 16530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896467596129345e-05,
+      "loss": 2.5925,
+      "step": 16535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896404809945665e-05,
+      "loss": 2.6601,
+      "step": 16540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89634200492902e-05,
+      "loss": 2.5606,
+      "step": 16545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896279181079653e-05,
+      "loss": 2.6312,
+      "step": 16550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896216338397804e-05,
+      "loss": 2.7688,
+      "step": 16555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896153476883714e-05,
+      "loss": 2.5688,
+      "step": 16560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896090596537623e-05,
+      "loss": 2.5159,
+      "step": 16565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89602769735978e-05,
+      "loss": 2.7096,
+      "step": 16570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89596477935042e-05,
+      "loss": 2.8078,
+      "step": 16575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895901842509787e-05,
+      "loss": 2.7446,
+      "step": 16580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895838886838123e-05,
+      "loss": 2.5919,
+      "step": 16585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895775912335671e-05,
+      "loss": 2.6369,
+      "step": 16590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895712919002673e-05,
+      "loss": 2.4509,
+      "step": 16595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895649906839371e-05,
+      "loss": 2.7325,
+      "step": 16600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895586875846009e-05,
+      "loss": 2.643,
+      "step": 16605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895523826022826e-05,
+      "loss": 2.6679,
+      "step": 16610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895460757370065e-05,
+      "loss": 2.7297,
+      "step": 16615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895397669887972e-05,
+      "loss": 2.6116,
+      "step": 16620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895334563576786e-05,
+      "loss": 2.5917,
+      "step": 16625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895271438436752e-05,
+      "loss": 2.6297,
+      "step": 16630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895208294468112e-05,
+      "loss": 2.6148,
+      "step": 16635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89514513167111e-05,
+      "loss": 2.6457,
+      "step": 16640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895081950045986e-05,
+      "loss": 2.8422,
+      "step": 16645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895018749592987e-05,
+      "loss": 2.6515,
+      "step": 16650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894955530312353e-05,
+      "loss": 2.6112,
+      "step": 16655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894892292204328e-05,
+      "loss": 2.6125,
+      "step": 16660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894829035269156e-05,
+      "loss": 2.6305,
+      "step": 16665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894765759507079e-05,
+      "loss": 2.6029,
+      "step": 16670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89470246491834e-05,
+      "loss": 2.5953,
+      "step": 16675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894639151503185e-05,
+      "loss": 2.6936,
+      "step": 16680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894575819261856e-05,
+      "loss": 2.6965,
+      "step": 16685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894512468194594e-05,
+      "loss": 2.6161,
+      "step": 16690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894449098301648e-05,
+      "loss": 2.6402,
+      "step": 16695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894385709583258e-05,
+      "loss": 2.6403,
+      "step": 16700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89432230203967e-05,
+      "loss": 2.6315,
+      "step": 16705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894258875671125e-05,
+      "loss": 2.5406,
+      "step": 16710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89419543047787e-05,
+      "loss": 2.7094,
+      "step": 16715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894131966460147e-05,
+      "loss": 2.5046,
+      "step": 16720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894068483618202e-05,
+      "loss": 2.7532,
+      "step": 16725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894004981952276e-05,
+      "loss": 2.6446,
+      "step": 16730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893941461462617e-05,
+      "loss": 2.6444,
+      "step": 16735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893877922149467e-05,
+      "loss": 2.6763,
+      "step": 16740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893814364013071e-05,
+      "loss": 2.5387,
+      "step": 16745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893750787053673e-05,
+      "loss": 2.4795,
+      "step": 16750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893687191271518e-05,
+      "loss": 2.6237,
+      "step": 16755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89362357666685e-05,
+      "loss": 2.6452,
+      "step": 16760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893559943239914e-05,
+      "loss": 2.5427,
+      "step": 16765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893496290990956e-05,
+      "loss": 2.6614,
+      "step": 16770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89343261992022e-05,
+      "loss": 2.6836,
+      "step": 16775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89336893002795e-05,
+      "loss": 2.6347,
+      "step": 16780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893305221314393e-05,
+      "loss": 2.7228,
+      "step": 16785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89324149377979e-05,
+      "loss": 2.6339,
+      "step": 16790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893177747424391e-05,
+      "loss": 2.6228,
+      "step": 16795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893113982248439e-05,
+      "loss": 2.6058,
+      "step": 16800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893050198252178e-05,
+      "loss": 2.6394,
+      "step": 16805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892986395435855e-05,
+      "loss": 2.62,
+      "step": 16810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892922573799714e-05,
+      "loss": 2.5235,
+      "step": 16815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892858733344003e-05,
+      "loss": 2.6142,
+      "step": 16820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892794874068966e-05,
+      "loss": 2.7245,
+      "step": 16825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892730995974848e-05,
+      "loss": 2.5799,
+      "step": 16830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892667099061896e-05,
+      "loss": 2.6451,
+      "step": 16835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892603183330354e-05,
+      "loss": 2.6348,
+      "step": 16840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89253924878047e-05,
+      "loss": 2.5583,
+      "step": 16845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892475295412489e-05,
+      "loss": 2.5705,
+      "step": 16850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892411323226656e-05,
+      "loss": 2.789,
+      "step": 16855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892347332223217e-05,
+      "loss": 2.6617,
+      "step": 16860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89228332240242e-05,
+      "loss": 2.6175,
+      "step": 16865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892219293764511e-05,
+      "loss": 2.4878,
+      "step": 16870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892155246309735e-05,
+      "loss": 2.6384,
+      "step": 16875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892091180038339e-05,
+      "loss": 2.5365,
+      "step": 16880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892027094950568e-05,
+      "loss": 2.6399,
+      "step": 16885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89196299104667e-05,
+      "loss": 2.5161,
+      "step": 16890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891898868326893e-05,
+      "loss": 2.503,
+      "step": 16895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89183472679148e-05,
+      "loss": 2.5204,
+      "step": 16900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89177056644068e-05,
+      "loss": 2.5925,
+      "step": 16905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89170638727474e-05,
+      "loss": 2.6466,
+      "step": 16910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891642189293905e-05,
+      "loss": 2.5651,
+      "step": 16915
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891577972498424e-05,
+      "loss": 2.5727,
+      "step": 16920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891513736888544e-05,
+      "loss": 2.69,
+      "step": 16925
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891449482464512e-05,
+      "loss": 2.5969,
+      "step": 16930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891385209226573e-05,
+      "loss": 2.5212,
+      "step": 16935
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891320917174975e-05,
+      "loss": 2.5165,
+      "step": 16940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891256606309966e-05,
+      "loss": 2.6079,
+      "step": 16945
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891192276631795e-05,
+      "loss": 2.6088,
+      "step": 16950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891127928140706e-05,
+      "loss": 2.7319,
+      "step": 16955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891063560836949e-05,
+      "loss": 2.7218,
+      "step": 16960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89099917472077e-05,
+      "loss": 2.557,
+      "step": 16965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890934769792417e-05,
+      "loss": 2.4974,
+      "step": 16970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89087034605214e-05,
+      "loss": 2.5713,
+      "step": 16975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890805903500183e-05,
+      "loss": 2.6062,
+      "step": 16980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890741442136797e-05,
+      "loss": 2.5452,
+      "step": 16985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890676961962228e-05,
+      "loss": 2.5404,
+      "step": 16990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890612462976725e-05,
+      "loss": 2.471,
+      "step": 16995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890547945180535e-05,
+      "loss": 2.6693,
+      "step": 17000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890483408573909e-05,
+      "loss": 2.6758,
+      "step": 17005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89041885315709e-05,
+      "loss": 2.4403,
+      "step": 17010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890354278930333e-05,
+      "loss": 2.5681,
+      "step": 17015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890289685893881e-05,
+      "loss": 2.6526,
+      "step": 17020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890225074047985e-05,
+      "loss": 2.6248,
+      "step": 17025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890160443392893e-05,
+      "loss": 2.7704,
+      "step": 17030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890095793928851e-05,
+      "loss": 2.6355,
+      "step": 17035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890031125656113e-05,
+      "loss": 2.5176,
+      "step": 17040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889966438574923e-05,
+      "loss": 2.7338,
+      "step": 17045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889901732685533e-05,
+      "loss": 2.6052,
+      "step": 17050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88983700798819e-05,
+      "loss": 2.7263,
+      "step": 17055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889772264483142e-05,
+      "loss": 2.7301,
+      "step": 17060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889707502170642e-05,
+      "loss": 2.7498,
+      "step": 17065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889642721050933e-05,
+      "loss": 2.7347,
+      "step": 17070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889577921124271e-05,
+      "loss": 2.6006,
+      "step": 17075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889513102390902e-05,
+      "loss": 2.5089,
+      "step": 17080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889448264851074e-05,
+      "loss": 2.6003,
+      "step": 17085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889383408505037e-05,
+      "loss": 2.5372,
+      "step": 17090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889318533353041e-05,
+      "loss": 2.5594,
+      "step": 17095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889253639395337e-05,
+      "loss": 2.5662,
+      "step": 17100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889188726632172e-05,
+      "loss": 2.7431,
+      "step": 17105
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889123795063797e-05,
+      "loss": 2.5861,
+      "step": 17110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889058844690463e-05,
+      "loss": 2.6171,
+      "step": 17115
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888993875512418e-05,
+      "loss": 2.5328,
+      "step": 17120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888928887529911e-05,
+      "loss": 2.5721,
+      "step": 17125
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888863880743195e-05,
+      "loss": 2.7475,
+      "step": 17130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888798855152517e-05,
+      "loss": 2.5495,
+      "step": 17135
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88873381075813e-05,
+      "loss": 2.6289,
+      "step": 17140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888668747560281e-05,
+      "loss": 2.668,
+      "step": 17145
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888603665559223e-05,
+      "loss": 2.6564,
+      "step": 17150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888538564755206e-05,
+      "loss": 2.5678,
+      "step": 17155
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888473445148477e-05,
+      "loss": 2.6561,
+      "step": 17160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888408306739291e-05,
+      "loss": 2.5641,
+      "step": 17165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888343149527896e-05,
+      "loss": 2.6038,
+      "step": 17170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888277973514544e-05,
+      "loss": 2.5793,
+      "step": 17175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888212778699483e-05,
+      "loss": 2.7473,
+      "step": 17180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888147565082967e-05,
+      "loss": 2.7056,
+      "step": 17185
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888082332665245e-05,
+      "loss": 2.6864,
+      "step": 17190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88801708144657e-05,
+      "loss": 2.5993,
+      "step": 17195
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887951811427189e-05,
+      "loss": 2.6319,
+      "step": 17200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887886522607357e-05,
+      "loss": 2.5281,
+      "step": 17205
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887821214987323e-05,
+      "loss": 2.4901,
+      "step": 17210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88775588856734e-05,
+      "loss": 2.5394,
+      "step": 17215
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887690543347656e-05,
+      "loss": 2.6394,
+      "step": 17220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887625179328526e-05,
+      "loss": 2.6881,
+      "step": 17225
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887559796510199e-05,
+      "loss": 2.7272,
+      "step": 17230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887494394892926e-05,
+      "loss": 2.46,
+      "step": 17235
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887428974476962e-05,
+      "loss": 2.6133,
+      "step": 17240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887363535262555e-05,
+      "loss": 2.6595,
+      "step": 17245
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887298077249958e-05,
+      "loss": 2.6526,
+      "step": 17250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887232600439424e-05,
+      "loss": 2.6383,
+      "step": 17255
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887167104831203e-05,
+      "loss": 2.6456,
+      "step": 17260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887101590425547e-05,
+      "loss": 2.6898,
+      "step": 17265
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88703605722271e-05,
+      "loss": 2.4909,
+      "step": 17270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886970505222943e-05,
+      "loss": 2.7625,
+      "step": 17275
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886904934426495e-05,
+      "loss": 2.5151,
+      "step": 17280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886839344833624e-05,
+      "loss": 2.602,
+      "step": 17285
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886773736444577e-05,
+      "loss": 2.545,
+      "step": 17290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88670810925961e-05,
+      "loss": 2.5736,
+      "step": 17295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886642463278974e-05,
+      "loss": 2.5142,
+      "step": 17300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886576798502922e-05,
+      "loss": 2.7131,
+      "step": 17305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886511114931704e-05,
+      "loss": 2.7615,
+      "step": 17310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886445412565578e-05,
+      "loss": 2.6276,
+      "step": 17315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886379691404791e-05,
+      "loss": 2.678,
+      "step": 17320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886313951449599e-05,
+      "loss": 2.6866,
+      "step": 17325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886248192700254e-05,
+      "loss": 2.5034,
+      "step": 17330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88618241515701e-05,
+      "loss": 2.5627,
+      "step": 17335
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886116618820118e-05,
+      "loss": 2.5451,
+      "step": 17340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886050803689832e-05,
+      "loss": 2.6219,
+      "step": 17345
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885984969766406e-05,
+      "loss": 2.5282,
+      "step": 17350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885919117050091e-05,
+      "loss": 2.5856,
+      "step": 17355
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885853245541143e-05,
+      "loss": 2.7023,
+      "step": 17360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885787355239813e-05,
+      "loss": 2.7155,
+      "step": 17365
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885721446146357e-05,
+      "loss": 2.6624,
+      "step": 17370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885655518261026e-05,
+      "loss": 2.669,
+      "step": 17375
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885589571584075e-05,
+      "loss": 2.5463,
+      "step": 17380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885523606115757e-05,
+      "loss": 2.5605,
+      "step": 17385
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885457621856326e-05,
+      "loss": 2.6376,
+      "step": 17390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885391618806037e-05,
+      "loss": 2.7188,
+      "step": 17395
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885325596965141e-05,
+      "loss": 2.492,
+      "step": 17400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885259556333894e-05,
+      "loss": 2.6278,
+      "step": 17405
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88519349691255e-05,
+      "loss": 2.6091,
+      "step": 17410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885127418701364e-05,
+      "loss": 2.6874,
+      "step": 17415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885061321700586e-05,
+      "loss": 2.723,
+      "step": 17420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884995205910475e-05,
+      "loss": 2.5807,
+      "step": 17425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884929071331283e-05,
+      "loss": 2.6611,
+      "step": 17430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884862917963264e-05,
+      "loss": 2.7547,
+      "step": 17435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884796745806675e-05,
+      "loss": 2.7068,
+      "step": 17440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884730554861767e-05,
+      "loss": 2.6242,
+      "step": 17445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884664345128796e-05,
+      "loss": 2.6276,
+      "step": 17450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884598116608018e-05,
+      "loss": 2.4709,
+      "step": 17455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884531869299687e-05,
+      "loss": 2.625,
+      "step": 17460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884465603204056e-05,
+      "loss": 2.6915,
+      "step": 17465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884399318321383e-05,
+      "loss": 2.6378,
+      "step": 17470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884333014651922e-05,
+      "loss": 2.6045,
+      "step": 17475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884266692195925e-05,
+      "loss": 2.664,
+      "step": 17480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884200350953649e-05,
+      "loss": 2.6951,
+      "step": 17485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884133990925352e-05,
+      "loss": 2.5934,
+      "step": 17490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884067612111286e-05,
+      "loss": 2.7152,
+      "step": 17495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884001214511706e-05,
+      "loss": 2.5921,
+      "step": 17500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88393479812687e-05,
+      "loss": 2.658,
+      "step": 17505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883868362957032e-05,
+      "loss": 2.5922,
+      "step": 17510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883801909002447e-05,
+      "loss": 2.6239,
+      "step": 17515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883735436263371e-05,
+      "loss": 2.6626,
+      "step": 17520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88366894474006e-05,
+      "loss": 2.604,
+      "step": 17525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883602434432769e-05,
+      "loss": 2.5396,
+      "step": 17530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883535905341756e-05,
+      "loss": 2.5553,
+      "step": 17535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883469357467275e-05,
+      "loss": 2.495,
+      "step": 17540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883402790809579e-05,
+      "loss": 2.5712,
+      "step": 17545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883336205368931e-05,
+      "loss": 2.5279,
+      "step": 17550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883269601145582e-05,
+      "loss": 2.7673,
+      "step": 17555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88320297813979e-05,
+      "loss": 2.4306,
+      "step": 17560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88313633635181e-05,
+      "loss": 2.548,
+      "step": 17565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.8830696757819e-05,
+      "loss": 2.7265,
+      "step": 17570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883002996430313e-05,
+      "loss": 2.5653,
+      "step": 17575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882936298297312e-05,
+      "loss": 2.563,
+      "step": 17580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882869581383148e-05,
+      "loss": 2.9151,
+      "step": 17585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882802845688077e-05,
+      "loss": 2.5479,
+      "step": 17590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882736091212359e-05,
+      "loss": 2.6502,
+      "step": 17595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88266931795625e-05,
+      "loss": 2.4859,
+      "step": 17600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882602525920006e-05,
+      "loss": 2.5981,
+      "step": 17605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882535715103885e-05,
+      "loss": 2.4999,
+      "step": 17610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882468885508144e-05,
+      "loss": 2.686,
+      "step": 17615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882402037133038e-05,
+      "loss": 2.5546,
+      "step": 17620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882335169978824e-05,
+      "loss": 2.5046,
+      "step": 17625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882268284045764e-05,
+      "loss": 2.4681,
+      "step": 17630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88220137933411e-05,
+      "loss": 2.594,
+      "step": 17635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882134455844121e-05,
+      "loss": 2.4694,
+      "step": 17640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882067513576055e-05,
+      "loss": 2.5626,
+      "step": 17645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88200055253017e-05,
+      "loss": 2.6168,
+      "step": 17650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88193357270672e-05,
+      "loss": 2.7517,
+      "step": 17655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88186657410597e-05,
+      "loss": 2.6291,
+      "step": 17660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881799556728169e-05,
+      "loss": 2.6777,
+      "step": 17665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88173252057358e-05,
+      "loss": 2.6149,
+      "step": 17670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88166546564246e-05,
+      "loss": 2.7446,
+      "step": 17675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881598391935065e-05,
+      "loss": 2.5216,
+      "step": 17680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881531299451656e-05,
+      "loss": 2.4914,
+      "step": 17685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881464188192488e-05,
+      "loss": 2.7263,
+      "step": 17690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881397058157822e-05,
+      "loss": 2.4853,
+      "step": 17695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881329909347914e-05,
+      "loss": 2.6562,
+      "step": 17700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881262741763025e-05,
+      "loss": 2.6333,
+      "step": 17705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881195555403409e-05,
+      "loss": 2.6237,
+      "step": 17710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881128350269329e-05,
+      "loss": 2.5485,
+      "step": 17715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881061126361039e-05,
+      "loss": 2.4697,
+      "step": 17720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880993883678802e-05,
+      "loss": 2.5781,
+      "step": 17725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880926622222875e-05,
+      "loss": 2.6443,
+      "step": 17730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880859341993513e-05,
+      "loss": 2.6492,
+      "step": 17735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880792042990981e-05,
+      "loss": 2.623,
+      "step": 17740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880724725215536e-05,
+      "loss": 2.666,
+      "step": 17745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880657388667434e-05,
+      "loss": 2.6722,
+      "step": 17750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880590033346935e-05,
+      "loss": 2.5991,
+      "step": 17755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.8805226592543e-05,
+      "loss": 2.6595,
+      "step": 17760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880455266389788e-05,
+      "loss": 2.5079,
+      "step": 17765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880387854753654e-05,
+      "loss": 2.6606,
+      "step": 17770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880320424346163e-05,
+      "loss": 2.5627,
+      "step": 17775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880252975167572e-05,
+      "loss": 2.6329,
+      "step": 17780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88018550721814e-05,
+      "loss": 2.7299,
+      "step": 17785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880118020498129e-05,
+      "loss": 2.64,
+      "step": 17790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880050515007793e-05,
+      "loss": 2.6498,
+      "step": 17795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879982990747396e-05,
+      "loss": 2.5636,
+      "step": 17800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879915447717197e-05,
+      "loss": 2.6729,
+      "step": 17805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879847885917456e-05,
+      "loss": 2.6698,
+      "step": 17810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879780305348432e-05,
+      "loss": 2.5715,
+      "step": 17815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879712706010386e-05,
+      "loss": 2.4979,
+      "step": 17820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879645087903577e-05,
+      "loss": 2.7042,
+      "step": 17825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879577451028264e-05,
+      "loss": 2.6657,
+      "step": 17830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879509795384712e-05,
+      "loss": 2.4646,
+      "step": 17835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879442120973174e-05,
+      "loss": 2.7037,
+      "step": 17840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879374427793917e-05,
+      "loss": 2.5471,
+      "step": 17845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879306715847197e-05,
+      "loss": 2.6664,
+      "step": 17850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879238985133276e-05,
+      "loss": 2.71,
+      "step": 17855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879171235652415e-05,
+      "loss": 2.7278,
+      "step": 17860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879103467404873e-05,
+      "loss": 2.6551,
+      "step": 17865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879035680390913e-05,
+      "loss": 2.4821,
+      "step": 17870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878967874610794e-05,
+      "loss": 2.606,
+      "step": 17875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878900050064776e-05,
+      "loss": 2.6113,
+      "step": 17880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878832206753123e-05,
+      "loss": 2.5519,
+      "step": 17885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878764344676092e-05,
+      "loss": 2.5905,
+      "step": 17890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878696463833948e-05,
+      "loss": 2.6951,
+      "step": 17895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878628564226948e-05,
+      "loss": 2.5865,
+      "step": 17900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878560645855357e-05,
+      "loss": 2.5682,
+      "step": 17905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878492708719433e-05,
+      "loss": 2.77,
+      "step": 17910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878424752819442e-05,
+      "loss": 2.447,
+      "step": 17915
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87835677815564e-05,
+      "loss": 2.5871,
+      "step": 17920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87828878472829e-05,
+      "loss": 2.5071,
+      "step": 17925
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878220772537654e-05,
+      "loss": 2.4882,
+      "step": 17930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878152741583995e-05,
+      "loss": 2.6301,
+      "step": 17935
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878084691867573e-05,
+      "loss": 2.5454,
+      "step": 17940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87801662338865e-05,
+      "loss": 2.604,
+      "step": 17945
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877948536147489e-05,
+      "loss": 2.5737,
+      "step": 17950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87788043014435e-05,
+      "loss": 2.694,
+      "step": 17955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877812305379496e-05,
+      "loss": 2.7007,
+      "step": 17960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877744161853188e-05,
+      "loss": 2.6631,
+      "step": 17965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877675999565689e-05,
+      "loss": 2.6066,
+      "step": 17970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877607818517263e-05,
+      "loss": 2.6889,
+      "step": 17975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877539618708169e-05,
+      "loss": 2.6728,
+      "step": 17980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877471400138671e-05,
+      "loss": 2.6742,
+      "step": 17985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877403162809031e-05,
+      "loss": 2.6947,
+      "step": 17990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877334906719511e-05,
+      "loss": 2.6352,
+      "step": 17995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877266631870374e-05,
+      "loss": 2.6219,
+      "step": 18000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877198338261883e-05,
+      "loss": 2.62,
+      "step": 18005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877130025894301e-05,
+      "loss": 2.709,
+      "step": 18010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87706169476789e-05,
+      "loss": 2.6506,
+      "step": 18015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876993344882913e-05,
+      "loss": 2.7966,
+      "step": 18020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876924976239631e-05,
+      "loss": 2.5334,
+      "step": 18025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876856588838311e-05,
+      "loss": 2.6007,
+      "step": 18030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876788182679213e-05,
+      "loss": 2.5591,
+      "step": 18035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876719757762601e-05,
+      "loss": 2.5713,
+      "step": 18040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876651314088739e-05,
+      "loss": 2.6262,
+      "step": 18045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876582851657888e-05,
+      "loss": 2.5589,
+      "step": 18050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876514370470312e-05,
+      "loss": 2.6198,
+      "step": 18055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876445870526276e-05,
+      "loss": 2.5923,
+      "step": 18060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876377351826043e-05,
+      "loss": 2.6283,
+      "step": 18065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876308814369875e-05,
+      "loss": 2.5078,
+      "step": 18070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876240258158038e-05,
+      "loss": 2.546,
+      "step": 18075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876171683190793e-05,
+      "loss": 2.7003,
+      "step": 18080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876103089468407e-05,
+      "loss": 2.5488,
+      "step": 18085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87603447699114e-05,
+      "loss": 2.5717,
+      "step": 18090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87596584575926e-05,
+      "loss": 2.5591,
+      "step": 18095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875897195773028e-05,
+      "loss": 2.6128,
+      "step": 18100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875828527032707e-05,
+      "loss": 2.6021,
+      "step": 18105
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875759839538565e-05,
+      "loss": 2.6766,
+      "step": 18110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875691133290865e-05,
+      "loss": 2.5632,
+      "step": 18115
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875622408289868e-05,
+      "loss": 2.6119,
+      "step": 18120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875553664535841e-05,
+      "loss": 2.6695,
+      "step": 18125
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875484902029051e-05,
+      "loss": 2.5874,
+      "step": 18130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875416120769758e-05,
+      "loss": 2.5509,
+      "step": 18135
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875347320758228e-05,
+      "loss": 2.5776,
+      "step": 18140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875278501994726e-05,
+      "loss": 2.5701,
+      "step": 18145
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875209664479517e-05,
+      "loss": 2.5594,
+      "step": 18150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875140808212865e-05,
+      "loss": 2.6454,
+      "step": 18155
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875071933195034e-05,
+      "loss": 2.638,
+      "step": 18160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875003039426293e-05,
+      "loss": 2.569,
+      "step": 18165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874934126906902e-05,
+      "loss": 2.482,
+      "step": 18170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874865195637129e-05,
+      "loss": 2.6168,
+      "step": 18175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874796245617237e-05,
+      "loss": 2.7527,
+      "step": 18180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874727276847493e-05,
+      "loss": 2.5574,
+      "step": 18185
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874658289328162e-05,
+      "loss": 2.5801,
+      "step": 18190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87458928305951e-05,
+      "loss": 2.6751,
+      "step": 18195
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874520258041801e-05,
+      "loss": 2.6507,
+      "step": 18200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.8744512142753e-05,
+      "loss": 2.6064,
+      "step": 18205
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874382151760274e-05,
+      "loss": 2.6701,
+      "step": 18210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874313070496989e-05,
+      "loss": 2.679,
+      "step": 18215
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874243970485709e-05,
+      "loss": 2.6439,
+      "step": 18220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.8741748517267e-05,
+      "loss": 2.635,
+      "step": 18225
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874105714220232e-05,
+      "loss": 2.604,
+      "step": 18230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874036557966564e-05,
+      "loss": 2.6494,
+      "step": 18235
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873967382965966e-05,
+      "loss": 2.6333,
+      "step": 18240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873898189218704e-05,
+      "loss": 2.6359,
+      "step": 18245
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873828976725044e-05,
+      "loss": 2.6849,
+      "step": 18250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873759745485252e-05,
+      "loss": 2.5298,
+      "step": 18255
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873690495499592e-05,
+      "loss": 2.5536,
+      "step": 18260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873621226768335e-05,
+      "loss": 2.5529,
+      "step": 18265
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873551939291744e-05,
+      "loss": 2.5503,
+      "step": 18270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873482633070085e-05,
+      "loss": 2.6311,
+      "step": 18275
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87341330810363e-05,
+      "loss": 2.5191,
+      "step": 18280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873343964392639e-05,
+      "loss": 2.4813,
+      "step": 18285
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873274601937382e-05,
+      "loss": 2.6812,
+      "step": 18290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873205220738123e-05,
+      "loss": 2.5195,
+      "step": 18295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873135820795132e-05,
+      "loss": 2.6155,
+      "step": 18300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873066402108676e-05,
+      "loss": 2.5389,
+      "step": 18305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872996964679021e-05,
+      "loss": 2.6524,
+      "step": 18310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872927508506433e-05,
+      "loss": 2.6664,
+      "step": 18315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872858033591181e-05,
+      "loss": 2.5563,
+      "step": 18320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87278853993353e-05,
+      "loss": 2.6043,
+      "step": 18325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872719027533751e-05,
+      "loss": 2.4786,
+      "step": 18330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872649496392108e-05,
+      "loss": 2.636,
+      "step": 18335
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872579946508868e-05,
+      "loss": 2.4224,
+      "step": 18340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872510377884302e-05,
+      "loss": 2.4993,
+      "step": 18345
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872440790518676e-05,
+      "loss": 2.6849,
+      "step": 18350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872371184412254e-05,
+      "loss": 2.6405,
+      "step": 18355
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87230155956531e-05,
+      "loss": 2.6378,
+      "step": 18360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872231915978108e-05,
+      "loss": 2.7695,
+      "step": 18365
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872162253650916e-05,
+      "loss": 2.5353,
+      "step": 18370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872092572584004e-05,
+      "loss": 2.6202,
+      "step": 18375
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872022872777636e-05,
+      "loss": 2.6509,
+      "step": 18380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871953154232085e-05,
+      "loss": 2.7318,
+      "step": 18385
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871883416947616e-05,
+      "loss": 2.6087,
+      "step": 18390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871813660924498e-05,
+      "loss": 2.4203,
+      "step": 18395
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871743886162999e-05,
+      "loss": 2.6903,
+      "step": 18400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871674092663387e-05,
+      "loss": 2.5404,
+      "step": 18405
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871604280425933e-05,
+      "loss": 2.6073,
+      "step": 18410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871534449450903e-05,
+      "loss": 2.6983,
+      "step": 18415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871464599738567e-05,
+      "loss": 2.6085,
+      "step": 18420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871394731289192e-05,
+      "loss": 2.5864,
+      "step": 18425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871324844103047e-05,
+      "loss": 2.7143,
+      "step": 18430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871254938180403e-05,
+      "loss": 2.5775,
+      "step": 18435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871185013521527e-05,
+      "loss": 2.6271,
+      "step": 18440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871115070126687e-05,
+      "loss": 2.6141,
+      "step": 18445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871045107996155e-05,
+      "loss": 2.6211,
+      "step": 18450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870975127130197e-05,
+      "loss": 2.6122,
+      "step": 18455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870905127529084e-05,
+      "loss": 2.6831,
+      "step": 18460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870835109193086e-05,
+      "loss": 2.5945,
+      "step": 18465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87076507212247e-05,
+      "loss": 2.583,
+      "step": 18470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870695016317505e-05,
+      "loss": 2.5682,
+      "step": 18475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870624941778464e-05,
+      "loss": 2.6902,
+      "step": 18480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870554848505614e-05,
+      "loss": 2.6211,
+      "step": 18485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870484736499226e-05,
+      "loss": 2.5958,
+      "step": 18490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870414605759569e-05,
+      "loss": 2.6346,
+      "step": 18495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870344456286911e-05,
+      "loss": 2.4936,
+      "step": 18500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870274288081523e-05,
+      "loss": 2.7925,
+      "step": 18505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870204101143676e-05,
+      "loss": 2.5249,
+      "step": 18510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870133895473639e-05,
+      "loss": 2.5276,
+      "step": 18515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870063671071684e-05,
+      "loss": 2.6532,
+      "step": 18520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869993427938076e-05,
+      "loss": 2.6492,
+      "step": 18525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869923166073089e-05,
+      "loss": 2.5772,
+      "step": 18530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869852885476995e-05,
+      "loss": 2.5902,
+      "step": 18535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86978258615006e-05,
+      "loss": 2.4589,
+      "step": 18540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869712268092559e-05,
+      "loss": 2.6481,
+      "step": 18545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869641931304759e-05,
+      "loss": 2.5612,
+      "step": 18550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869571575786929e-05,
+      "loss": 2.571,
+      "step": 18555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869501201539344e-05,
+      "loss": 2.5242,
+      "step": 18560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869430808562273e-05,
+      "loss": 2.5381,
+      "step": 18565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869360396855984e-05,
+      "loss": 2.5018,
+      "step": 18570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869289966420754e-05,
+      "loss": 2.6089,
+      "step": 18575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869219517256846e-05,
+      "loss": 2.6605,
+      "step": 18580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86914904936454e-05,
+      "loss": 2.5459,
+      "step": 18585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869078562744098e-05,
+      "loss": 2.671,
+      "step": 18590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869008057395799e-05,
+      "loss": 2.6264,
+      "step": 18595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868937533319908e-05,
+      "loss": 2.5475,
+      "step": 18600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868866990516698e-05,
+      "loss": 2.6233,
+      "step": 18605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868796428986443e-05,
+      "loss": 2.586,
+      "step": 18610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868725848729412e-05,
+      "loss": 2.4503,
+      "step": 18615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868655249745878e-05,
+      "loss": 2.6538,
+      "step": 18620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868584632036109e-05,
+      "loss": 2.5452,
+      "step": 18625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86851399560038e-05,
+      "loss": 2.6174,
+      "step": 18630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868443340438963e-05,
+      "loss": 2.6341,
+      "step": 18635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868372666552127e-05,
+      "loss": 2.4883,
+      "step": 18640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868301973940146e-05,
+      "loss": 2.4817,
+      "step": 18645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868231262603292e-05,
+      "loss": 2.6435,
+      "step": 18650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868160532541834e-05,
+      "loss": 2.7304,
+      "step": 18655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86808978375605e-05,
+      "loss": 2.4513,
+      "step": 18660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868019016246205e-05,
+      "loss": 2.6529,
+      "step": 18665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867948230012576e-05,
+      "loss": 2.5442,
+      "step": 18670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867877425055434e-05,
+      "loss": 2.6243,
+      "step": 18675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867806601375052e-05,
+      "loss": 2.6352,
+      "step": 18680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.8677357589717e-05,
+      "loss": 2.5118,
+      "step": 18685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867664897845653e-05,
+      "loss": 2.6567,
+      "step": 18690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867594017997183e-05,
+      "loss": 2.5307,
+      "step": 18695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867523119426561e-05,
+      "loss": 2.4841,
+      "step": 18700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867452202134062e-05,
+      "loss": 2.5794,
+      "step": 18705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867381266119957e-05,
+      "loss": 2.4785,
+      "step": 18710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867310311384522e-05,
+      "loss": 2.6694,
+      "step": 18715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867239337928025e-05,
+      "loss": 2.7025,
+      "step": 18720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867168345750742e-05,
+      "loss": 2.5705,
+      "step": 18725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867097334852945e-05,
+      "loss": 2.6765,
+      "step": 18730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867026305234907e-05,
+      "loss": 2.5087,
+      "step": 18735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866955256896904e-05,
+      "loss": 2.587,
+      "step": 18740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866884189839208e-05,
+      "loss": 2.5931,
+      "step": 18745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86681310406209e-05,
+      "loss": 2.68,
+      "step": 18750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866741999565823e-05,
+      "loss": 2.5588,
+      "step": 18755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866670876350685e-05,
+      "loss": 2.6781,
+      "step": 18760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866599734416945e-05,
+      "loss": 2.5683,
+      "step": 18765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866528573764881e-05,
+      "loss": 2.6631,
+      "step": 18770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866457394394763e-05,
+      "loss": 2.5853,
+      "step": 18775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866386196306864e-05,
+      "loss": 2.5686,
+      "step": 18780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866314979501464e-05,
+      "loss": 2.619,
+      "step": 18785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86624374397883e-05,
+      "loss": 2.5316,
+      "step": 18790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866172489739241e-05,
+      "loss": 2.5795,
+      "step": 18795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866101216782966e-05,
+      "loss": 2.6115,
+      "step": 18800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866029925110284e-05,
+      "loss": 2.7223,
+      "step": 18805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865958614721466e-05,
+      "loss": 2.6189,
+      "step": 18810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865887285616788e-05,
+      "loss": 2.7774,
+      "step": 18815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865815937796526e-05,
+      "loss": 2.6211,
+      "step": 18820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865744571260948e-05,
+      "loss": 2.5325,
+      "step": 18825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865673186010336e-05,
+      "loss": 2.6925,
+      "step": 18830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865601782044962e-05,
+      "loss": 2.6218,
+      "step": 18835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865530359365099e-05,
+      "loss": 2.5439,
+      "step": 18840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865458917971021e-05,
+      "loss": 2.7039,
+      "step": 18845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865387457863006e-05,
+      "loss": 2.4883,
+      "step": 18850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865315979041327e-05,
+      "loss": 2.6312,
+      "step": 18855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86524448150626e-05,
+      "loss": 2.5156,
+      "step": 18860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865172965258079e-05,
+      "loss": 2.5531,
+      "step": 18865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865101430297058e-05,
+      "loss": 2.6794,
+      "step": 18870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.865029876623475e-05,
+      "loss": 2.6765,
+      "step": 18875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864958304237603e-05,
+      "loss": 2.6141,
+      "step": 18880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864886713139719e-05,
+      "loss": 2.6848,
+      "step": 18885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864815103330095e-05,
+      "loss": 2.4786,
+      "step": 18890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86474347480901e-05,
+      "loss": 2.6815,
+      "step": 18895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864671827576739e-05,
+      "loss": 2.7102,
+      "step": 18900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864600161633557e-05,
+      "loss": 2.6149,
+      "step": 18905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864528476979738e-05,
+      "loss": 2.5423,
+      "step": 18910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86445677361556e-05,
+      "loss": 2.5393,
+      "step": 18915
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864385051541298e-05,
+      "loss": 2.5631,
+      "step": 18920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86431331075723e-05,
+      "loss": 2.6005,
+      "step": 18925
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864241551263627e-05,
+      "loss": 2.6052,
+      "step": 18930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86416977306077e-05,
+      "loss": 2.5753,
+      "step": 18935
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864097976148931e-05,
+      "loss": 2.6202,
+      "step": 18940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.864026160528388e-05,
+      "loss": 2.5835,
+      "step": 18945
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.863954326199419e-05,
+      "loss": 2.6764,
+      "step": 18950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.863882473162297e-05,
+      "loss": 2.5815,
+      "step": 18955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.8638106014173e-05,
+      "loss": 2.6123,
+      "step": 18960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.863738710964706e-05,
+      "loss": 2.6666,
+      "step": 18965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.863666801804787e-05,
+      "loss": 2.5167,
+      "step": 18970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.863594873937824e-05,
+      "loss": 2.5411,
+      "step": 18975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.863522927364093e-05,
+      "loss": 2.4664,
+      "step": 18980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86345096208387e-05,
+      "loss": 2.6019,
+      "step": 18985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.86337897809743e-05,
+      "loss": 2.5343,
+      "step": 18990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.863306975405053e-05,
+      "loss": 2.6576,
+      "step": 18995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.863234954007015e-05,
+      "loss": 2.6432,
+      "step": 19000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.863162913903591e-05,
+      "loss": 2.6162,
+      "step": 19005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.86309085509506e-05,
+      "loss": 2.5508,
+      "step": 19010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.863018777581699e-05,
+      "loss": 2.689,
+      "step": 19015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862946681363786e-05,
+      "loss": 2.5391,
+      "step": 19020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862874566441595e-05,
+      "loss": 2.6275,
+      "step": 19025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862802432815405e-05,
+      "loss": 2.6248,
+      "step": 19030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862730280485496e-05,
+      "loss": 2.6741,
+      "step": 19035
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862658109452145e-05,
+      "loss": 2.715,
+      "step": 19040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862585919715625e-05,
+      "loss": 2.6675,
+      "step": 19045
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862513711276218e-05,
+      "loss": 2.5767,
+      "step": 19050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862441484134202e-05,
+      "loss": 2.6511,
+      "step": 19055
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862369238289853e-05,
+      "loss": 2.6686,
+      "step": 19060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862296973743448e-05,
+      "loss": 2.605,
+      "step": 19065
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862224690495267e-05,
+      "loss": 2.5432,
+      "step": 19070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862152388545587e-05,
+      "loss": 2.7939,
+      "step": 19075
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862080067894685e-05,
+      "loss": 2.6423,
+      "step": 19080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862007728542844e-05,
+      "loss": 2.7534,
+      "step": 19085
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861935370490336e-05,
+      "loss": 2.6397,
+      "step": 19090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861862993737442e-05,
+      "loss": 2.5763,
+      "step": 19095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861790598284442e-05,
+      "loss": 2.4929,
+      "step": 19100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861718184131611e-05,
+      "loss": 2.5207,
+      "step": 19105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861645751279232e-05,
+      "loss": 2.4337,
+      "step": 19110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.86157329972758e-05,
+      "loss": 2.5216,
+      "step": 19115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861500829476933e-05,
+      "loss": 2.5066,
+      "step": 19120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861428340527572e-05,
+      "loss": 2.6274,
+      "step": 19125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861355832879777e-05,
+      "loss": 2.4737,
+      "step": 19130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861283306533824e-05,
+      "loss": 2.8666,
+      "step": 19135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861210761489991e-05,
+      "loss": 2.5186,
+      "step": 19140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861138197748563e-05,
+      "loss": 2.665,
+      "step": 19145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861065615309813e-05,
+      "loss": 2.5734,
+      "step": 19150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860993014174023e-05,
+      "loss": 2.7064,
+      "step": 19155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.86092039434147e-05,
+      "loss": 2.6486,
+      "step": 19160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860847755812437e-05,
+      "loss": 2.6444,
+      "step": 19165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8607750985872e-05,
+      "loss": 2.6078,
+      "step": 19170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860702422666042e-05,
+      "loss": 2.7073,
+      "step": 19175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860629728049238e-05,
+      "loss": 2.6208,
+      "step": 19180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.86055701473707e-05,
+      "loss": 2.5937,
+      "step": 19185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.86048428272982e-05,
+      "loss": 2.7042,
+      "step": 19190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860411532027763e-05,
+      "loss": 2.6545,
+      "step": 19195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860338762631182e-05,
+      "loss": 2.68,
+      "step": 19200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860265974540356e-05,
+      "loss": 2.6276,
+      "step": 19205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860193167755565e-05,
+      "loss": 2.7275,
+      "step": 19210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860120342277088e-05,
+      "loss": 2.6865,
+      "step": 19215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860047498105207e-05,
+      "loss": 2.536,
+      "step": 19220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859974635240202e-05,
+      "loss": 2.6388,
+      "step": 19225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859901753682352e-05,
+      "loss": 2.7152,
+      "step": 19230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859828853431938e-05,
+      "loss": 2.5812,
+      "step": 19235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859755934489242e-05,
+      "loss": 2.577,
+      "step": 19240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85968299685454e-05,
+      "loss": 2.4797,
+      "step": 19245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859610040528115e-05,
+      "loss": 2.4889,
+      "step": 19250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85953706551025e-05,
+      "loss": 2.4875,
+      "step": 19255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859464071801223e-05,
+      "loss": 2.6782,
+      "step": 19260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859391059401317e-05,
+      "loss": 2.752,
+      "step": 19265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859318028310809e-05,
+      "loss": 2.5127,
+      "step": 19270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859244978529982e-05,
+      "loss": 2.5128,
+      "step": 19275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859171910059117e-05,
+      "loss": 2.5454,
+      "step": 19280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859098822898496e-05,
+      "loss": 2.6445,
+      "step": 19285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859025717048401e-05,
+      "loss": 2.5814,
+      "step": 19290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858952592509108e-05,
+      "loss": 2.554,
+      "step": 19295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858879449280904e-05,
+      "loss": 2.7609,
+      "step": 19300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858806287364066e-05,
+      "loss": 2.6188,
+      "step": 19305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85873310675888e-05,
+      "loss": 2.6684,
+      "step": 19310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858659907465621e-05,
+      "loss": 2.6447,
+      "step": 19315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858586689484579e-05,
+      "loss": 2.5518,
+      "step": 19320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858513452816028e-05,
+      "loss": 2.5519,
+      "step": 19325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858440197460253e-05,
+      "loss": 2.3692,
+      "step": 19330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858366923417538e-05,
+      "loss": 2.5561,
+      "step": 19335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85829363068816e-05,
+      "loss": 2.605,
+      "step": 19340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858220319272403e-05,
+      "loss": 2.5158,
+      "step": 19345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858146989170549e-05,
+      "loss": 2.5924,
+      "step": 19350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858073640382881e-05,
+      "loss": 2.6365,
+      "step": 19355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858000272909681e-05,
+      "loss": 2.599,
+      "step": 19360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85792688675123e-05,
+      "loss": 2.6246,
+      "step": 19365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85785348190781e-05,
+      "loss": 2.7448,
+      "step": 19370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857780058379707e-05,
+      "loss": 2.6672,
+      "step": 19375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857706616167197e-05,
+      "loss": 2.7679,
+      "step": 19380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857633155270569e-05,
+      "loss": 2.5788,
+      "step": 19385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8575596756901e-05,
+      "loss": 2.5725,
+      "step": 19390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857486177426077e-05,
+      "loss": 2.7278,
+      "step": 19395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857412660478779e-05,
+      "loss": 2.6422,
+      "step": 19400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857339124848492e-05,
+      "loss": 2.6063,
+      "step": 19405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8572655705355e-05,
+      "loss": 2.6831,
+      "step": 19410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857191997540079e-05,
+      "loss": 2.6876,
+      "step": 19415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857118405862518e-05,
+      "loss": 2.7013,
+      "step": 19420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857044795503099e-05,
+      "loss": 2.6592,
+      "step": 19425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856971166462106e-05,
+      "loss": 2.5771,
+      "step": 19430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856897518739818e-05,
+      "loss": 2.6479,
+      "step": 19435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856823852336523e-05,
+      "loss": 2.682,
+      "step": 19440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856750167252502e-05,
+      "loss": 2.7259,
+      "step": 19445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856676463488037e-05,
+      "loss": 2.5224,
+      "step": 19450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856602741043416e-05,
+      "loss": 2.6331,
+      "step": 19455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856528999918918e-05,
+      "loss": 2.5832,
+      "step": 19460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85645524011483e-05,
+      "loss": 2.6085,
+      "step": 19465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856381461631433e-05,
+      "loss": 2.4503,
+      "step": 19470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856307664469012e-05,
+      "loss": 2.6381,
+      "step": 19475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856233848627851e-05,
+      "loss": 2.5893,
+      "step": 19480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856160014108234e-05,
+      "loss": 2.6708,
+      "step": 19485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856086160910445e-05,
+      "loss": 2.586,
+      "step": 19490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856012289034767e-05,
+      "loss": 2.4925,
+      "step": 19495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855938398481485e-05,
+      "loss": 2.6556,
+      "step": 19500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855864489250883e-05,
+      "loss": 2.4846,
+      "step": 19505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855790561343248e-05,
+      "loss": 2.7162,
+      "step": 19510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855716614758859e-05,
+      "loss": 2.6199,
+      "step": 19515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855642649498002e-05,
+      "loss": 2.6951,
+      "step": 19520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855568665560966e-05,
+      "loss": 2.5668,
+      "step": 19525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855494662948031e-05,
+      "loss": 2.6359,
+      "step": 19530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855420641659481e-05,
+      "loss": 2.6476,
+      "step": 19535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855346601695605e-05,
+      "loss": 2.6179,
+      "step": 19540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855272543056684e-05,
+      "loss": 2.5828,
+      "step": 19545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855198465743005e-05,
+      "loss": 2.5266,
+      "step": 19550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855124369754851e-05,
+      "loss": 2.6758,
+      "step": 19555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855050255092508e-05,
+      "loss": 2.6105,
+      "step": 19560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854976121756262e-05,
+      "loss": 2.5438,
+      "step": 19565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854901969746398e-05,
+      "loss": 2.675,
+      "step": 19570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8548277990632e-05,
+      "loss": 2.611,
+      "step": 19575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854753609706953e-05,
+      "loss": 2.5571,
+      "step": 19580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854679401677944e-05,
+      "loss": 2.6766,
+      "step": 19585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854605174976456e-05,
+      "loss": 2.6439,
+      "step": 19590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85453092960278e-05,
+      "loss": 2.6302,
+      "step": 19595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854456665557192e-05,
+      "loss": 2.6124,
+      "step": 19600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854382382839988e-05,
+      "loss": 2.5292,
+      "step": 19605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854308081451447e-05,
+      "loss": 2.5577,
+      "step": 19610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854233761391856e-05,
+      "loss": 2.6006,
+      "step": 19615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854159422661503e-05,
+      "loss": 2.6567,
+      "step": 19620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854085065260672e-05,
+      "loss": 2.6603,
+      "step": 19625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85401068918965e-05,
+      "loss": 2.5942,
+      "step": 19630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853936294448722e-05,
+      "loss": 2.6024,
+      "step": 19635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853861881038176e-05,
+      "loss": 2.4964,
+      "step": 19640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853787448958296e-05,
+      "loss": 2.4549,
+      "step": 19645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853712998209369e-05,
+      "loss": 2.6178,
+      "step": 19650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853638528791683e-05,
+      "loss": 2.673,
+      "step": 19655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85356404070552e-05,
+      "loss": 2.603,
+      "step": 19660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853489533951172e-05,
+      "loss": 2.5919,
+      "step": 19665
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853415008528924e-05,
+      "loss": 2.6189,
+      "step": 19670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85334046443906e-05,
+      "loss": 2.696,
+      "step": 19675
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85326590168187e-05,
+      "loss": 2.5468,
+      "step": 19680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853191320257638e-05,
+      "loss": 2.667,
+      "step": 19685
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853116720166653e-05,
+      "loss": 2.7404,
+      "step": 19690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8530421014092e-05,
+      "loss": 2.6027,
+      "step": 19695
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852967463985568e-05,
+      "loss": 2.5407,
+      "step": 19700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852892807896043e-05,
+      "loss": 2.5936,
+      "step": 19705
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852818133140912e-05,
+      "loss": 2.6689,
+      "step": 19710
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852743439720462e-05,
+      "loss": 2.5526,
+      "step": 19715
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852668727634983e-05,
+      "loss": 2.5295,
+      "step": 19720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852593996884759e-05,
+      "loss": 2.6329,
+      "step": 19725
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852519247470078e-05,
+      "loss": 2.5324,
+      "step": 19730
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852444479391228e-05,
+      "loss": 2.4354,
+      "step": 19735
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852369692648498e-05,
+      "loss": 2.7118,
+      "step": 19740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852294887242173e-05,
+      "loss": 2.6598,
+      "step": 19745
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852220063172544e-05,
+      "loss": 2.6811,
+      "step": 19750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852145220439894e-05,
+      "loss": 2.656,
+      "step": 19755
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852070359044517e-05,
+      "loss": 2.4529,
+      "step": 19760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851995478986696e-05,
+      "loss": 2.5689,
+      "step": 19765
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851920580266721e-05,
+      "loss": 2.5938,
+      "step": 19770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85184566288488e-05,
+      "loss": 2.6034,
+      "step": 19775
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85177072684146e-05,
+      "loss": 2.6844,
+      "step": 19780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85169577213675e-05,
+      "loss": 2.3932,
+      "step": 19785
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851620798771039e-05,
+      "loss": 2.6035,
+      "step": 19790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851545806744616e-05,
+      "loss": 2.6215,
+      "step": 19795
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851470796057767e-05,
+      "loss": 2.6609,
+      "step": 19800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851395766710783e-05,
+      "loss": 2.5573,
+      "step": 19805
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85132071870395e-05,
+      "loss": 2.612,
+      "step": 19810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851245652037559e-05,
+      "loss": 2.4786,
+      "step": 19815
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851170566711896e-05,
+      "loss": 2.5878,
+      "step": 19820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851095462727254e-05,
+      "loss": 2.7489,
+      "step": 19825
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851020340083917e-05,
+      "loss": 2.5356,
+      "step": 19830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850945198782178e-05,
+      "loss": 2.5695,
+      "step": 19835
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850870038822323e-05,
+      "loss": 2.5351,
+      "step": 19840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850794860204644e-05,
+      "loss": 2.5124,
+      "step": 19845
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850719662929428e-05,
+      "loss": 2.5515,
+      "step": 19850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850644446996967e-05,
+      "loss": 2.5469,
+      "step": 19855
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850569212407544e-05,
+      "loss": 2.6658,
+      "step": 19860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850493959161456e-05,
+      "loss": 2.5895,
+      "step": 19865
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850418687258986e-05,
+      "loss": 2.5744,
+      "step": 19870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850343396700429e-05,
+      "loss": 2.5943,
+      "step": 19875
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850268087486071e-05,
+      "loss": 2.6854,
+      "step": 19880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850192759616203e-05,
+      "loss": 2.6443,
+      "step": 19885
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850117413091113e-05,
+      "loss": 2.6369,
+      "step": 19890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850042047911094e-05,
+      "loss": 2.5838,
+      "step": 19895
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849966664076435e-05,
+      "loss": 2.5287,
+      "step": 19900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849891261587423e-05,
+      "loss": 2.6086,
+      "step": 19905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849815840444352e-05,
+      "loss": 2.6324,
+      "step": 19910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849740400647509e-05,
+      "loss": 2.5165,
+      "step": 19915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849664942197186e-05,
+      "loss": 2.6697,
+      "step": 19920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849589465093672e-05,
+      "loss": 2.6284,
+      "step": 19925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849513969337259e-05,
+      "loss": 2.6623,
+      "step": 19930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849438454928235e-05,
+      "loss": 2.5039,
+      "step": 19935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849362921866892e-05,
+      "loss": 2.6599,
+      "step": 19940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84928737015352e-05,
+      "loss": 2.5023,
+      "step": 19945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84921179978841e-05,
+      "loss": 2.6157,
+      "step": 19950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849136210771853e-05,
+      "loss": 2.6228,
+      "step": 19955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849060603104139e-05,
+      "loss": 2.7026,
+      "step": 19960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848984976785561e-05,
+      "loss": 2.4923,
+      "step": 19965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848909331816405e-05,
+      "loss": 2.6803,
+      "step": 19970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848833668196967e-05,
+      "loss": 2.663,
+      "step": 19975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848757985927534e-05,
+      "loss": 2.4786,
+      "step": 19980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848682285008399e-05,
+      "loss": 2.5967,
+      "step": 19985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848606565439855e-05,
+      "loss": 2.6066,
+      "step": 19990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848530827222191e-05,
+      "loss": 2.5164,
+      "step": 19995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848455070355698e-05,
+      "loss": 2.6442,
+      "step": 20000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848379294840668e-05,
+      "loss": 2.6761,
+      "step": 20005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848303500677394e-05,
+      "loss": 2.5236,
+      "step": 20010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848227687866165e-05,
+      "loss": 2.4905,
+      "step": 20015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848151856407273e-05,
+      "loss": 2.6154,
+      "step": 20020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848076006301012e-05,
+      "loss": 2.5747,
+      "step": 20025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848000137547671e-05,
+      "loss": 2.6679,
+      "step": 20030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847924250147543e-05,
+      "loss": 2.778,
+      "step": 20035
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84784834410092e-05,
+      "loss": 2.6523,
+      "step": 20040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847772419408095e-05,
+      "loss": 2.529,
+      "step": 20045
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847696476069357e-05,
+      "loss": 2.6112,
+      "step": 20050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847620514085001e-05,
+      "loss": 2.5891,
+      "step": 20055
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847544533455318e-05,
+      "loss": 2.7276,
+      "step": 20060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847468534180599e-05,
+      "loss": 2.6453,
+      "step": 20065
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847392516261137e-05,
+      "loss": 2.5951,
+      "step": 20070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847316479697228e-05,
+      "loss": 2.5495,
+      "step": 20075
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847240424489159e-05,
+      "loss": 2.5609,
+      "step": 20080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847164350637227e-05,
+      "loss": 2.502,
+      "step": 20085
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847088258141721e-05,
+      "loss": 2.4632,
+      "step": 20090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847012147002936e-05,
+      "loss": 2.57,
+      "step": 20095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846936017221164e-05,
+      "loss": 2.8086,
+      "step": 20100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846859868796698e-05,
+      "loss": 2.5946,
+      "step": 20105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84678370172983e-05,
+      "loss": 2.6219,
+      "step": 20110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846707516020855e-05,
+      "loss": 2.8232,
+      "step": 20115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846631311670062e-05,
+      "loss": 2.5891,
+      "step": 20120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846555088677749e-05,
+      "loss": 2.4553,
+      "step": 20125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846478847044207e-05,
+      "loss": 2.6259,
+      "step": 20130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846402586769729e-05,
+      "loss": 2.5938,
+      "step": 20135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846326307854609e-05,
+      "loss": 2.4333,
+      "step": 20140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846250010299139e-05,
+      "loss": 2.6368,
+      "step": 20145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846173694103614e-05,
+      "loss": 2.5268,
+      "step": 20150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846097359268327e-05,
+      "loss": 2.5327,
+      "step": 20155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846021005793572e-05,
+      "loss": 2.6985,
+      "step": 20160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845944633679641e-05,
+      "loss": 2.6134,
+      "step": 20165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845868242926829e-05,
+      "loss": 2.638,
+      "step": 20170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84579183353543e-05,
+      "loss": 2.6313,
+      "step": 20175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845715405505738e-05,
+      "loss": 2.6463,
+      "step": 20180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845638958838047e-05,
+      "loss": 2.6147,
+      "step": 20185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84556249353265e-05,
+      "loss": 2.7408,
+      "step": 20190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845486009589841e-05,
+      "loss": 2.5633,
+      "step": 20195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845409507009918e-05,
+      "loss": 2.56,
+      "step": 20200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84533298579317e-05,
+      "loss": 2.5838,
+      "step": 20205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845256445939894e-05,
+      "loss": 2.6604,
+      "step": 20210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845179887450383e-05,
+      "loss": 2.4462,
+      "step": 20215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845103310324933e-05,
+      "loss": 2.681,
+      "step": 20220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845026714563838e-05,
+      "loss": 2.5699,
+      "step": 20225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844950100167394e-05,
+      "loss": 2.6757,
+      "step": 20230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844873467135892e-05,
+      "loss": 2.6467,
+      "step": 20235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84479681546963e-05,
+      "loss": 2.6386,
+      "step": 20240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844720145168902e-05,
+      "loss": 2.6009,
+      "step": 20245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844643456234002e-05,
+      "loss": 2.6173,
+      "step": 20250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844566748665226e-05,
+      "loss": 2.6054,
+      "step": 20255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844490022462868e-05,
+      "loss": 2.6037,
+      "step": 20260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844413277627227e-05,
+      "loss": 2.6535,
+      "step": 20265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844336514158592e-05,
+      "loss": 2.5749,
+      "step": 20270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844259732057262e-05,
+      "loss": 2.4363,
+      "step": 20275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844182931323531e-05,
+      "loss": 2.6012,
+      "step": 20280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844106111957696e-05,
+      "loss": 2.7103,
+      "step": 20285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844029273960052e-05,
+      "loss": 2.5647,
+      "step": 20290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843952417330892e-05,
+      "loss": 2.5443,
+      "step": 20295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843875542070514e-05,
+      "loss": 2.6025,
+      "step": 20300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843798648179215e-05,
+      "loss": 2.446,
+      "step": 20305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843721735657288e-05,
+      "loss": 2.7254,
+      "step": 20310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843644804505028e-05,
+      "loss": 2.4915,
+      "step": 20315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843567854722735e-05,
+      "loss": 2.6593,
+      "step": 20320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843490886310702e-05,
+      "loss": 2.3846,
+      "step": 20325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843413899269227e-05,
+      "loss": 2.4817,
+      "step": 20330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843336893598603e-05,
+      "loss": 2.5845,
+      "step": 20335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84325986929913e-05,
+      "loss": 2.6258,
+      "step": 20340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843182826371101e-05,
+      "loss": 2.7075,
+      "step": 20345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843105764814814e-05,
+      "loss": 2.6776,
+      "step": 20350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843028684630565e-05,
+      "loss": 2.5659,
+      "step": 20355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842951585818652e-05,
+      "loss": 2.5731,
+      "step": 20360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842874468379369e-05,
+      "loss": 2.6235,
+      "step": 20365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842797332313011e-05,
+      "loss": 2.557,
+      "step": 20370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842720177619882e-05,
+      "loss": 2.5234,
+      "step": 20375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842643004300271e-05,
+      "loss": 2.5488,
+      "step": 20380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842565812354479e-05,
+      "loss": 2.6775,
+      "step": 20385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842488601782803e-05,
+      "loss": 2.6081,
+      "step": 20390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842411372585536e-05,
+      "loss": 2.6654,
+      "step": 20395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842334124762979e-05,
+      "loss": 2.5794,
+      "step": 20400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84225685831543e-05,
+      "loss": 2.64,
+      "step": 20405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842179573243184e-05,
+      "loss": 2.5522,
+      "step": 20410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842102269546536e-05,
+      "loss": 2.6227,
+      "step": 20415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842024947225787e-05,
+      "loss": 2.5934,
+      "step": 20420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841947606281235e-05,
+      "loss": 2.6346,
+      "step": 20425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841870246713172e-05,
+      "loss": 2.488,
+      "step": 20430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841792868521902e-05,
+      "loss": 2.674,
+      "step": 20435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841715471707719e-05,
+      "loss": 2.6454,
+      "step": 20440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841638056270922e-05,
+      "loss": 2.7102,
+      "step": 20445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841560622211808e-05,
+      "loss": 2.5481,
+      "step": 20450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841483169530674e-05,
+      "loss": 2.5255,
+      "step": 20455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841405698227821e-05,
+      "loss": 2.4866,
+      "step": 20460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841328208303545e-05,
+      "loss": 2.596,
+      "step": 20465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841250699758142e-05,
+      "loss": 2.5829,
+      "step": 20470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841173172591913e-05,
+      "loss": 2.5415,
+      "step": 20475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841095626805156e-05,
+      "loss": 2.5673,
+      "step": 20480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841018062398169e-05,
+      "loss": 2.5558,
+      "step": 20485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840940479371249e-05,
+      "loss": 2.6021,
+      "step": 20490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840862877724695e-05,
+      "loss": 2.6659,
+      "step": 20495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840785257458807e-05,
+      "loss": 2.6179,
+      "step": 20500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840707618573882e-05,
+      "loss": 2.596,
+      "step": 20505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84062996107022e-05,
+      "loss": 2.5686,
+      "step": 20510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840552284948116e-05,
+      "loss": 2.5041,
+      "step": 20515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840474590207874e-05,
+      "loss": 2.5609,
+      "step": 20520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840396876849789e-05,
+      "loss": 2.5036,
+      "step": 20525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840319144874163e-05,
+      "loss": 2.4862,
+      "step": 20530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840241394281291e-05,
+      "loss": 2.54,
+      "step": 20535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840163625071476e-05,
+      "loss": 2.6082,
+      "step": 20540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840085837245014e-05,
+      "loss": 2.5965,
+      "step": 20545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840008030802205e-05,
+      "loss": 2.6125,
+      "step": 20550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839930205743351e-05,
+      "loss": 2.6449,
+      "step": 20555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839852362068749e-05,
+      "loss": 2.6999,
+      "step": 20560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839774499778697e-05,
+      "loss": 2.5896,
+      "step": 20565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839696618873497e-05,
+      "loss": 2.5146,
+      "step": 20570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839618719353447e-05,
+      "loss": 2.4653,
+      "step": 20575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839540801218848e-05,
+      "loss": 2.6932,
+      "step": 20580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83946286447e-05,
+      "loss": 2.4979,
+      "step": 20585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8393849091072e-05,
+      "loss": 2.6639,
+      "step": 20590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83930693513075e-05,
+      "loss": 2.6004,
+      "step": 20595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839228942540948e-05,
+      "loss": 2.581,
+      "step": 20600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839150931338097e-05,
+      "loss": 2.6463,
+      "step": 20605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839072901522495e-05,
+      "loss": 2.6193,
+      "step": 20610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838994853094443e-05,
+      "loss": 2.5207,
+      "step": 20615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838916786054241e-05,
+      "loss": 2.6601,
+      "step": 20620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838838700402188e-05,
+      "loss": 2.7014,
+      "step": 20625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838760596138588e-05,
+      "loss": 2.5314,
+      "step": 20630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838682473263736e-05,
+      "loss": 2.7793,
+      "step": 20635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838604331777937e-05,
+      "loss": 2.6036,
+      "step": 20640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838526171681487e-05,
+      "loss": 2.6134,
+      "step": 20645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838447992974693e-05,
+      "loss": 2.5362,
+      "step": 20650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838369795657852e-05,
+      "loss": 2.743,
+      "step": 20655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838291579731263e-05,
+      "loss": 2.6594,
+      "step": 20660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838213345195228e-05,
+      "loss": 2.6023,
+      "step": 20665
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838135092050051e-05,
+      "loss": 2.6181,
+      "step": 20670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838056820296031e-05,
+      "loss": 2.5219,
+      "step": 20675
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837978529933468e-05,
+      "loss": 2.6276,
+      "step": 20680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837900220962663e-05,
+      "loss": 2.5177,
+      "step": 20685
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837821893383918e-05,
+      "loss": 2.5078,
+      "step": 20690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837743547197536e-05,
+      "loss": 2.6605,
+      "step": 20695
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837665182403815e-05,
+      "loss": 2.6092,
+      "step": 20700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837586799003059e-05,
+      "loss": 2.7422,
+      "step": 20705
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837508396995568e-05,
+      "loss": 2.5289,
+      "step": 20710
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837429976381645e-05,
+      "loss": 2.494,
+      "step": 20715
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83735153716159e-05,
+      "loss": 2.5383,
+      "step": 20720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837273079335706e-05,
+      "loss": 2.4893,
+      "step": 20725
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837194602904294e-05,
+      "loss": 2.6909,
+      "step": 20730
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837116107867657e-05,
+      "loss": 2.5321,
+      "step": 20735
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837037594226095e-05,
+      "loss": 2.4765,
+      "step": 20740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836959061979911e-05,
+      "loss": 2.4753,
+      "step": 20745
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836880511129407e-05,
+      "loss": 2.6529,
+      "step": 20750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836801941674885e-05,
+      "loss": 2.5788,
+      "step": 20755
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836723353616648e-05,
+      "loss": 2.5847,
+      "step": 20760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836644746954997e-05,
+      "loss": 2.5125,
+      "step": 20765
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836566121690236e-05,
+      "loss": 2.5267,
+      "step": 20770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836487477822664e-05,
+      "loss": 2.5117,
+      "step": 20775
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83640881535259e-05,
+      "loss": 2.5103,
+      "step": 20780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836330134280308e-05,
+      "loss": 2.8117,
+      "step": 20785
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836251434606127e-05,
+      "loss": 2.6921,
+      "step": 20790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836172716330349e-05,
+      "loss": 2.6836,
+      "step": 20795
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836093979453274e-05,
+      "loss": 2.7044,
+      "step": 20800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836015223975207e-05,
+      "loss": 2.6213,
+      "step": 20805
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83593644989645e-05,
+      "loss": 2.7352,
+      "step": 20810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.835857657217307e-05,
+      "loss": 2.5758,
+      "step": 20815
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83577884593808e-05,
+      "loss": 2.6045,
+      "step": 20820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83570001605907e-05,
+      "loss": 2.5304,
+      "step": 20825
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.835621167580586e-05,
+      "loss": 2.614,
+      "step": 20830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.835542300502927e-05,
+      "loss": 2.532,
+      "step": 20835
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.835463414826398e-05,
+      "loss": 2.7538,
+      "step": 20840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8353845105513e-05,
+      "loss": 2.5874,
+      "step": 20845
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83530558767794e-05,
+      "loss": 2.5645,
+      "step": 20850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83522664620662e-05,
+      "loss": 2.6067,
+      "step": 20855
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83514768613764e-05,
+      "loss": 2.5732,
+      "step": 20860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.835068707471311e-05,
+      "loss": 2.6638,
+      "step": 20865
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834989710207931e-05,
+      "loss": 2.5465,
+      "step": 20870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834910694347806e-05,
+      "loss": 2.7172,
+      "step": 20875
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83483165989124e-05,
+      "loss": 2.5372,
+      "step": 20880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834752606838536e-05,
+      "loss": 2.545,
+      "step": 20885
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834673535189999e-05,
+      "loss": 2.5023,
+      "step": 20890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834594444945932e-05,
+      "loss": 2.5279,
+      "step": 20895
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834515336106642e-05,
+      "loss": 2.5076,
+      "step": 20900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83443620867243e-05,
+      "loss": 2.6619,
+      "step": 20905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834357062643603e-05,
+      "loss": 2.5186,
+      "step": 20910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83427789802046e-05,
+      "loss": 2.6659,
+      "step": 20915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834198714803313e-05,
+      "loss": 2.7182,
+      "step": 20920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834119512992464e-05,
+      "loss": 2.4936,
+      "step": 20925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834040292588215e-05,
+      "loss": 2.6289,
+      "step": 20930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833961053590872e-05,
+      "loss": 2.6457,
+      "step": 20935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833881796000742e-05,
+      "loss": 2.7142,
+      "step": 20940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833802519818128e-05,
+      "loss": 2.5532,
+      "step": 20945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833723225043333e-05,
+      "loss": 2.5362,
+      "step": 20950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833643911676665e-05,
+      "loss": 2.5995,
+      "step": 20955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833564579718428e-05,
+      "loss": 2.7211,
+      "step": 20960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833485229168927e-05,
+      "loss": 2.5281,
+      "step": 20965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833405860028468e-05,
+      "loss": 2.7013,
+      "step": 20970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833326472297354e-05,
+      "loss": 2.5358,
+      "step": 20975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833247065975895e-05,
+      "loss": 2.6221,
+      "step": 20980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833167641064392e-05,
+      "loss": 2.579,
+      "step": 20985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833088197563151e-05,
+      "loss": 2.6051,
+      "step": 20990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83300873547248e-05,
+      "loss": 2.6685,
+      "step": 20995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832929254792681e-05,
+      "loss": 2.5878,
+      "step": 21000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832849755524064e-05,
+      "loss": 2.5357,
+      "step": 21005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832770237666933e-05,
+      "loss": 2.6801,
+      "step": 21010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832690701221592e-05,
+      "loss": 2.7019,
+      "step": 21015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832611146188349e-05,
+      "loss": 2.6998,
+      "step": 21020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832531572567507e-05,
+      "loss": 2.7007,
+      "step": 21025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832451980359378e-05,
+      "loss": 2.5985,
+      "step": 21030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832372369564262e-05,
+      "loss": 2.7509,
+      "step": 21035
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83229274018247e-05,
+      "loss": 2.5985,
+      "step": 21040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832213092214304e-05,
+      "loss": 2.6737,
+      "step": 21045
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832133425660073e-05,
+      "loss": 2.6518,
+      "step": 21050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832053740520083e-05,
+      "loss": 2.602,
+      "step": 21055
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83197403679464e-05,
+      "loss": 2.4755,
+      "step": 21060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831894314484051e-05,
+      "loss": 2.4191,
+      "step": 21065
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831814573588623e-05,
+      "loss": 2.662,
+      "step": 21070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831734814108661e-05,
+      "loss": 2.5583,
+      "step": 21075
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831655036044474e-05,
+      "loss": 2.5754,
+      "step": 21080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831575239396368e-05,
+      "loss": 2.4943,
+      "step": 21085
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83149542416465e-05,
+      "loss": 2.6789,
+      "step": 21090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831415590349626e-05,
+      "loss": 2.6294,
+      "step": 21095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831335737951604e-05,
+      "loss": 2.5195,
+      "step": 21100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83125586697089e-05,
+      "loss": 2.6572,
+      "step": 21105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831175977407794e-05,
+      "loss": 2.5531,
+      "step": 21110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83109606926262e-05,
+      "loss": 2.7711,
+      "step": 21115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831016142535676e-05,
+      "loss": 2.6338,
+      "step": 21120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830936197227272e-05,
+      "loss": 2.5423,
+      "step": 21125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830856233337712e-05,
+      "loss": 2.7452,
+      "step": 21130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830776250867308e-05,
+      "loss": 2.63,
+      "step": 21135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830696249816363e-05,
+      "loss": 2.5872,
+      "step": 21140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830616230185187e-05,
+      "loss": 2.604,
+      "step": 21145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830536191974087e-05,
+      "loss": 2.6137,
+      "step": 21150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830456135183372e-05,
+      "loss": 2.5364,
+      "step": 21155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830376059813349e-05,
+      "loss": 2.5347,
+      "step": 21160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830295965864325e-05,
+      "loss": 2.6251,
+      "step": 21165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83021585333661e-05,
+      "loss": 2.6352,
+      "step": 21170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830135722230511e-05,
+      "loss": 2.5865,
+      "step": 21175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830055572546337e-05,
+      "loss": 2.4833,
+      "step": 21180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829975404284395e-05,
+      "loss": 2.5813,
+      "step": 21185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829895217444995e-05,
+      "loss": 2.6372,
+      "step": 21190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829815012028445e-05,
+      "loss": 2.5951,
+      "step": 21195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829734788035053e-05,
+      "loss": 2.5419,
+      "step": 21200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829654545465126e-05,
+      "loss": 2.497,
+      "step": 21205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829574284318976e-05,
+      "loss": 2.4248,
+      "step": 21210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82949400459691e-05,
+      "loss": 2.6265,
+      "step": 21215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829413706299235e-05,
+      "loss": 2.459,
+      "step": 21220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829333389426262e-05,
+      "loss": 2.5431,
+      "step": 21225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829253053978302e-05,
+      "loss": 2.6837,
+      "step": 21230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82917269995566e-05,
+      "loss": 2.7206,
+      "step": 21235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829092327358647e-05,
+      "loss": 2.4866,
+      "step": 21240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829011936187571e-05,
+      "loss": 2.7083,
+      "step": 21245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.828931526442742e-05,
+      "loss": 2.6497,
+      "step": 21250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82885109812447e-05,
+      "loss": 2.6893,
+      "step": 21255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.828770651233063e-05,
+      "loss": 2.5434,
+      "step": 21260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82869018576883e-05,
+      "loss": 2.6195,
+      "step": 21265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.828609701732083e-05,
+      "loss": 2.6989,
+      "step": 21270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.828529199123128e-05,
+      "loss": 2.5465,
+      "step": 21275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82844867794228e-05,
+      "loss": 2.4818,
+      "step": 21280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.828368138189843e-05,
+      "loss": 2.6544,
+      "step": 21285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82828757986613e-05,
+      "loss": 2.6596,
+      "step": 21290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82820700297145e-05,
+      "loss": 2.6113,
+      "step": 21295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.828126407506112e-05,
+      "loss": 2.4628,
+      "step": 21300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.828045793470429e-05,
+      "loss": 2.5672,
+      "step": 21305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827965160864707e-05,
+      "loss": 2.5603,
+      "step": 21310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82788450968926e-05,
+      "loss": 2.6215,
+      "step": 21315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827803839944394e-05,
+      "loss": 2.6467,
+      "step": 21320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827723151630424e-05,
+      "loss": 2.4595,
+      "step": 21325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827642444747656e-05,
+      "loss": 2.6875,
+      "step": 21330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827561719296403e-05,
+      "loss": 2.5965,
+      "step": 21335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827480975276975e-05,
+      "loss": 2.4884,
+      "step": 21340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827400212689682e-05,
+      "loss": 2.5035,
+      "step": 21345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827319431534836e-05,
+      "loss": 2.6227,
+      "step": 21350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827238631812747e-05,
+      "loss": 2.6821,
+      "step": 21355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827157813523724e-05,
+      "loss": 2.6498,
+      "step": 21360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.827076976668081e-05,
+      "loss": 2.5716,
+      "step": 21365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826996121246127e-05,
+      "loss": 2.7393,
+      "step": 21370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826915247258172e-05,
+      "loss": 2.7082,
+      "step": 21375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82683435470453e-05,
+      "loss": 2.568,
+      "step": 21380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826753443585511e-05,
+      "loss": 2.5775,
+      "step": 21385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826672513901425e-05,
+      "loss": 2.5525,
+      "step": 21390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826591565652583e-05,
+      "loss": 2.5589,
+      "step": 21395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8265105988393e-05,
+      "loss": 2.6005,
+      "step": 21400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826429613461883e-05,
+      "loss": 2.5807,
+      "step": 21405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826348609520644e-05,
+      "loss": 2.5339,
+      "step": 21410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826267587015898e-05,
+      "loss": 2.4852,
+      "step": 21415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826186545947953e-05,
+      "loss": 2.5154,
+      "step": 21420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.826105486317123e-05,
+      "loss": 2.5649,
+      "step": 21425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82602440812372e-05,
+      "loss": 2.5388,
+      "step": 21430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825943311368054e-05,
+      "loss": 2.5296,
+      "step": 21435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825862196050437e-05,
+      "loss": 2.6668,
+      "step": 21440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825781062171182e-05,
+      "loss": 2.5271,
+      "step": 21445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825699909730602e-05,
+      "loss": 2.6047,
+      "step": 21450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825618738729005e-05,
+      "loss": 2.6482,
+      "step": 21455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825537549166708e-05,
+      "loss": 2.6121,
+      "step": 21460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825456341044022e-05,
+      "loss": 2.5144,
+      "step": 21465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825375114361258e-05,
+      "loss": 2.5302,
+      "step": 21470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82529386911873e-05,
+      "loss": 2.5959,
+      "step": 21475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825212605316749e-05,
+      "loss": 2.6076,
+      "step": 21480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825131322955628e-05,
+      "loss": 2.5461,
+      "step": 21485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82505002203568e-05,
+      "loss": 2.5673,
+      "step": 21490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.824968702557218e-05,
+      "loss": 2.5999,
+      "step": 21495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.824887364520555e-05,
+      "loss": 2.503,
+      "step": 21500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.824806007926002e-05,
+      "loss": 2.5238,
+      "step": 21505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.824724632773873e-05,
+      "loss": 2.6132,
+      "step": 21510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.824643239064484e-05,
+      "loss": 2.6523,
+      "step": 21515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.824561826798143e-05,
+      "loss": 2.4973,
+      "step": 21520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.824480395975164e-05,
+      "loss": 2.5977,
+      "step": 21525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.824398946595863e-05,
+      "loss": 2.4952,
+      "step": 21530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.824317478660552e-05,
+      "loss": 2.6367,
+      "step": 21535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.824235992169543e-05,
+      "loss": 2.5819,
+      "step": 21540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.824154487123152e-05,
+      "loss": 2.5817,
+      "step": 21545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82407296352169e-05,
+      "loss": 2.7324,
+      "step": 21550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823991421365472e-05,
+      "loss": 2.6665,
+      "step": 21555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823909860654812e-05,
+      "loss": 2.3963,
+      "step": 21560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823828281390022e-05,
+      "loss": 2.6201,
+      "step": 21565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823746683571417e-05,
+      "loss": 2.6306,
+      "step": 21570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823665067199312e-05,
+      "loss": 2.5922,
+      "step": 21575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823583432274018e-05,
+      "loss": 2.6484,
+      "step": 21580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823501778795851e-05,
+      "loss": 2.5677,
+      "step": 21585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823420106765124e-05,
+      "loss": 2.6509,
+      "step": 21590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823338416182151e-05,
+      "loss": 2.5484,
+      "step": 21595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823256707047249e-05,
+      "loss": 2.5332,
+      "step": 21600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82317497936073e-05,
+      "loss": 2.5717,
+      "step": 21605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823093233122907e-05,
+      "loss": 2.6623,
+      "step": 21610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823011468334096e-05,
+      "loss": 2.5532,
+      "step": 21615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822929684994612e-05,
+      "loss": 2.6641,
+      "step": 21620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82284788310477e-05,
+      "loss": 2.5036,
+      "step": 21625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822766062664884e-05,
+      "loss": 2.5275,
+      "step": 21630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822684223675266e-05,
+      "loss": 2.5559,
+      "step": 21635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822602366136234e-05,
+      "loss": 2.6536,
+      "step": 21640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822520490048101e-05,
+      "loss": 2.613,
+      "step": 21645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822438595411185e-05,
+      "loss": 2.5121,
+      "step": 21650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822356682225798e-05,
+      "loss": 2.503,
+      "step": 21655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822274750492257e-05,
+      "loss": 2.5165,
+      "step": 21660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822192800210874e-05,
+      "loss": 2.6397,
+      "step": 21665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822110831381967e-05,
+      "loss": 2.5527,
+      "step": 21670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822028844005852e-05,
+      "loss": 2.6291,
+      "step": 21675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821946838082841e-05,
+      "loss": 2.7439,
+      "step": 21680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821864813613253e-05,
+      "loss": 2.6757,
+      "step": 21685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.8217827705974e-05,
+      "loss": 2.4879,
+      "step": 21690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821700709035601e-05,
+      "loss": 2.5364,
+      "step": 21695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821618628928169e-05,
+      "loss": 2.648,
+      "step": 21700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821536530275421e-05,
+      "loss": 2.6075,
+      "step": 21705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821454413077674e-05,
+      "loss": 2.4379,
+      "step": 21710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821372277335239e-05,
+      "loss": 2.5425,
+      "step": 21715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821290123048438e-05,
+      "loss": 2.5254,
+      "step": 21720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821207950217584e-05,
+      "loss": 2.6738,
+      "step": 21725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82112575884299e-05,
+      "loss": 2.6746,
+      "step": 21730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82104354892498e-05,
+      "loss": 2.4672,
+      "step": 21735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820961320463864e-05,
+      "loss": 2.6776,
+      "step": 21740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820879073459959e-05,
+      "loss": 2.5475,
+      "step": 21745
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820796807913582e-05,
+      "loss": 2.6066,
+      "step": 21750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820714523825051e-05,
+      "loss": 2.5533,
+      "step": 21755
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820632221194681e-05,
+      "loss": 2.5472,
+      "step": 21760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820549900022788e-05,
+      "loss": 2.5623,
+      "step": 21765
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82046756030969e-05,
+      "loss": 2.566,
+      "step": 21770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820385202055702e-05,
+      "loss": 2.6632,
+      "step": 21775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820302825261143e-05,
+      "loss": 2.5684,
+      "step": 21780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820220429926328e-05,
+      "loss": 2.6717,
+      "step": 21785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820138016051574e-05,
+      "loss": 2.7298,
+      "step": 21790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.8200555836372e-05,
+      "loss": 2.4795,
+      "step": 21795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81997313268352e-05,
+      "loss": 2.5044,
+      "step": 21800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819890663190854e-05,
+      "loss": 2.4917,
+      "step": 21805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819808175159518e-05,
+      "loss": 2.6649,
+      "step": 21810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81972566858983e-05,
+      "loss": 2.4557,
+      "step": 21815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819643143482105e-05,
+      "loss": 2.6194,
+      "step": 21820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819560599836663e-05,
+      "loss": 2.5206,
+      "step": 21825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81947803765382e-05,
+      "loss": 2.6152,
+      "step": 21830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819395456933892e-05,
+      "loss": 2.5715,
+      "step": 21835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819312857677202e-05,
+      "loss": 2.7125,
+      "step": 21840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819230239884063e-05,
+      "loss": 2.4725,
+      "step": 21845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819147603554794e-05,
+      "loss": 2.6009,
+      "step": 21850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819064948689713e-05,
+      "loss": 2.5299,
+      "step": 21855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818982275289137e-05,
+      "loss": 2.4891,
+      "step": 21860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818899583353387e-05,
+      "loss": 2.6359,
+      "step": 21865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818816872882777e-05,
+      "loss": 2.5408,
+      "step": 21870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818734143877629e-05,
+      "loss": 2.5903,
+      "step": 21875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818651396338258e-05,
+      "loss": 2.4832,
+      "step": 21880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818568630264984e-05,
+      "loss": 2.4593,
+      "step": 21885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818485845658123e-05,
+      "loss": 2.5722,
+      "step": 21890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818403042517997e-05,
+      "loss": 2.5854,
+      "step": 21895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818320220844924e-05,
+      "loss": 2.6508,
+      "step": 21900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818237380639219e-05,
+      "loss": 2.5768,
+      "step": 21905
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818154521901204e-05,
+      "loss": 2.5357,
+      "step": 21910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818071644631196e-05,
+      "loss": 2.5776,
+      "step": 21915
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817988748829516e-05,
+      "loss": 2.5444,
+      "step": 21920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817905834496479e-05,
+      "loss": 2.6553,
+      "step": 21925
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817822901632407e-05,
+      "loss": 2.4913,
+      "step": 21930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817739950237619e-05,
+      "loss": 2.6749,
+      "step": 21935
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817656980312432e-05,
+      "loss": 2.4458,
+      "step": 21940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817573991857167e-05,
+      "loss": 2.5444,
+      "step": 21945
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817490984872143e-05,
+      "loss": 2.5144,
+      "step": 21950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817407959357678e-05,
+      "loss": 2.501,
+      "step": 21955
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817324915314092e-05,
+      "loss": 2.6443,
+      "step": 21960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817241852741705e-05,
+      "loss": 2.5698,
+      "step": 21965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817158771640837e-05,
+      "loss": 2.737,
+      "step": 21970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817075672011806e-05,
+      "loss": 2.5543,
+      "step": 21975
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816992553854931e-05,
+      "loss": 2.5492,
+      "step": 21980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816909417170535e-05,
+      "loss": 2.684,
+      "step": 21985
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816826261958934e-05,
+      "loss": 2.5458,
+      "step": 21990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81674308822045e-05,
+      "loss": 2.5369,
+      "step": 21995
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816659895955403e-05,
+      "loss": 2.4857,
+      "step": 22000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816576685164111e-05,
+      "loss": 2.5995,
+      "step": 22005
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816493455846896e-05,
+      "loss": 2.5969,
+      "step": 22010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816410208004078e-05,
+      "loss": 2.5218,
+      "step": 22015
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816326941635976e-05,
+      "loss": 2.6308,
+      "step": 22020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816243656742911e-05,
+      "loss": 2.5076,
+      "step": 22025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816160353325206e-05,
+      "loss": 2.4458,
+      "step": 22030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816077031383176e-05,
+      "loss": 2.6927,
+      "step": 22035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815993690917144e-05,
+      "loss": 2.607,
+      "step": 22040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815910331927434e-05,
+      "loss": 2.5663,
+      "step": 22045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81582695441436e-05,
+      "loss": 2.5316,
+      "step": 22050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815743558378248e-05,
+      "loss": 2.772,
+      "step": 22055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815660143819416e-05,
+      "loss": 2.5145,
+      "step": 22060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815576710738186e-05,
+      "loss": 2.6679,
+      "step": 22065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815493259134879e-05,
+      "loss": 2.6874,
+      "step": 22070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815409789009815e-05,
+      "loss": 2.6086,
+      "step": 22075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815326300363316e-05,
+      "loss": 2.5565,
+      "step": 22080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815242793195703e-05,
+      "loss": 2.5162,
+      "step": 22085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815159267507296e-05,
+      "loss": 2.7101,
+      "step": 22090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815075723298419e-05,
+      "loss": 2.6181,
+      "step": 22095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814992160569391e-05,
+      "loss": 2.6242,
+      "step": 22100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814908579320534e-05,
+      "loss": 2.5455,
+      "step": 22105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814824979552168e-05,
+      "loss": 2.5523,
+      "step": 22110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814741361264617e-05,
+      "loss": 2.4874,
+      "step": 22115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814657724458203e-05,
+      "loss": 2.5381,
+      "step": 22120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814574069133245e-05,
+      "loss": 2.6559,
+      "step": 22125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814490395290065e-05,
+      "loss": 2.5226,
+      "step": 22130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814406702928989e-05,
+      "loss": 2.6992,
+      "step": 22135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814322992050333e-05,
+      "loss": 2.5289,
+      "step": 22140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814239262654423e-05,
+      "loss": 2.5443,
+      "step": 22145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814155514741578e-05,
+      "loss": 2.4716,
+      "step": 22150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814071748312125e-05,
+      "loss": 2.7111,
+      "step": 22155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813987963366382e-05,
+      "loss": 2.41,
+      "step": 22160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813904159904672e-05,
+      "loss": 2.5656,
+      "step": 22165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813820337927318e-05,
+      "loss": 2.5858,
+      "step": 22170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813736497434641e-05,
+      "loss": 2.3222,
+      "step": 22175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813652638426967e-05,
+      "loss": 2.6393,
+      "step": 22180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813568760904614e-05,
+      "loss": 2.591,
+      "step": 22185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813484864867907e-05,
+      "loss": 2.7803,
+      "step": 22190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813400950317169e-05,
+      "loss": 2.5055,
+      "step": 22195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813317017252724e-05,
+      "loss": 2.4321,
+      "step": 22200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813233065674889e-05,
+      "loss": 2.6018,
+      "step": 22205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813149095583995e-05,
+      "loss": 2.7022,
+      "step": 22210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813065106980359e-05,
+      "loss": 2.5995,
+      "step": 22215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812981099864307e-05,
+      "loss": 2.6194,
+      "step": 22220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81289707423616e-05,
+      "loss": 2.5854,
+      "step": 22225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812813030096242e-05,
+      "loss": 2.533,
+      "step": 22230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812728967444878e-05,
+      "loss": 2.4006,
+      "step": 22235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812644886282388e-05,
+      "loss": 2.6834,
+      "step": 22240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812560786609099e-05,
+      "loss": 2.5961,
+      "step": 22245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812476668425333e-05,
+      "loss": 2.4595,
+      "step": 22250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812392531731411e-05,
+      "loss": 2.7524,
+      "step": 22255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812308376527661e-05,
+      "loss": 2.4289,
+      "step": 22260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812224202814403e-05,
+      "loss": 2.5046,
+      "step": 22265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812140010591963e-05,
+      "loss": 2.5303,
+      "step": 22270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812055799860665e-05,
+      "loss": 2.4451,
+      "step": 22275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811971570620832e-05,
+      "loss": 2.5146,
+      "step": 22280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811887322872786e-05,
+      "loss": 2.4872,
+      "step": 22285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811803056616855e-05,
+      "loss": 2.5267,
+      "step": 22290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811718771853361e-05,
+      "loss": 2.5849,
+      "step": 22295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811634468582629e-05,
+      "loss": 2.5943,
+      "step": 22300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81155014680498e-05,
+      "loss": 2.6846,
+      "step": 22305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811465806520743e-05,
+      "loss": 2.6003,
+      "step": 22310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811381447730242e-05,
+      "loss": 2.6607,
+      "step": 22315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811297070433797e-05,
+      "loss": 2.5027,
+      "step": 22320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811212674631735e-05,
+      "loss": 2.645,
+      "step": 22325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811128260324382e-05,
+      "loss": 2.6901,
+      "step": 22330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811043827512063e-05,
+      "loss": 2.5572,
+      "step": 22335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.8109593761951e-05,
+      "loss": 2.52,
+      "step": 22340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810874906373819e-05,
+      "loss": 2.5026,
+      "step": 22345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810790418048545e-05,
+      "loss": 2.5461,
+      "step": 22350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810705911219604e-05,
+      "loss": 2.5599,
+      "step": 22355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810621385887321e-05,
+      "loss": 2.5369,
+      "step": 22360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810536842052019e-05,
+      "loss": 2.5747,
+      "step": 22365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810452279714024e-05,
+      "loss": 2.637,
+      "step": 22370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810367698873663e-05,
+      "loss": 2.7106,
+      "step": 22375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810283099531259e-05,
+      "loss": 2.6656,
+      "step": 22380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810198481687138e-05,
+      "loss": 2.6148,
+      "step": 22385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810113845341627e-05,
+      "loss": 2.724,
+      "step": 22390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810029190495049e-05,
+      "loss": 2.5816,
+      "step": 22395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809944517147732e-05,
+      "loss": 2.5591,
+      "step": 22400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.8098598253e-05,
+      "loss": 2.6102,
+      "step": 22405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80977511495218e-05,
+      "loss": 2.5192,
+      "step": 22410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809690386104596e-05,
+      "loss": 2.3876,
+      "step": 22415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809605638757578e-05,
+      "loss": 2.5504,
+      "step": 22420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809520872911447e-05,
+      "loss": 2.6185,
+      "step": 22425
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80943608856653e-05,
+      "loss": 2.5134,
+      "step": 22430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809351285723156e-05,
+      "loss": 2.582,
+      "step": 22435
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809266464381647e-05,
+      "loss": 2.6177,
+      "step": 22440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809181624542335e-05,
+      "loss": 2.6741,
+      "step": 22445
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809096766205542e-05,
+      "loss": 2.6722,
+      "step": 22450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809011889371595e-05,
+      "loss": 2.5299,
+      "step": 22455
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80892699404082e-05,
+      "loss": 2.6466,
+      "step": 22460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808842080213545e-05,
+      "loss": 2.7483,
+      "step": 22465
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808757147890097e-05,
+      "loss": 2.6111,
+      "step": 22470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.8086721970708e-05,
+      "loss": 2.5426,
+      "step": 22475
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808587227755983e-05,
+      "loss": 2.6374,
+      "step": 22480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808502239945971e-05,
+      "loss": 2.5509,
+      "step": 22485
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808417233641093e-05,
+      "loss": 2.5805,
+      "step": 22490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808332208841676e-05,
+      "loss": 2.6564,
+      "step": 22495
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808247165548046e-05,
+      "loss": 2.6962,
+      "step": 22500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808162103760529e-05,
+      "loss": 2.6058,
+      "step": 22505
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808077023479453e-05,
+      "loss": 2.5314,
+      "step": 22510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807991924705147e-05,
+      "loss": 2.4335,
+      "step": 22515
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807906807437937e-05,
+      "loss": 2.625,
+      "step": 22520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807821671678149e-05,
+      "loss": 2.6224,
+      "step": 22525
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807736517426111e-05,
+      "loss": 2.6442,
+      "step": 22530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807651344682153e-05,
+      "loss": 2.5431,
+      "step": 22535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807566153446601e-05,
+      "loss": 2.4634,
+      "step": 22540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807480943719782e-05,
+      "loss": 2.5802,
+      "step": 22545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807395715502024e-05,
+      "loss": 2.6859,
+      "step": 22550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807310468793655e-05,
+      "loss": 2.6169,
+      "step": 22555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807225203595003e-05,
+      "loss": 2.5515,
+      "step": 22560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807139919906396e-05,
+      "loss": 2.5711,
+      "step": 22565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807054617728163e-05,
+      "loss": 2.5212,
+      "step": 22570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80696929706063e-05,
+      "loss": 2.5886,
+      "step": 22575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806883957904126e-05,
+      "loss": 2.5935,
+      "step": 22580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806798600258979e-05,
+      "loss": 2.6463,
+      "step": 22585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806713224125519e-05,
+      "loss": 2.5525,
+      "step": 22590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806627829504073e-05,
+      "loss": 2.5571,
+      "step": 22595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806542416394968e-05,
+      "loss": 2.5782,
+      "step": 22600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806456984798536e-05,
+      "loss": 2.4827,
+      "step": 22605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806371534715103e-05,
+      "loss": 2.5555,
+      "step": 22610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806286066144998e-05,
+      "loss": 2.66,
+      "step": 22615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806200579088551e-05,
+      "loss": 2.5225,
+      "step": 22620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806115073546089e-05,
+      "loss": 2.4869,
+      "step": 22625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806029549517944e-05,
+      "loss": 2.5472,
+      "step": 22630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80594400700444e-05,
+      "loss": 2.6295,
+      "step": 22635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805858446005909e-05,
+      "loss": 2.6926,
+      "step": 22640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80577286652268e-05,
+      "loss": 2.5602,
+      "step": 22645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805687268555084e-05,
+      "loss": 2.3744,
+      "step": 22650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805601652103447e-05,
+      "loss": 2.4893,
+      "step": 22655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805516017168098e-05,
+      "loss": 2.6916,
+      "step": 22660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80543036374937e-05,
+      "loss": 2.6125,
+      "step": 22665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805344691847589e-05,
+      "loss": 2.565,
+      "step": 22670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805259001463088e-05,
+      "loss": 2.6773,
+      "step": 22675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805173292596192e-05,
+      "loss": 2.4745,
+      "step": 22680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805087565247234e-05,
+      "loss": 2.5055,
+      "step": 22685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805001819416544e-05,
+      "loss": 2.5728,
+      "step": 22690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80491605510445e-05,
+      "loss": 2.5927,
+      "step": 22695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80483027231128e-05,
+      "loss": 2.6162,
+      "step": 22700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80474447103737e-05,
+      "loss": 2.5298,
+      "step": 22705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804658651283045e-05,
+      "loss": 2.4731,
+      "step": 22710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804572813048637e-05,
+      "loss": 2.6015,
+      "step": 22715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804486956334475e-05,
+      "loss": 2.5909,
+      "step": 22720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804401081140891e-05,
+      "loss": 2.5954,
+      "step": 22725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804315187468214e-05,
+      "loss": 2.5425,
+      "step": 22730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804229275316774e-05,
+      "loss": 2.5105,
+      "step": 22735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804143344686903e-05,
+      "loss": 2.5445,
+      "step": 22740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804057395578931e-05,
+      "loss": 2.5335,
+      "step": 22745
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803971427993189e-05,
+      "loss": 2.5793,
+      "step": 22750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803885441930004e-05,
+      "loss": 2.7839,
+      "step": 22755
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803799437389712e-05,
+      "loss": 2.6569,
+      "step": 22760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80371341437264e-05,
+      "loss": 2.3839,
+      "step": 22765
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80362737287912e-05,
+      "loss": 2.4543,
+      "step": 22770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803541312909484e-05,
+      "loss": 2.6401,
+      "step": 22775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803455234464062e-05,
+      "loss": 2.6012,
+      "step": 22780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803369137543185e-05,
+      "loss": 2.4493,
+      "step": 22785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803283022147185e-05,
+      "loss": 2.5407,
+      "step": 22790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803196888276392e-05,
+      "loss": 2.493,
+      "step": 22795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803110735931139e-05,
+      "loss": 2.5706,
+      "step": 22800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803024565111755e-05,
+      "loss": 2.571,
+      "step": 22805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802938375818574e-05,
+      "loss": 2.5409,
+      "step": 22810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802852168051926e-05,
+      "loss": 2.6449,
+      "step": 22815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802765941812141e-05,
+      "loss": 2.6679,
+      "step": 22820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802679697099555e-05,
+      "loss": 2.6145,
+      "step": 22825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802593433914498e-05,
+      "loss": 2.6716,
+      "step": 22830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802507152257299e-05,
+      "loss": 2.4522,
+      "step": 22835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802420852128292e-05,
+      "loss": 2.5268,
+      "step": 22840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802334533527808e-05,
+      "loss": 2.6644,
+      "step": 22845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802248196456181e-05,
+      "loss": 2.5841,
+      "step": 22850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802161840913742e-05,
+      "loss": 2.6347,
+      "step": 22855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802075466900825e-05,
+      "loss": 2.6004,
+      "step": 22860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801989074417757e-05,
+      "loss": 2.6817,
+      "step": 22865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801902663464875e-05,
+      "loss": 2.5117,
+      "step": 22870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801816234042509e-05,
+      "loss": 2.5202,
+      "step": 22875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801729786150993e-05,
+      "loss": 2.4855,
+      "step": 22880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801643319790659e-05,
+      "loss": 2.6446,
+      "step": 22885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801556834961837e-05,
+      "loss": 2.7414,
+      "step": 22890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801470331664865e-05,
+      "loss": 2.6477,
+      "step": 22895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801383809900073e-05,
+      "loss": 2.627,
+      "step": 22900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801297269667793e-05,
+      "loss": 2.5664,
+      "step": 22905
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801210710968359e-05,
+      "loss": 2.5733,
+      "step": 22910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801124133802101e-05,
+      "loss": 2.6117,
+      "step": 22915
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801037538169355e-05,
+      "loss": 2.6845,
+      "step": 22920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.800950924070455e-05,
+      "loss": 2.5885,
+      "step": 22925
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.800864291505732e-05,
+      "loss": 2.6474,
+      "step": 22930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80077764047552e-05,
+      "loss": 2.57,
+      "step": 22935
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80069097098015e-05,
+      "loss": 2.5397,
+      "step": 22940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80060428301996e-05,
+      "loss": 2.5661,
+      "step": 22945
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.800517576595281e-05,
+      "loss": 2.5354,
+      "step": 22950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.800430851706444e-05,
+      "loss": 2.5462,
+      "step": 22955
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.800344108353788e-05,
+      "loss": 2.5884,
+      "step": 22960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80025734653764e-05,
+      "loss": 2.6079,
+      "step": 22965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.800170566258339e-05,
+      "loss": 2.6697,
+      "step": 22970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.800083767516217e-05,
+      "loss": 2.5382,
+      "step": 22975
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799996950311609e-05,
+      "loss": 2.5515,
+      "step": 22980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799910114644846e-05,
+      "loss": 2.6413,
+      "step": 22985
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799823260516265e-05,
+      "loss": 2.6515,
+      "step": 22990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.7997363879262e-05,
+      "loss": 2.6534,
+      "step": 22995
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799649496874981e-05,
+      "loss": 2.5209,
+      "step": 23000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799562587362947e-05,
+      "loss": 2.4816,
+      "step": 23005
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799475659390432e-05,
+      "loss": 2.5482,
+      "step": 23010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799388712957767e-05,
+      "loss": 2.5294,
+      "step": 23015
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799301748065287e-05,
+      "loss": 2.5683,
+      "step": 23020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79921476471333e-05,
+      "loss": 2.542,
+      "step": 23025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799127762902227e-05,
+      "loss": 2.6411,
+      "step": 23030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799040742632315e-05,
+      "loss": 2.5009,
+      "step": 23035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798953703903927e-05,
+      "loss": 2.5603,
+      "step": 23040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.7988666467174e-05,
+      "loss": 2.6501,
+      "step": 23045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798779571073065e-05,
+      "loss": 2.6234,
+      "step": 23050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798692476971261e-05,
+      "loss": 2.5907,
+      "step": 23055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79860536441232e-05,
+      "loss": 2.4768,
+      "step": 23060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798518233396579e-05,
+      "loss": 2.5885,
+      "step": 23065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798431083924371e-05,
+      "loss": 2.6091,
+      "step": 23070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798343915996035e-05,
+      "loss": 2.5482,
+      "step": 23075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798256729611903e-05,
+      "loss": 2.6381,
+      "step": 23080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798169524772309e-05,
+      "loss": 2.6557,
+      "step": 23085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798082301477594e-05,
+      "loss": 2.5708,
+      "step": 23090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797995059728088e-05,
+      "loss": 2.6927,
+      "step": 23095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797907799524129e-05,
+      "loss": 2.7383,
+      "step": 23100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797820520866052e-05,
+      "loss": 2.5997,
+      "step": 23105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797733223754194e-05,
+      "loss": 2.6008,
+      "step": 23110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797645908188889e-05,
+      "loss": 2.5747,
+      "step": 23115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797558574170474e-05,
+      "loss": 2.6096,
+      "step": 23120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797471221699284e-05,
+      "loss": 2.5801,
+      "step": 23125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797383850775656e-05,
+      "loss": 2.5876,
+      "step": 23130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797296461399926e-05,
+      "loss": 2.5014,
+      "step": 23135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797209053572429e-05,
+      "loss": 2.4884,
+      "step": 23140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797121627293501e-05,
+      "loss": 2.5655,
+      "step": 23145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797034182563481e-05,
+      "loss": 2.448,
+      "step": 23150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796946719382702e-05,
+      "loss": 2.5808,
+      "step": 23155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796859237751504e-05,
+      "loss": 2.6406,
+      "step": 23160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79677173767022e-05,
+      "loss": 2.6488,
+      "step": 23165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796684219139188e-05,
+      "loss": 2.4263,
+      "step": 23170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796596682158744e-05,
+      "loss": 2.6132,
+      "step": 23175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796509126729225e-05,
+      "loss": 2.5891,
+      "step": 23180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79642155285097e-05,
+      "loss": 2.7334,
+      "step": 23185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796333960524311e-05,
+      "loss": 2.6325,
+      "step": 23190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79624634974959e-05,
+      "loss": 2.6779,
+      "step": 23195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796158720527139e-05,
+      "loss": 2.5894,
+      "step": 23200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796071072857302e-05,
+      "loss": 2.5142,
+      "step": 23205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79598340674041e-05,
+      "loss": 2.6415,
+      "step": 23210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.7958957221768e-05,
+      "loss": 2.6661,
+      "step": 23215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795808019166812e-05,
+      "loss": 2.5864,
+      "step": 23220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795720297710784e-05,
+      "loss": 2.6865,
+      "step": 23225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795632557809052e-05,
+      "loss": 2.5461,
+      "step": 23230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795544799461953e-05,
+      "loss": 2.492,
+      "step": 23235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795457022669825e-05,
+      "loss": 2.5519,
+      "step": 23240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795369227433007e-05,
+      "loss": 2.4814,
+      "step": 23245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795281413751833e-05,
+      "loss": 2.5692,
+      "step": 23250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795193581626645e-05,
+      "loss": 2.5695,
+      "step": 23255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795105731057779e-05,
+      "loss": 2.6763,
+      "step": 23260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795017862045572e-05,
+      "loss": 2.47,
+      "step": 23265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794929974590363e-05,
+      "loss": 2.6111,
+      "step": 23270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794842068692489e-05,
+      "loss": 2.489,
+      "step": 23275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79475414435229e-05,
+      "loss": 2.5747,
+      "step": 23280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794666201570103e-05,
+      "loss": 2.6472,
+      "step": 23285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794578240346265e-05,
+      "loss": 2.6009,
+      "step": 23290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794490260681119e-05,
+      "loss": 2.6878,
+      "step": 23295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794402262574997e-05,
+      "loss": 2.6999,
+      "step": 23300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794314246028242e-05,
+      "loss": 2.8197,
+      "step": 23305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794226211041191e-05,
+      "loss": 2.538,
+      "step": 23310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794138157614183e-05,
+      "loss": 2.6953,
+      "step": 23315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794050085747557e-05,
+      "loss": 2.5588,
+      "step": 23320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79396199544165e-05,
+      "loss": 2.483,
+      "step": 23325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793873886696801e-05,
+      "loss": 2.5496,
+      "step": 23330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793785759513352e-05,
+      "loss": 2.6563,
+      "step": 23335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79369761389164e-05,
+      "loss": 2.6583,
+      "step": 23340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793609449832002e-05,
+      "loss": 2.6267,
+      "step": 23345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79352126733478e-05,
+      "loss": 2.6129,
+      "step": 23350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793433066400311e-05,
+      "loss": 2.6208,
+      "step": 23355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793344847028937e-05,
+      "loss": 2.4896,
+      "step": 23360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793256609220996e-05,
+      "loss": 2.5878,
+      "step": 23365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793168352976825e-05,
+      "loss": 2.6034,
+      "step": 23370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793080078296767e-05,
+      "loss": 2.5378,
+      "step": 23375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792991785181161e-05,
+      "loss": 2.5247,
+      "step": 23380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792903473630344e-05,
+      "loss": 2.6275,
+      "step": 23385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792815143644657e-05,
+      "loss": 2.6143,
+      "step": 23390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79272679522444e-05,
+      "loss": 2.5718,
+      "step": 23395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792638428370034e-05,
+      "loss": 2.6967,
+      "step": 23400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792550043081778e-05,
+      "loss": 2.5703,
+      "step": 23405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792461639360012e-05,
+      "loss": 2.4334,
+      "step": 23410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792373217205075e-05,
+      "loss": 2.5382,
+      "step": 23415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792284776617309e-05,
+      "loss": 2.6058,
+      "step": 23420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792196317597051e-05,
+      "loss": 2.4468,
+      "step": 23425
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792107840144644e-05,
+      "loss": 2.5275,
+      "step": 23430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792019344260429e-05,
+      "loss": 2.3915,
+      "step": 23435
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791930829944745e-05,
+      "loss": 2.5992,
+      "step": 23440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791842297197931e-05,
+      "loss": 2.7348,
+      "step": 23445
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791753746020329e-05,
+      "loss": 2.5489,
+      "step": 23450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791665176412279e-05,
+      "loss": 2.6353,
+      "step": 23455
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791576588374123e-05,
+      "loss": 2.6693,
+      "step": 23460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791487981906202e-05,
+      "loss": 2.6378,
+      "step": 23465
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791399357008854e-05,
+      "loss": 2.6021,
+      "step": 23470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791310713682423e-05,
+      "loss": 2.4346,
+      "step": 23475
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791222051927247e-05,
+      "loss": 2.4929,
+      "step": 23480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79113337174367e-05,
+      "loss": 2.5252,
+      "step": 23485
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791044673132031e-05,
+      "loss": 2.6317,
+      "step": 23490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790955956092673e-05,
+      "loss": 2.6959,
+      "step": 23495
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790867220625935e-05,
+      "loss": 2.5874,
+      "step": 23500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79077846673216e-05,
+      "loss": 2.4203,
+      "step": 23505
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790689694411687e-05,
+      "loss": 2.4269,
+      "step": 23510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790600903664862e-05,
+      "loss": 2.5862,
+      "step": 23515
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790512094492021e-05,
+      "loss": 2.5893,
+      "step": 23520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790423266893509e-05,
+      "loss": 2.5705,
+      "step": 23525
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790334420869668e-05,
+      "loss": 2.5333,
+      "step": 23530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790245556420839e-05,
+      "loss": 2.6047,
+      "step": 23535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790156673547361e-05,
+      "loss": 2.614,
+      "step": 23540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790067772249581e-05,
+      "loss": 2.4906,
+      "step": 23545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789978852527836e-05,
+      "loss": 2.6418,
+      "step": 23550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789889914382472e-05,
+      "loss": 2.6166,
+      "step": 23555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789800957813828e-05,
+      "loss": 2.5862,
+      "step": 23560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78971198282225e-05,
+      "loss": 2.5932,
+      "step": 23565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789622989408075e-05,
+      "loss": 2.5854,
+      "step": 23570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789533977571648e-05,
+      "loss": 2.4511,
+      "step": 23575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789444947313313e-05,
+      "loss": 2.6096,
+      "step": 23580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78935589863341e-05,
+      "loss": 2.5574,
+      "step": 23585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789266831532283e-05,
+      "loss": 2.7997,
+      "step": 23590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789177746010274e-05,
+      "loss": 2.5909,
+      "step": 23595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789088642067726e-05,
+      "loss": 2.5564,
+      "step": 23600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78899951970498e-05,
+      "loss": 2.6193,
+      "step": 23605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788910378922381e-05,
+      "loss": 2.6102,
+      "step": 23610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788821219720269e-05,
+      "loss": 2.7019,
+      "step": 23615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78873204209899e-05,
+      "loss": 2.572,
+      "step": 23620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788642846058886e-05,
+      "loss": 2.6759,
+      "step": 23625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788553631600302e-05,
+      "loss": 2.6325,
+      "step": 23630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788464398723576e-05,
+      "loss": 2.6698,
+      "step": 23635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788375147429055e-05,
+      "loss": 2.5995,
+      "step": 23640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788285877717083e-05,
+      "loss": 2.4919,
+      "step": 23645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788196589588e-05,
+      "loss": 2.6969,
+      "step": 23650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788107283042152e-05,
+      "loss": 2.4336,
+      "step": 23655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788017958079883e-05,
+      "loss": 2.7497,
+      "step": 23660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787928614701535e-05,
+      "loss": 2.5341,
+      "step": 23665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787839252907451e-05,
+      "loss": 2.5858,
+      "step": 23670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787749872697978e-05,
+      "loss": 2.4319,
+      "step": 23675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787660474073455e-05,
+      "loss": 2.5225,
+      "step": 23680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78757105703423e-05,
+      "loss": 2.3883,
+      "step": 23685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787481621580646e-05,
+      "loss": 2.6164,
+      "step": 23690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787392167713044e-05,
+      "loss": 2.6453,
+      "step": 23695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787302695431773e-05,
+      "loss": 2.6144,
+      "step": 23700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787213204737171e-05,
+      "loss": 2.5603,
+      "step": 23705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787123695629589e-05,
+      "loss": 2.5032,
+      "step": 23710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787034168109367e-05,
+      "loss": 2.4587,
+      "step": 23715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78694462217685e-05,
+      "loss": 2.6498,
+      "step": 23720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786855057832384e-05,
+      "loss": 2.6017,
+      "step": 23725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786765475076312e-05,
+      "loss": 2.5545,
+      "step": 23730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786675873908977e-05,
+      "loss": 2.4362,
+      "step": 23735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786586254330727e-05,
+      "loss": 2.5734,
+      "step": 23740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786496616341904e-05,
+      "loss": 2.4639,
+      "step": 23745
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786406959942857e-05,
+      "loss": 2.4919,
+      "step": 23750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786317285133926e-05,
+      "loss": 2.7643,
+      "step": 23755
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786227591915457e-05,
+      "loss": 2.6286,
+      "step": 23760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786137880287796e-05,
+      "loss": 2.5988,
+      "step": 23765
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786048150251288e-05,
+      "loss": 2.7235,
+      "step": 23770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785958401806276e-05,
+      "loss": 2.5496,
+      "step": 23775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785868634953111e-05,
+      "loss": 2.5324,
+      "step": 23780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785778849692132e-05,
+      "loss": 2.6539,
+      "step": 23785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785689046023686e-05,
+      "loss": 2.6747,
+      "step": 23790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78559922394812e-05,
+      "loss": 2.5395,
+      "step": 23795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785509383465778e-05,
+      "loss": 2.5746,
+      "step": 23800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785419524577007e-05,
+      "loss": 2.6308,
+      "step": 23805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785329647282153e-05,
+      "loss": 2.6345,
+      "step": 23810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785239751581559e-05,
+      "loss": 2.5518,
+      "step": 23815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785149837475571e-05,
+      "loss": 2.4488,
+      "step": 23820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785059904964537e-05,
+      "loss": 2.5755,
+      "step": 23825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784969954048803e-05,
+      "loss": 2.6598,
+      "step": 23830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784879984728712e-05,
+      "loss": 2.5528,
+      "step": 23835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784789997004615e-05,
+      "loss": 2.5262,
+      "step": 23840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784699990876852e-05,
+      "loss": 2.4941,
+      "step": 23845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784609966345774e-05,
+      "loss": 2.665,
+      "step": 23850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784519923411725e-05,
+      "loss": 2.5328,
+      "step": 23855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78442986207505e-05,
+      "loss": 2.4779,
+      "step": 23860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784339782336101e-05,
+      "loss": 2.5408,
+      "step": 23865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784249684195219e-05,
+      "loss": 2.7517,
+      "step": 23870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784159567652751e-05,
+      "loss": 2.4814,
+      "step": 23875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.784069432709046e-05,
+      "loss": 2.5154,
+      "step": 23880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78397927936445e-05,
+      "loss": 2.6128,
+      "step": 23885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.783889107619309e-05,
+      "loss": 2.7398,
+      "step": 23890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.783798917473971e-05,
+      "loss": 2.5424,
+      "step": 23895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78370870892878e-05,
+      "loss": 2.6,
+      "step": 23900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.783618481984087e-05,
+      "loss": 2.647,
+      "step": 23905
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.783528236640237e-05,
+      "loss": 2.5716,
+      "step": 23910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.783437972897576e-05,
+      "loss": 2.4811,
+      "step": 23915
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.783347690756454e-05,
+      "loss": 2.4467,
+      "step": 23920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.783257390217216e-05,
+      "loss": 2.57,
+      "step": 23925
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.783167071280208e-05,
+      "loss": 2.546,
+      "step": 23930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.783076733945782e-05,
+      "loss": 2.6297,
+      "step": 23935
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78298637821428e-05,
+      "loss": 2.5926,
+      "step": 23940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.782896004086056e-05,
+      "loss": 2.5481,
+      "step": 23945
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.782805611561451e-05,
+      "loss": 2.4712,
+      "step": 23950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.782715200640818e-05,
+      "loss": 2.4071,
+      "step": 23955
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.7826247713245e-05,
+      "loss": 2.5277,
+      "step": 23960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.782534323612848e-05,
+      "loss": 2.6343,
+      "step": 23965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78244385750621e-05,
+      "loss": 2.7132,
+      "step": 23970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.782353373004931e-05,
+      "loss": 2.4462,
+      "step": 23975
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.782262870109363e-05,
+      "loss": 2.4127,
+      "step": 23980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78217234881985e-05,
+      "loss": 2.4978,
+      "step": 23985
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.782081809136744e-05,
+      "loss": 2.546,
+      "step": 23990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.781991251060392e-05,
+      "loss": 2.6086,
+      "step": 23995
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78190067459114e-05,
+      "loss": 2.5661,
+      "step": 24000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78181007972934e-05,
+      "loss": 2.7344,
+      "step": 24005
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.781719466475338e-05,
+      "loss": 2.6031,
+      "step": 24010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.781628834829484e-05,
+      "loss": 2.5951,
+      "step": 24015
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.781538184792125e-05,
+      "loss": 2.5453,
+      "step": 24020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78144751636361e-05,
+      "loss": 2.4294,
+      "step": 24025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.781356829544288e-05,
+      "loss": 2.5685,
+      "step": 24030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.781266124334509e-05,
+      "loss": 2.6505,
+      "step": 24035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.781175400734621e-05,
+      "loss": 2.5769,
+      "step": 24040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.781084658744972e-05,
+      "loss": 2.6395,
+      "step": 24045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.780993898365914e-05,
+      "loss": 2.5367,
+      "step": 24050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.780903119597791e-05,
+      "loss": 2.6112,
+      "step": 24055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.780812322440958e-05,
+      "loss": 2.5257,
+      "step": 24060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.78072150689576e-05,
+      "loss": 2.5249,
+      "step": 24065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.780630672962547e-05,
+      "loss": 2.6857,
+      "step": 24070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.78053982064167e-05,
+      "loss": 2.5238,
+      "step": 24075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.780448949933477e-05,
+      "loss": 2.5179,
+      "step": 24080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.780358060838319e-05,
+      "loss": 2.54,
+      "step": 24085
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.780267153356543e-05,
+      "loss": 2.5133,
+      "step": 24090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.780176227488501e-05,
+      "loss": 2.6021,
+      "step": 24095
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.780085283234543e-05,
+      "loss": 2.6831,
+      "step": 24100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779994320595017e-05,
+      "loss": 2.5558,
+      "step": 24105
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779903339570274e-05,
+      "loss": 2.4794,
+      "step": 24110
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779812340160662e-05,
+      "loss": 2.5029,
+      "step": 24115
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779721322366533e-05,
+      "loss": 2.5098,
+      "step": 24120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779630286188237e-05,
+      "loss": 2.5977,
+      "step": 24125
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779539231626125e-05,
+      "loss": 2.3583,
+      "step": 24130
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779448158680545e-05,
+      "loss": 2.6478,
+      "step": 24135
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779357067351848e-05,
+      "loss": 2.5284,
+      "step": 24140
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779265957640384e-05,
+      "loss": 2.505,
+      "step": 24145
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779174829546506e-05,
+      "loss": 2.4493,
+      "step": 24150
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779083683070563e-05,
+      "loss": 2.5259,
+      "step": 24155
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778992518212902e-05,
+      "loss": 2.6154,
+      "step": 24160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77890133497388e-05,
+      "loss": 2.6001,
+      "step": 24165
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778810133353842e-05,
+      "loss": 2.6261,
+      "step": 24170
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778718913353145e-05,
+      "loss": 2.5026,
+      "step": 24175
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778627674972133e-05,
+      "loss": 2.4204,
+      "step": 24180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778536418211162e-05,
+      "loss": 2.4715,
+      "step": 24185
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778445143070579e-05,
+      "loss": 2.5387,
+      "step": 24190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778353849550739e-05,
+      "loss": 2.6134,
+      "step": 24195
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77826253765199e-05,
+      "loss": 2.6121,
+      "step": 24200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778171207374687e-05,
+      "loss": 2.6979,
+      "step": 24205
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778079858719175e-05,
+      "loss": 2.4024,
+      "step": 24210
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777988491685812e-05,
+      "loss": 2.5511,
+      "step": 24215
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777897106274947e-05,
+      "loss": 2.5217,
+      "step": 24220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777805702486931e-05,
+      "loss": 2.5476,
+      "step": 24225
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777714280322115e-05,
+      "loss": 2.7003,
+      "step": 24230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77762283978085e-05,
+      "loss": 2.5719,
+      "step": 24235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777531380863491e-05,
+      "loss": 2.5941,
+      "step": 24240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777439903570387e-05,
+      "loss": 2.5701,
+      "step": 24245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77734840790189e-05,
+      "loss": 2.6567,
+      "step": 24250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777256893858353e-05,
+      "loss": 2.5708,
+      "step": 24255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777165361440129e-05,
+      "loss": 2.619,
+      "step": 24260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777073810647567e-05,
+      "loss": 2.4939,
+      "step": 24265
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776982241481021e-05,
+      "loss": 2.4336,
+      "step": 24270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776890653940842e-05,
+      "loss": 2.5144,
+      "step": 24275
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776799048027385e-05,
+      "loss": 2.6401,
+      "step": 24280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776707423741e-05,
+      "loss": 2.6651,
+      "step": 24285
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77661578108204e-05,
+      "loss": 2.535,
+      "step": 24290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776524120050858e-05,
+      "loss": 2.4656,
+      "step": 24295
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776432440647805e-05,
+      "loss": 2.6028,
+      "step": 24300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776340742873235e-05,
+      "loss": 2.5828,
+      "step": 24305
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.7762490267275e-05,
+      "loss": 2.6193,
+      "step": 24310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776157292210954e-05,
+      "loss": 2.4858,
+      "step": 24315
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776065539323947e-05,
+      "loss": 2.713,
+      "step": 24320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775973768066835e-05,
+      "loss": 2.4993,
+      "step": 24325
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775881978439971e-05,
+      "loss": 2.5901,
+      "step": 24330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775790170443706e-05,
+      "loss": 2.6586,
+      "step": 24335
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775698344078393e-05,
+      "loss": 2.4984,
+      "step": 24340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775606499344387e-05,
+      "loss": 2.6324,
+      "step": 24345
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775514636242041e-05,
+      "loss": 2.4811,
+      "step": 24350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775422754771706e-05,
+      "loss": 2.5795,
+      "step": 24355
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775330854933739e-05,
+      "loss": 2.7628,
+      "step": 24360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775238936728491e-05,
+      "loss": 2.5261,
+      "step": 24365
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775147000156317e-05,
+      "loss": 2.5095,
+      "step": 24370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77505504521757e-05,
+      "loss": 2.6148,
+      "step": 24375
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774963071912602e-05,
+      "loss": 2.5319,
+      "step": 24380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774871080241767e-05,
+      "loss": 2.5853,
+      "step": 24385
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774779070205423e-05,
+      "loss": 2.6088,
+      "step": 24390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774687041803921e-05,
+      "loss": 2.7299,
+      "step": 24395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774594995037613e-05,
+      "loss": 2.5852,
+      "step": 24400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774502929906856e-05,
+      "loss": 2.7418,
+      "step": 24405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774410846412003e-05,
+      "loss": 2.6293,
+      "step": 24410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774318744553407e-05,
+      "loss": 2.6506,
+      "step": 24415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774226624331425e-05,
+      "loss": 2.4911,
+      "step": 24420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774134485746409e-05,
+      "loss": 2.4905,
+      "step": 24425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774042328798716e-05,
+      "loss": 2.5514,
+      "step": 24430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773950153488697e-05,
+      "loss": 2.5713,
+      "step": 24435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773857959816707e-05,
+      "loss": 2.6905,
+      "step": 24440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773765747783105e-05,
+      "loss": 2.3828,
+      "step": 24445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77367351738824e-05,
+      "loss": 2.5506,
+      "step": 24450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77358126863247e-05,
+      "loss": 2.467,
+      "step": 24455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77348900151615e-05,
+      "loss": 2.5925,
+      "step": 24460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773396716039633e-05,
+      "loss": 2.7377,
+      "step": 24465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773304412203275e-05,
+      "loss": 2.5452,
+      "step": 24470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773212090007431e-05,
+      "loss": 2.3364,
+      "step": 24475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773119749452455e-05,
+      "loss": 2.5384,
+      "step": 24480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773027390538705e-05,
+      "loss": 2.5142,
+      "step": 24485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772935013266533e-05,
+      "loss": 2.5397,
+      "step": 24490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772842617636297e-05,
+      "loss": 2.6353,
+      "step": 24495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77275020364835e-05,
+      "loss": 2.6232,
+      "step": 24500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772657771303048e-05,
+      "loss": 2.651,
+      "step": 24505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772565320600749e-05,
+      "loss": 2.5373,
+      "step": 24510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772472851541806e-05,
+      "loss": 2.5683,
+      "step": 24515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772380364126576e-05,
+      "loss": 2.4376,
+      "step": 24520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772287858355413e-05,
+      "loss": 2.6718,
+      "step": 24525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772195334228673e-05,
+      "loss": 2.5684,
+      "step": 24530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772102791746715e-05,
+      "loss": 2.5699,
+      "step": 24535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772010230909892e-05,
+      "loss": 2.5185,
+      "step": 24540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77191765171856e-05,
+      "loss": 2.6406,
+      "step": 24545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.771825054173076e-05,
+      "loss": 2.631,
+      "step": 24550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.771732438273796e-05,
+      "loss": 2.5558,
+      "step": 24555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.771639804021076e-05,
+      "loss": 2.4809,
+      "step": 24560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.771547151415272e-05,
+      "loss": 2.5581,
+      "step": 24565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77145448045674e-05,
+      "loss": 2.4553,
+      "step": 24570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.771361791145841e-05,
+      "loss": 2.4973,
+      "step": 24575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.771269083482926e-05,
+      "loss": 2.5369,
+      "step": 24580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.771176357468352e-05,
+      "loss": 2.5813,
+      "step": 24585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.771083613102478e-05,
+      "loss": 2.6838,
+      "step": 24590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77099085038566e-05,
+      "loss": 2.4937,
+      "step": 24595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770898069318254e-05,
+      "loss": 2.5259,
+      "step": 24600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770805269900618e-05,
+      "loss": 2.4021,
+      "step": 24605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770712452133109e-05,
+      "loss": 2.6718,
+      "step": 24610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770619616016082e-05,
+      "loss": 2.5862,
+      "step": 24615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770526761549893e-05,
+      "loss": 2.5877,
+      "step": 24620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770433888734906e-05,
+      "loss": 2.5343,
+      "step": 24625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770340997571472e-05,
+      "loss": 2.5445,
+      "step": 24630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77024808805995e-05,
+      "loss": 2.6083,
+      "step": 24635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770155160200698e-05,
+      "loss": 2.6593,
+      "step": 24640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770062213994072e-05,
+      "loss": 2.5048,
+      "step": 24645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769969249440431e-05,
+      "loss": 2.5038,
+      "step": 24650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769876266540132e-05,
+      "loss": 2.7876,
+      "step": 24655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769783265293533e-05,
+      "loss": 2.5721,
+      "step": 24660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76969024570099e-05,
+      "loss": 2.7327,
+      "step": 24665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769597207762862e-05,
+      "loss": 2.5213,
+      "step": 24670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769504151479507e-05,
+      "loss": 2.614,
+      "step": 24675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769411076851284e-05,
+      "loss": 2.7535,
+      "step": 24680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76931798387855e-05,
+      "loss": 2.6463,
+      "step": 24685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76922487256166e-05,
+      "loss": 2.6484,
+      "step": 24690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769131742900976e-05,
+      "loss": 2.4661,
+      "step": 24695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769038594896858e-05,
+      "loss": 2.5643,
+      "step": 24700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.768945428549658e-05,
+      "loss": 2.5626,
+      "step": 24705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76885224385974e-05,
+      "loss": 2.5046,
+      "step": 24710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76875904082746e-05,
+      "loss": 2.4525,
+      "step": 24715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.768665819453175e-05,
+      "loss": 2.6956,
+      "step": 24720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.768572579737248e-05,
+      "loss": 2.6844,
+      "step": 24725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.768479321680033e-05,
+      "loss": 2.7266,
+      "step": 24730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76838604528189e-05,
+      "loss": 2.5143,
+      "step": 24735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76829275054318e-05,
+      "loss": 2.6007,
+      "step": 24740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76819943746426e-05,
+      "loss": 2.4823,
+      "step": 24745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76810610604549e-05,
+      "loss": 2.5033,
+      "step": 24750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.768012756287225e-05,
+      "loss": 2.6039,
+      "step": 24755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76791938818983e-05,
+      "loss": 2.6434,
+      "step": 24760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767826001753659e-05,
+      "loss": 2.6214,
+      "step": 24765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767732596979076e-05,
+      "loss": 2.568,
+      "step": 24770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767639173866435e-05,
+      "loss": 2.4654,
+      "step": 24775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767545732416099e-05,
+      "loss": 2.5957,
+      "step": 24780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767452272628428e-05,
+      "loss": 2.5167,
+      "step": 24785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767358794503779e-05,
+      "loss": 2.4156,
+      "step": 24790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767265298042511e-05,
+      "loss": 2.5658,
+      "step": 24795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767171783244986e-05,
+      "loss": 2.4496,
+      "step": 24800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767078250111563e-05,
+      "loss": 2.5819,
+      "step": 24805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766984698642603e-05,
+      "loss": 2.6793,
+      "step": 24810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766891128838462e-05,
+      "loss": 2.5675,
+      "step": 24815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766797540699504e-05,
+      "loss": 2.6331,
+      "step": 24820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766703934226085e-05,
+      "loss": 2.6388,
+      "step": 24825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766610309418569e-05,
+      "loss": 2.5516,
+      "step": 24830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766516666277315e-05,
+      "loss": 2.5979,
+      "step": 24835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76642300480268e-05,
+      "loss": 2.6262,
+      "step": 24840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76632932499503e-05,
+      "loss": 2.5682,
+      "step": 24845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766235626854719e-05,
+      "loss": 2.6115,
+      "step": 24850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766141910382112e-05,
+      "loss": 2.5724,
+      "step": 24855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766048175577566e-05,
+      "loss": 2.55,
+      "step": 24860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765954422441447e-05,
+      "loss": 2.3899,
+      "step": 24865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76586065097411e-05,
+      "loss": 2.4872,
+      "step": 24870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765766861175918e-05,
+      "loss": 2.578,
+      "step": 24875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765673053047231e-05,
+      "loss": 2.6193,
+      "step": 24880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765579226588411e-05,
+      "loss": 2.6543,
+      "step": 24885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765485381799818e-05,
+      "loss": 2.4952,
+      "step": 24890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765391518681813e-05,
+      "loss": 2.6229,
+      "step": 24895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765297637234759e-05,
+      "loss": 2.519,
+      "step": 24900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765203737459012e-05,
+      "loss": 2.3967,
+      "step": 24905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765109819354938e-05,
+      "loss": 2.569,
+      "step": 24910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765015882922897e-05,
+      "loss": 2.5073,
+      "step": 24915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764921928163248e-05,
+      "loss": 2.5759,
+      "step": 24920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764827955076357e-05,
+      "loss": 2.4073,
+      "step": 24925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764733963662583e-05,
+      "loss": 2.5294,
+      "step": 24930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764639953922285e-05,
+      "loss": 2.5171,
+      "step": 24935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764545925855828e-05,
+      "loss": 2.529,
+      "step": 24940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764451879463573e-05,
+      "loss": 2.6635,
+      "step": 24945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764357814745881e-05,
+      "loss": 2.5623,
+      "step": 24950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764263731703112e-05,
+      "loss": 2.5471,
+      "step": 24955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764169630335633e-05,
+      "loss": 2.4932,
+      "step": 24960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764075510643802e-05,
+      "loss": 2.6697,
+      "step": 24965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763981372627982e-05,
+      "loss": 2.6445,
+      "step": 24970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763887216288535e-05,
+      "loss": 2.5381,
+      "step": 24975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763793041625821e-05,
+      "loss": 2.4886,
+      "step": 24980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763698848640206e-05,
+      "loss": 2.5761,
+      "step": 24985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76360463733205e-05,
+      "loss": 2.5238,
+      "step": 24990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763510407701717e-05,
+      "loss": 2.6104,
+      "step": 24995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763416159749569e-05,
+      "loss": 2.5568,
+      "step": 25000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763321893475968e-05,
+      "loss": 2.6189,
+      "step": 25005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763227608881274e-05,
+      "loss": 2.4007,
+      "step": 25010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763133305965855e-05,
+      "loss": 2.6516,
+      "step": 25015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763038984730068e-05,
+      "loss": 2.4661,
+      "step": 25020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762944645174281e-05,
+      "loss": 2.4073,
+      "step": 25025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762850287298855e-05,
+      "loss": 2.5631,
+      "step": 25030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762755911104151e-05,
+      "loss": 2.449,
+      "step": 25035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762661516590535e-05,
+      "loss": 2.7344,
+      "step": 25040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762567103758367e-05,
+      "loss": 2.3606,
+      "step": 25045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762472672608012e-05,
+      "loss": 2.5916,
+      "step": 25050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762378223139832e-05,
+      "loss": 2.4087,
+      "step": 25055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762283755354191e-05,
+      "loss": 2.5648,
+      "step": 25060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762189269251453e-05,
+      "loss": 2.5404,
+      "step": 25065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762094764831983e-05,
+      "loss": 2.4615,
+      "step": 25070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762000242096141e-05,
+      "loss": 2.5992,
+      "step": 25075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761905701044292e-05,
+      "loss": 2.5876,
+      "step": 25080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.7618111416768e-05,
+      "loss": 2.5544,
+      "step": 25085
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761716563994028e-05,
+      "loss": 2.4556,
+      "step": 25090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76162196799634e-05,
+      "loss": 2.5759,
+      "step": 25095
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761527353684098e-05,
+      "loss": 2.587,
+      "step": 25100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761432721057671e-05,
+      "loss": 2.5802,
+      "step": 25105
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761338070117419e-05,
+      "loss": 2.5694,
+      "step": 25110
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761243400863705e-05,
+      "loss": 2.5933,
+      "step": 25115
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761148713296897e-05,
+      "loss": 2.6009,
+      "step": 25120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761054007417356e-05,
+      "loss": 2.5969,
+      "step": 25125
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760959283225449e-05,
+      "loss": 2.6192,
+      "step": 25130
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760864540721537e-05,
+      "loss": 2.6285,
+      "step": 25135
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760769779905987e-05,
+      "loss": 2.5924,
+      "step": 25140
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760675000779163e-05,
+      "loss": 2.5342,
+      "step": 25145
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760580203341429e-05,
+      "loss": 2.5366,
+      "step": 25150
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76048538759315e-05,
+      "loss": 2.6944,
+      "step": 25155
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76039055353469e-05,
+      "loss": 2.4967,
+      "step": 25160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760295701166415e-05,
+      "loss": 2.4186,
+      "step": 25165
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760200830488688e-05,
+      "loss": 2.3471,
+      "step": 25170
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760105941501875e-05,
+      "loss": 2.6597,
+      "step": 25175
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760011034206342e-05,
+      "loss": 2.6615,
+      "step": 25180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.759916108602454e-05,
+      "loss": 2.5915,
+      "step": 25185
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.759821164690574e-05,
+      "loss": 2.4086,
+      "step": 25190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.759726202471067e-05,
+      "loss": 2.6431,
+      "step": 25195
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.759631221944301e-05,
+      "loss": 2.6344,
+      "step": 25200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75953622311064e-05,
+      "loss": 2.6066,
+      "step": 25205
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75944120597045e-05,
+      "loss": 2.6318,
+      "step": 25210
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.759346170524095e-05,
+      "loss": 2.6329,
+      "step": 25215
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75925111677194e-05,
+      "loss": 2.5897,
+      "step": 25220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.759156044714352e-05,
+      "loss": 2.7473,
+      "step": 25225
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.759060954351698e-05,
+      "loss": 2.4778,
+      "step": 25230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758965845684343e-05,
+      "loss": 2.4734,
+      "step": 25235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75887071871265e-05,
+      "loss": 2.5703,
+      "step": 25240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758775573436989e-05,
+      "loss": 2.6108,
+      "step": 25245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758680409857722e-05,
+      "loss": 2.6273,
+      "step": 25250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758585227975219e-05,
+      "loss": 2.6501,
+      "step": 25255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758490027789842e-05,
+      "loss": 2.6579,
+      "step": 25260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758394809301961e-05,
+      "loss": 2.4842,
+      "step": 25265
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75829957251194e-05,
+      "loss": 2.5395,
+      "step": 25270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758204317420145e-05,
+      "loss": 2.5528,
+      "step": 25275
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758109044026944e-05,
+      "loss": 2.5956,
+      "step": 25280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758013752332702e-05,
+      "loss": 2.5771,
+      "step": 25285
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757918442337786e-05,
+      "loss": 2.5322,
+      "step": 25290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757823114042564e-05,
+      "loss": 2.5693,
+      "step": 25295
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757727767447401e-05,
+      "loss": 2.5277,
+      "step": 25300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757632402552663e-05,
+      "loss": 2.4018,
+      "step": 25305
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75753701935872e-05,
+      "loss": 2.6467,
+      "step": 25310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757441617865935e-05,
+      "loss": 2.4143,
+      "step": 25315
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757346198074677e-05,
+      "loss": 2.5668,
+      "step": 25320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757250759985313e-05,
+      "loss": 2.5335,
+      "step": 25325
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757155303598211e-05,
+      "loss": 2.5344,
+      "step": 25330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757059828913735e-05,
+      "loss": 2.6387,
+      "step": 25335
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756964335932255e-05,
+      "loss": 2.6403,
+      "step": 25340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756868824654137e-05,
+      "loss": 2.5892,
+      "step": 25345
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75677329507975e-05,
+      "loss": 2.6564,
+      "step": 25350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75667774720946e-05,
+      "loss": 2.3799,
+      "step": 25355
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756582181043634e-05,
+      "loss": 2.5954,
+      "step": 25360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756486596582641e-05,
+      "loss": 2.4251,
+      "step": 25365
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756390993826849e-05,
+      "loss": 2.601,
+      "step": 25370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756295372776622e-05,
+      "loss": 2.4051,
+      "step": 25375
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756199733432335e-05,
+      "loss": 2.6348,
+      "step": 25380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756104075794348e-05,
+      "loss": 2.5976,
+      "step": 25385
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756008399863032e-05,
+      "loss": 2.5961,
+      "step": 25390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755912705638757e-05,
+      "loss": 2.5663,
+      "step": 25395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755816993121888e-05,
+      "loss": 2.6809,
+      "step": 25400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755721262312795e-05,
+      "loss": 2.6456,
+      "step": 25405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755625513211846e-05,
+      "loss": 2.3916,
+      "step": 25410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75552974581941e-05,
+      "loss": 2.6,
+      "step": 25415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755433960135852e-05,
+      "loss": 2.5064,
+      "step": 25420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755338156161544e-05,
+      "loss": 2.431,
+      "step": 25425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755242333896853e-05,
+      "loss": 2.4808,
+      "step": 25430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755146493342149e-05,
+      "loss": 2.6446,
+      "step": 25435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755050634497798e-05,
+      "loss": 2.5005,
+      "step": 25440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75495475736417e-05,
+      "loss": 2.7064,
+      "step": 25445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.754858861941635e-05,
+      "loss": 2.632,
+      "step": 25450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75476294823056e-05,
+      "loss": 2.4986,
+      "step": 25455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.754667016231315e-05,
+      "loss": 2.6115,
+      "step": 25460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.754571065944269e-05,
+      "loss": 2.51,
+      "step": 25465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75447509736979e-05,
+      "loss": 2.5412,
+      "step": 25470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.754379110508247e-05,
+      "loss": 2.4819,
+      "step": 25475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.754283105360012e-05,
+      "loss": 2.5453,
+      "step": 25480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.754187081925452e-05,
+      "loss": 2.5695,
+      "step": 25485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.754091040204935e-05,
+      "loss": 2.5999,
+      "step": 25490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753994980198832e-05,
+      "loss": 2.6668,
+      "step": 25495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753898901907515e-05,
+      "loss": 2.5881,
+      "step": 25500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753802805331348e-05,
+      "loss": 2.6196,
+      "step": 25505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753706690470706e-05,
+      "loss": 2.5361,
+      "step": 25510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753610557325956e-05,
+      "loss": 2.5391,
+      "step": 25515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753514405897468e-05,
+      "loss": 2.6415,
+      "step": 25520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753418236185612e-05,
+      "loss": 2.6547,
+      "step": 25525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753322048190756e-05,
+      "loss": 2.5497,
+      "step": 25530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753225841913273e-05,
+      "loss": 2.4481,
+      "step": 25535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753129617353532e-05,
+      "loss": 2.4639,
+      "step": 25540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753033374511903e-05,
+      "loss": 2.5255,
+      "step": 25545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752937113388756e-05,
+      "loss": 2.4571,
+      "step": 25550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752840833984462e-05,
+      "loss": 2.5666,
+      "step": 25555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75274453629939e-05,
+      "loss": 2.5276,
+      "step": 25560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752648220333911e-05,
+      "loss": 2.5879,
+      "step": 25565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752551886088394e-05,
+      "loss": 2.4972,
+      "step": 25570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752455533563214e-05,
+      "loss": 2.4788,
+      "step": 25575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752359162758737e-05,
+      "loss": 2.4932,
+      "step": 25580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752262773675336e-05,
+      "loss": 2.6201,
+      "step": 25585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75216636631338e-05,
+      "loss": 2.5365,
+      "step": 25590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752069940673241e-05,
+      "loss": 2.547,
+      "step": 25595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751973496755291e-05,
+      "loss": 2.5649,
+      "step": 25600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751877034559898e-05,
+      "loss": 2.6804,
+      "step": 25605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751780554087437e-05,
+      "loss": 2.4938,
+      "step": 25610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751684055338275e-05,
+      "loss": 2.4751,
+      "step": 25615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751587538312785e-05,
+      "loss": 2.6164,
+      "step": 25620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751491003011339e-05,
+      "loss": 2.4644,
+      "step": 25625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751394449434306e-05,
+      "loss": 2.5684,
+      "step": 25630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751297877582059e-05,
+      "loss": 2.5421,
+      "step": 25635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751201287454972e-05,
+      "loss": 2.5131,
+      "step": 25640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751104679053412e-05,
+      "loss": 2.6433,
+      "step": 25645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751008052377753e-05,
+      "loss": 2.4866,
+      "step": 25650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750911407428365e-05,
+      "loss": 2.4605,
+      "step": 25655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75081474420562e-05,
+      "loss": 2.463,
+      "step": 25660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750718062709893e-05,
+      "loss": 2.5991,
+      "step": 25665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750621362941552e-05,
+      "loss": 2.6173,
+      "step": 25670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750524644900971e-05,
+      "loss": 2.5703,
+      "step": 25675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750427908588521e-05,
+      "loss": 2.6071,
+      "step": 25680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750331154004575e-05,
+      "loss": 2.6449,
+      "step": 25685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750234381149504e-05,
+      "loss": 2.6152,
+      "step": 25690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75013759002368e-05,
+      "loss": 2.3576,
+      "step": 25695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750040780627478e-05,
+      "loss": 2.616,
+      "step": 25700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749943952961267e-05,
+      "loss": 2.4635,
+      "step": 25705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749847107025421e-05,
+      "loss": 2.5247,
+      "step": 25710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749750242820312e-05,
+      "loss": 2.6628,
+      "step": 25715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749653360346315e-05,
+      "loss": 2.4987,
+      "step": 25720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749556459603799e-05,
+      "loss": 2.4359,
+      "step": 25725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749459540593138e-05,
+      "loss": 2.4528,
+      "step": 25730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749362603314707e-05,
+      "loss": 2.53,
+      "step": 25735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749265647768875e-05,
+      "loss": 2.4394,
+      "step": 25740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749168673956018e-05,
+      "loss": 2.6036,
+      "step": 25745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749071681876507e-05,
+      "loss": 2.6028,
+      "step": 25750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748974671530717e-05,
+      "loss": 2.5225,
+      "step": 25755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748877642919019e-05,
+      "loss": 2.4272,
+      "step": 25760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748780596041788e-05,
+      "loss": 2.5524,
+      "step": 25765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748683530899397e-05,
+      "loss": 2.4858,
+      "step": 25770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748586447492217e-05,
+      "loss": 2.4564,
+      "step": 25775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748489345820624e-05,
+      "loss": 2.5159,
+      "step": 25780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74839222588499e-05,
+      "loss": 2.4919,
+      "step": 25785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748295087685691e-05,
+      "loss": 2.6478,
+      "step": 25790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748197931223097e-05,
+      "loss": 2.5953,
+      "step": 25795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748100756497585e-05,
+      "loss": 2.5611,
+      "step": 25800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748003563509527e-05,
+      "loss": 2.448,
+      "step": 25805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747906352259297e-05,
+      "loss": 2.5049,
+      "step": 25810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747809122747268e-05,
+      "loss": 2.3954,
+      "step": 25815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747711874973817e-05,
+      "loss": 2.6285,
+      "step": 25820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747614608939315e-05,
+      "loss": 2.4944,
+      "step": 25825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747517324644137e-05,
+      "loss": 2.6166,
+      "step": 25830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747420022088659e-05,
+      "loss": 2.7955,
+      "step": 25835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747322701273253e-05,
+      "loss": 2.5578,
+      "step": 25840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747225362198293e-05,
+      "loss": 2.4584,
+      "step": 25845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747128004864156e-05,
+      "loss": 2.7157,
+      "step": 25850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747030629271213e-05,
+      "loss": 2.5406,
+      "step": 25855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746933235419841e-05,
+      "loss": 2.5279,
+      "step": 25860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746835823310415e-05,
+      "loss": 2.6853,
+      "step": 25865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746738392943307e-05,
+      "loss": 2.4397,
+      "step": 25870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746640944318894e-05,
+      "loss": 2.6517,
+      "step": 25875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746543477437551e-05,
+      "loss": 2.7018,
+      "step": 25880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746445992299651e-05,
+      "loss": 2.4696,
+      "step": 25885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74634848890557e-05,
+      "loss": 2.6174,
+      "step": 25890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746250967255684e-05,
+      "loss": 2.5731,
+      "step": 25895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746153427350366e-05,
+      "loss": 2.5556,
+      "step": 25900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746055869189995e-05,
+      "loss": 2.6169,
+      "step": 25905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745958292774941e-05,
+      "loss": 2.4593,
+      "step": 25910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745860698105582e-05,
+      "loss": 2.4916,
+      "step": 25915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745763085182293e-05,
+      "loss": 2.3658,
+      "step": 25920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74566545400545e-05,
+      "loss": 2.5637,
+      "step": 25925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745567804575429e-05,
+      "loss": 2.6298,
+      "step": 25930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745470136892605e-05,
+      "loss": 2.6033,
+      "step": 25935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745372450957353e-05,
+      "loss": 2.5574,
+      "step": 25940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745274746770049e-05,
+      "loss": 2.7575,
+      "step": 25945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745177024331069e-05,
+      "loss": 2.4838,
+      "step": 25950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745079283640789e-05,
+      "loss": 2.6259,
+      "step": 25955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744981524699584e-05,
+      "loss": 2.5079,
+      "step": 25960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744883747507832e-05,
+      "loss": 2.561,
+      "step": 25965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744785952065908e-05,
+      "loss": 2.6423,
+      "step": 25970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744688138374187e-05,
+      "loss": 2.3528,
+      "step": 25975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744590306433047e-05,
+      "loss": 2.5436,
+      "step": 25980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744492456242864e-05,
+      "loss": 2.6477,
+      "step": 25985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744394587804013e-05,
+      "loss": 2.4165,
+      "step": 25990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744296701116871e-05,
+      "loss": 2.5699,
+      "step": 25995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744198796181815e-05,
+      "loss": 2.5696,
+      "step": 26000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744100872999222e-05,
+      "loss": 2.5632,
+      "step": 26005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744002931569468e-05,
+      "loss": 2.6597,
+      "step": 26010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743904971892929e-05,
+      "loss": 2.4994,
+      "step": 26015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743806993969982e-05,
+      "loss": 2.5236,
+      "step": 26020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743708997801006e-05,
+      "loss": 2.6055,
+      "step": 26025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743610983386376e-05,
+      "loss": 2.6242,
+      "step": 26030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743512950726469e-05,
+      "loss": 2.6808,
+      "step": 26035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743414899821663e-05,
+      "loss": 2.4714,
+      "step": 26040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743316830672334e-05,
+      "loss": 2.4078,
+      "step": 26045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74321874327886e-05,
+      "loss": 2.691,
+      "step": 26050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743120637641616e-05,
+      "loss": 2.5125,
+      "step": 26055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743022513760983e-05,
+      "loss": 2.5659,
+      "step": 26060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742924371637337e-05,
+      "loss": 2.5486,
+      "step": 26065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742826211271054e-05,
+      "loss": 2.576,
+      "step": 26070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742728032662514e-05,
+      "loss": 2.5668,
+      "step": 26075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742629835812092e-05,
+      "loss": 2.5869,
+      "step": 26080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742531620720168e-05,
+      "loss": 2.5743,
+      "step": 26085
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74243338738712e-05,
+      "loss": 2.5258,
+      "step": 26090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742335135813323e-05,
+      "loss": 2.4583,
+      "step": 26095
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742236865999158e-05,
+      "loss": 2.6402,
+      "step": 26100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742138577944999e-05,
+      "loss": 2.4407,
+      "step": 26105
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742040271651228e-05,
+      "loss": 2.594,
+      "step": 26110
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741941947118222e-05,
+      "loss": 2.5251,
+      "step": 26115
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741843604346359e-05,
+      "loss": 2.4923,
+      "step": 26120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741745243336015e-05,
+      "loss": 2.6185,
+      "step": 26125
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741646864087573e-05,
+      "loss": 2.6055,
+      "step": 26130
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741548466601408e-05,
+      "loss": 2.7523,
+      "step": 26135
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741450050877901e-05,
+      "loss": 2.4492,
+      "step": 26140
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741351616917427e-05,
+      "loss": 2.5061,
+      "step": 26145
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741253164720366e-05,
+      "loss": 2.4514,
+      "step": 26150
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741154694287098e-05,
+      "loss": 2.3437,
+      "step": 26155
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741056205618e-05,
+      "loss": 2.4308,
+      "step": 26160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740957698713453e-05,
+      "loss": 2.5335,
+      "step": 26165
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740859173573834e-05,
+      "loss": 2.6047,
+      "step": 26170
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740760630199522e-05,
+      "loss": 2.5444,
+      "step": 26175
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740662068590898e-05,
+      "loss": 2.5896,
+      "step": 26180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740563488748339e-05,
+      "loss": 2.5989,
+      "step": 26185
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740464890672224e-05,
+      "loss": 2.4778,
+      "step": 26190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740366274362933e-05,
+      "loss": 2.6669,
+      "step": 26195
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740267639820847e-05,
+      "loss": 2.6054,
+      "step": 26200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740168987046343e-05,
+      "loss": 2.5875,
+      "step": 26205
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.7400703160398e-05,
+      "loss": 2.5557,
+      "step": 26210
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.7399716268016e-05,
+      "loss": 2.5622,
+      "step": 26215
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739872919332121e-05,
+      "loss": 2.5468,
+      "step": 26220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739774193631743e-05,
+      "loss": 2.584,
+      "step": 26225
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739675449700846e-05,
+      "loss": 2.4463,
+      "step": 26230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739576687539809e-05,
+      "loss": 2.6635,
+      "step": 26235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739477907149012e-05,
+      "loss": 2.6644,
+      "step": 26240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739379108528836e-05,
+      "loss": 2.7353,
+      "step": 26245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73928029167966e-05,
+      "loss": 2.6066,
+      "step": 26250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739181456601864e-05,
+      "loss": 2.6514,
+      "step": 26255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739082603295829e-05,
+      "loss": 2.5561,
+      "step": 26260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738983731761935e-05,
+      "loss": 2.5622,
+      "step": 26265
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738884842000562e-05,
+      "loss": 2.6184,
+      "step": 26270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73878593401209e-05,
+      "loss": 2.5581,
+      "step": 26275
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738687007796899e-05,
+      "loss": 2.6309,
+      "step": 26280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738588063355371e-05,
+      "loss": 2.5972,
+      "step": 26285
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738489100687886e-05,
+      "loss": 2.4955,
+      "step": 26290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738390119794824e-05,
+      "loss": 2.4756,
+      "step": 26295
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738291120676568e-05,
+      "loss": 2.504,
+      "step": 26300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738192103333495e-05,
+      "loss": 2.5276,
+      "step": 26305
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73809306776599e-05,
+      "loss": 2.6266,
+      "step": 26310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73799401397443e-05,
+      "loss": 2.4863,
+      "step": 26315
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737894941959198e-05,
+      "loss": 2.5677,
+      "step": 26320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737795851720674e-05,
+      "loss": 2.4915,
+      "step": 26325
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737696743259242e-05,
+      "loss": 2.7066,
+      "step": 26330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737597616575279e-05,
+      "loss": 2.7077,
+      "step": 26335
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73749847166917e-05,
+      "loss": 2.5631,
+      "step": 26340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737399308541294e-05,
+      "loss": 2.6534,
+      "step": 26345
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737300127192033e-05,
+      "loss": 2.5311,
+      "step": 26350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73720092762177e-05,
+      "loss": 2.6065,
+      "step": 26355
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737101709830883e-05,
+      "loss": 2.514,
+      "step": 26360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737002473819757e-05,
+      "loss": 2.503,
+      "step": 26365
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.736903219588773e-05,
+      "loss": 2.6359,
+      "step": 26370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73680394713831e-05,
+      "loss": 2.4847,
+      "step": 26375
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.736704656468756e-05,
+      "loss": 2.6323,
+      "step": 26380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.736605347580487e-05,
+      "loss": 2.6514,
+      "step": 26385
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.736506020473886e-05,
+      "loss": 2.6062,
+      "step": 26390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.736406675149337e-05,
+      "loss": 2.4477,
+      "step": 26395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.736307311607221e-05,
+      "loss": 2.4107,
+      "step": 26400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.736207929847921e-05,
+      "loss": 2.4958,
+      "step": 26405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.736108529871817e-05,
+      "loss": 2.5667,
+      "step": 26410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.736009111679293e-05,
+      "loss": 2.4981,
+      "step": 26415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.735909675270731e-05,
+      "loss": 2.6285,
+      "step": 26420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.735810220646515e-05,
+      "loss": 2.5248,
+      "step": 26425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.735710747807026e-05,
+      "loss": 2.6481,
+      "step": 26430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.735611256752647e-05,
+      "loss": 2.5206,
+      "step": 26435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73551174748376e-05,
+      "loss": 2.4503,
+      "step": 26440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.735412220000748e-05,
+      "loss": 2.6214,
+      "step": 26445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.735312674303996e-05,
+      "loss": 2.6762,
+      "step": 26450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.735213110393882e-05,
+      "loss": 2.5787,
+      "step": 26455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.735113528270794e-05,
+      "loss": 2.5874,
+      "step": 26460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.735013927935112e-05,
+      "loss": 2.7367,
+      "step": 26465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.734914309387219e-05,
+      "loss": 2.615,
+      "step": 26470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.734814672627501e-05,
+      "loss": 2.5517,
+      "step": 26475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.734715017656339e-05,
+      "loss": 2.5334,
+      "step": 26480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.734615344474118e-05,
+      "loss": 2.6221,
+      "step": 26485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.734515653081218e-05,
+      "loss": 2.5789,
+      "step": 26490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.734415943478025e-05,
+      "loss": 2.6001,
+      "step": 26495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.734316215664923e-05,
+      "loss": 2.5513,
+      "step": 26500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.734216469642293e-05,
+      "loss": 2.7571,
+      "step": 26505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.734116705410521e-05,
+      "loss": 2.4355,
+      "step": 26510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73401692296999e-05,
+      "loss": 2.5944,
+      "step": 26515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733917122321083e-05,
+      "loss": 2.5652,
+      "step": 26520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733817303464187e-05,
+      "loss": 2.5271,
+      "step": 26525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733717466399681e-05,
+      "loss": 2.61,
+      "step": 26530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733617611127954e-05,
+      "loss": 2.6237,
+      "step": 26535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733517737649385e-05,
+      "loss": 2.6572,
+      "step": 26540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733417845964363e-05,
+      "loss": 2.6109,
+      "step": 26545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733317936073268e-05,
+      "loss": 2.6375,
+      "step": 26550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733218007976488e-05,
+      "loss": 2.7152,
+      "step": 26555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733118061674405e-05,
+      "loss": 2.6244,
+      "step": 26560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.733018097167404e-05,
+      "loss": 2.5361,
+      "step": 26565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.732918114455869e-05,
+      "loss": 2.5294,
+      "step": 26570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.732818113540186e-05,
+      "loss": 2.5226,
+      "step": 26575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.732718094420738e-05,
+      "loss": 2.4866,
+      "step": 26580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73261805709791e-05,
+      "loss": 2.5451,
+      "step": 26585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73251800157209e-05,
+      "loss": 2.5319,
+      "step": 26590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.732417927843658e-05,
+      "loss": 2.6141,
+      "step": 26595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.732317835913001e-05,
+      "loss": 2.5152,
+      "step": 26600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.732217725780503e-05,
+      "loss": 2.5099,
+      "step": 26605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.732117597446552e-05,
+      "loss": 2.5723,
+      "step": 26610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.73201745091153e-05,
+      "loss": 2.5693,
+      "step": 26615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731917286175824e-05,
+      "loss": 2.4951,
+      "step": 26620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731817103239817e-05,
+      "loss": 2.5539,
+      "step": 26625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731716902103898e-05,
+      "loss": 2.5771,
+      "step": 26630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.73161668276845e-05,
+      "loss": 2.5265,
+      "step": 26635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731516445233857e-05,
+      "loss": 2.5436,
+      "step": 26640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731416189500508e-05,
+      "loss": 2.5966,
+      "step": 26645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731315915568786e-05,
+      "loss": 2.6452,
+      "step": 26650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731215623439079e-05,
+      "loss": 2.5633,
+      "step": 26655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.73111531311177e-05,
+      "loss": 2.6502,
+      "step": 26660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731014984587246e-05,
+      "loss": 2.5296,
+      "step": 26665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730914637865894e-05,
+      "loss": 2.702,
+      "step": 26670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730814272948098e-05,
+      "loss": 2.5988,
+      "step": 26675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730713889834246e-05,
+      "loss": 2.5266,
+      "step": 26680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730613488524722e-05,
+      "loss": 2.5726,
+      "step": 26685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730513069019916e-05,
+      "loss": 2.5241,
+      "step": 26690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730412631320209e-05,
+      "loss": 2.5826,
+      "step": 26695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730312175425991e-05,
+      "loss": 2.4555,
+      "step": 26700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730211701337647e-05,
+      "loss": 2.5472,
+      "step": 26705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730111209055564e-05,
+      "loss": 2.5588,
+      "step": 26710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730010698580127e-05,
+      "loss": 2.4978,
+      "step": 26715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729910169911726e-05,
+      "loss": 2.5518,
+      "step": 26720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729809623050745e-05,
+      "loss": 2.6382,
+      "step": 26725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72970905799757e-05,
+      "loss": 2.5277,
+      "step": 26730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729608474752589e-05,
+      "loss": 2.5817,
+      "step": 26735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72950787331619e-05,
+      "loss": 2.5937,
+      "step": 26740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72940725368876e-05,
+      "loss": 2.5885,
+      "step": 26745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729306615870683e-05,
+      "loss": 2.5464,
+      "step": 26750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729205959862348e-05,
+      "loss": 2.6715,
+      "step": 26755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729105285664141e-05,
+      "loss": 2.5504,
+      "step": 26760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729004593276453e-05,
+      "loss": 2.4789,
+      "step": 26765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.728903882699668e-05,
+      "loss": 2.5897,
+      "step": 26770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.728803153934174e-05,
+      "loss": 2.6304,
+      "step": 26775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.728702406980357e-05,
+      "loss": 2.5575,
+      "step": 26780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.728601641838608e-05,
+      "loss": 2.425,
+      "step": 26785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72850085850931e-05,
+      "loss": 2.6935,
+      "step": 26790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.728400056992854e-05,
+      "loss": 2.5367,
+      "step": 26795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.728299237289628e-05,
+      "loss": 2.6186,
+      "step": 26800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.728198399400018e-05,
+      "loss": 2.6537,
+      "step": 26805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.728097543324412e-05,
+      "loss": 2.5942,
+      "step": 26810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727996669063199e-05,
+      "loss": 2.5616,
+      "step": 26815
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727895776616767e-05,
+      "loss": 2.5041,
+      "step": 26820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727794865985503e-05,
+      "loss": 2.5446,
+      "step": 26825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727693937169794e-05,
+      "loss": 2.5827,
+      "step": 26830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727592990170031e-05,
+      "loss": 2.5062,
+      "step": 26835
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.7274920249866e-05,
+      "loss": 2.6239,
+      "step": 26840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727391041619892e-05,
+      "loss": 2.5955,
+      "step": 26845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727290040070294e-05,
+      "loss": 2.5887,
+      "step": 26850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727189020338193e-05,
+      "loss": 2.6504,
+      "step": 26855
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727087982423979e-05,
+      "loss": 2.5428,
+      "step": 26860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726986926328041e-05,
+      "loss": 2.5678,
+      "step": 26865
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726885852050766e-05,
+      "loss": 2.5649,
+      "step": 26870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726784759592547e-05,
+      "loss": 2.3781,
+      "step": 26875
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726683648953767e-05,
+      "loss": 2.5714,
+      "step": 26880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726582520134817e-05,
+      "loss": 2.5954,
+      "step": 26885
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726481373136088e-05,
+      "loss": 2.4864,
+      "step": 26890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726380207957968e-05,
+      "loss": 2.5442,
+      "step": 26895
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726279024600845e-05,
+      "loss": 2.5487,
+      "step": 26900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72617782306511e-05,
+      "loss": 2.5058,
+      "step": 26905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72607660335115e-05,
+      "loss": 2.5091,
+      "step": 26910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725975365459355e-05,
+      "loss": 2.5914,
+      "step": 26915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725874109390117e-05,
+      "loss": 2.5247,
+      "step": 26920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725772835143821e-05,
+      "loss": 2.5,
+      "step": 26925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72567154272086e-05,
+      "loss": 2.6517,
+      "step": 26930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725570232121623e-05,
+      "loss": 2.5555,
+      "step": 26935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725468903346497e-05,
+      "loss": 2.6065,
+      "step": 26940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725367556395876e-05,
+      "loss": 2.6224,
+      "step": 26945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725266191270149e-05,
+      "loss": 2.3997,
+      "step": 26950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725164807969703e-05,
+      "loss": 2.5797,
+      "step": 26955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725063406494927e-05,
+      "loss": 2.642,
+      "step": 26960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724961986846217e-05,
+      "loss": 2.654,
+      "step": 26965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724860549023957e-05,
+      "loss": 2.3979,
+      "step": 26970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724759093028543e-05,
+      "loss": 2.4666,
+      "step": 26975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72465761886036e-05,
+      "loss": 2.6936,
+      "step": 26980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724556126519801e-05,
+      "loss": 2.4854,
+      "step": 26985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724454616007255e-05,
+      "loss": 2.6734,
+      "step": 26990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724353087323112e-05,
+      "loss": 2.3238,
+      "step": 26995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724251540467767e-05,
+      "loss": 2.6948,
+      "step": 27000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724149975441606e-05,
+      "loss": 2.439,
+      "step": 27005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724048392245019e-05,
+      "loss": 2.4563,
+      "step": 27010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723946790878399e-05,
+      "loss": 2.6906,
+      "step": 27015
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723845171342138e-05,
+      "loss": 2.4326,
+      "step": 27020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723743533636624e-05,
+      "loss": 2.7628,
+      "step": 27025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72364187776225e-05,
+      "loss": 2.6197,
+      "step": 27030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723540203719406e-05,
+      "loss": 2.6548,
+      "step": 27035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723438511508483e-05,
+      "loss": 2.5699,
+      "step": 27040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723336801129872e-05,
+      "loss": 2.5592,
+      "step": 27045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723235072583965e-05,
+      "loss": 2.5565,
+      "step": 27050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723133325871153e-05,
+      "loss": 2.5685,
+      "step": 27055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723031560991825e-05,
+      "loss": 2.5645,
+      "step": 27060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722929777946377e-05,
+      "loss": 2.6001,
+      "step": 27065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722827976735198e-05,
+      "loss": 2.4867,
+      "step": 27070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722726157358679e-05,
+      "loss": 2.6132,
+      "step": 27075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722624319817213e-05,
+      "loss": 2.6629,
+      "step": 27080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72252246411119e-05,
+      "loss": 2.609,
+      "step": 27085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722420590241002e-05,
+      "loss": 2.5299,
+      "step": 27090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722318698207044e-05,
+      "loss": 2.5812,
+      "step": 27095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722216788009704e-05,
+      "loss": 2.5594,
+      "step": 27100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722114859649376e-05,
+      "loss": 2.5365,
+      "step": 27105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72201291312645e-05,
+      "loss": 2.4612,
+      "step": 27110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721910948441321e-05,
+      "loss": 2.5179,
+      "step": 27115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72180896559438e-05,
+      "loss": 2.6211,
+      "step": 27120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721706964586019e-05,
+      "loss": 2.5694,
+      "step": 27125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721604945416629e-05,
+      "loss": 2.5004,
+      "step": 27130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721502908086606e-05,
+      "loss": 2.6036,
+      "step": 27135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721400852596339e-05,
+      "loss": 2.6072,
+      "step": 27140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721298778946222e-05,
+      "loss": 2.5434,
+      "step": 27145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721196687136647e-05,
+      "loss": 2.6537,
+      "step": 27150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721094577168006e-05,
+      "loss": 2.5177,
+      "step": 27155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720992449040694e-05,
+      "loss": 2.5932,
+      "step": 27160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720890302755103e-05,
+      "loss": 2.5508,
+      "step": 27165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720788138311625e-05,
+      "loss": 2.5187,
+      "step": 27170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720685955710654e-05,
+      "loss": 2.616,
+      "step": 27175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720583754952582e-05,
+      "loss": 2.6687,
+      "step": 27180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720481536037802e-05,
+      "loss": 2.4414,
+      "step": 27185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720379298966709e-05,
+      "loss": 2.4346,
+      "step": 27190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720277043739694e-05,
+      "loss": 2.4126,
+      "step": 27195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720174770357151e-05,
+      "loss": 2.3665,
+      "step": 27200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720072478819476e-05,
+      "loss": 2.4263,
+      "step": 27205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719970169127057e-05,
+      "loss": 2.6526,
+      "step": 27210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719867841280292e-05,
+      "loss": 2.583,
+      "step": 27215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719765495279573e-05,
+      "loss": 2.4905,
+      "step": 27220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719663131125294e-05,
+      "loss": 2.4766,
+      "step": 27225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719560748817849e-05,
+      "loss": 2.6881,
+      "step": 27230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719458348357629e-05,
+      "loss": 2.509,
+      "step": 27235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719355929745033e-05,
+      "loss": 2.671,
+      "step": 27240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719253492980451e-05,
+      "loss": 2.4793,
+      "step": 27245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719151038064278e-05,
+      "loss": 2.5036,
+      "step": 27250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719048564996909e-05,
+      "loss": 2.5869,
+      "step": 27255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718946073778737e-05,
+      "loss": 2.5709,
+      "step": 27260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718843564410155e-05,
+      "loss": 2.7704,
+      "step": 27265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71874103689156e-05,
+      "loss": 2.5905,
+      "step": 27270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718638491223345e-05,
+      "loss": 2.5717,
+      "step": 27275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718535927405906e-05,
+      "loss": 2.5388,
+      "step": 27280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718433345439635e-05,
+      "loss": 2.5892,
+      "step": 27285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718330745324926e-05,
+      "loss": 2.6616,
+      "step": 27290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718228127062178e-05,
+      "loss": 2.5378,
+      "step": 27295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718125490651781e-05,
+      "loss": 2.5443,
+      "step": 27300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718022836094134e-05,
+      "loss": 2.6272,
+      "step": 27305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717920163389626e-05,
+      "loss": 2.5719,
+      "step": 27310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71781747253866e-05,
+      "loss": 2.6186,
+      "step": 27315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717714763541623e-05,
+      "loss": 2.5129,
+      "step": 27320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717612036398915e-05,
+      "loss": 2.4917,
+      "step": 27325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717509291110928e-05,
+      "loss": 2.5005,
+      "step": 27330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717406527678062e-05,
+      "loss": 2.4631,
+      "step": 27335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717303746100706e-05,
+      "loss": 2.6751,
+      "step": 27340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717200946379261e-05,
+      "loss": 2.4304,
+      "step": 27345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717098128514118e-05,
+      "loss": 2.5604,
+      "step": 27350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716995292505674e-05,
+      "loss": 2.5111,
+      "step": 27355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716892438354326e-05,
+      "loss": 2.6201,
+      "step": 27360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716789566060467e-05,
+      "loss": 2.4215,
+      "step": 27365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716686675624495e-05,
+      "loss": 2.6013,
+      "step": 27370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716583767046805e-05,
+      "loss": 2.5008,
+      "step": 27375
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716480840327793e-05,
+      "loss": 2.566,
+      "step": 27380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716377895467854e-05,
+      "loss": 2.498,
+      "step": 27385
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716274932467385e-05,
+      "loss": 2.5568,
+      "step": 27390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71617195132678e-05,
+      "loss": 2.6226,
+      "step": 27395
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716068952046438e-05,
+      "loss": 2.5837,
+      "step": 27400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715965934626754e-05,
+      "loss": 2.5452,
+      "step": 27405
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715862899068122e-05,
+      "loss": 2.3166,
+      "step": 27410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715759845370943e-05,
+      "loss": 2.6082,
+      "step": 27415
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715656773535609e-05,
+      "loss": 2.5512,
+      "step": 27420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715553683562519e-05,
+      "loss": 2.5446,
+      "step": 27425
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715450575452067e-05,
+      "loss": 2.6337,
+      "step": 27430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715347449204653e-05,
+      "loss": 2.3938,
+      "step": 27435
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715244304820672e-05,
+      "loss": 2.3777,
+      "step": 27440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715141142300521e-05,
+      "loss": 2.5434,
+      "step": 27445
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715037961644595e-05,
+      "loss": 2.5657,
+      "step": 27450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714934762853292e-05,
+      "loss": 2.707,
+      "step": 27455
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71483154592701e-05,
+      "loss": 2.5699,
+      "step": 27460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714728310866144e-05,
+      "loss": 2.6632,
+      "step": 27465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714625057671095e-05,
+      "loss": 2.5392,
+      "step": 27470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714521786342254e-05,
+      "loss": 2.4679,
+      "step": 27475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714418496880025e-05,
+      "loss": 2.5594,
+      "step": 27480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714315189284799e-05,
+      "loss": 2.4923,
+      "step": 27485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714211863556978e-05,
+      "loss": 2.5448,
+      "step": 27490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714108519696957e-05,
+      "loss": 2.474,
+      "step": 27495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714005157705134e-05,
+      "loss": 2.5899,
+      "step": 27500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713901777581906e-05,
+      "loss": 2.5469,
+      "step": 27505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713798379327673e-05,
+      "loss": 2.6237,
+      "step": 27510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713694962942832e-05,
+      "loss": 2.6096,
+      "step": 27515
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713591528427777e-05,
+      "loss": 2.6137,
+      "step": 27520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71348807578291e-05,
+      "loss": 2.5446,
+      "step": 27525
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713384605008628e-05,
+      "loss": 2.5584,
+      "step": 27530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713281116105328e-05,
+      "loss": 2.6336,
+      "step": 27535
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713177609073407e-05,
+      "loss": 2.5989,
+      "step": 27540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713074083913268e-05,
+      "loss": 2.6589,
+      "step": 27545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712970540625303e-05,
+      "loss": 2.5454,
+      "step": 27550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712866979209915e-05,
+      "loss": 2.5316,
+      "step": 27555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712763399667499e-05,
+      "loss": 2.5516,
+      "step": 27560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712659801998456e-05,
+      "loss": 2.6614,
+      "step": 27565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712556186203184e-05,
+      "loss": 2.6624,
+      "step": 27570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712452552282079e-05,
+      "loss": 2.6597,
+      "step": 27575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712348900235541e-05,
+      "loss": 2.4983,
+      "step": 27580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712245230063973e-05,
+      "loss": 2.5131,
+      "step": 27585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712141541767765e-05,
+      "loss": 2.5293,
+      "step": 27590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712037835347324e-05,
+      "loss": 2.5403,
+      "step": 27595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711934110803044e-05,
+      "loss": 2.5573,
+      "step": 27600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711830368135327e-05,
+      "loss": 2.6897,
+      "step": 27605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71172660734457e-05,
+      "loss": 2.6824,
+      "step": 27610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711622828431173e-05,
+      "loss": 2.6011,
+      "step": 27615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711519031395534e-05,
+      "loss": 2.4837,
+      "step": 27620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711415216238053e-05,
+      "loss": 2.6706,
+      "step": 27625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711311382959129e-05,
+      "loss": 2.6548,
+      "step": 27630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711207531559163e-05,
+      "loss": 2.4635,
+      "step": 27635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711103662038553e-05,
+      "loss": 2.4832,
+      "step": 27640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710999774397699e-05,
+      "loss": 2.4803,
+      "step": 27645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710895868637e-05,
+      "loss": 2.5743,
+      "step": 27650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710791944756855e-05,
+      "loss": 2.6567,
+      "step": 27655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710688002757665e-05,
+      "loss": 2.5446,
+      "step": 27660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71058404263983e-05,
+      "loss": 2.4206,
+      "step": 27665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710480064403749e-05,
+      "loss": 2.4819,
+      "step": 27670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710376068049823e-05,
+      "loss": 2.4657,
+      "step": 27675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71027205357845e-05,
+      "loss": 2.4212,
+      "step": 27680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710168020990032e-05,
+      "loss": 2.53,
+      "step": 27685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710063970284968e-05,
+      "loss": 2.6663,
+      "step": 27690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709959901463659e-05,
+      "loss": 2.595,
+      "step": 27695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709855814526504e-05,
+      "loss": 2.5715,
+      "step": 27700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709751709473905e-05,
+      "loss": 2.4717,
+      "step": 27705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709647586306263e-05,
+      "loss": 2.5669,
+      "step": 27710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709543445023974e-05,
+      "loss": 2.4924,
+      "step": 27715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709439285627444e-05,
+      "loss": 2.5454,
+      "step": 27720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709335108117071e-05,
+      "loss": 2.4285,
+      "step": 27725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709230912493257e-05,
+      "loss": 2.5413,
+      "step": 27730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709126698756401e-05,
+      "loss": 2.5918,
+      "step": 27735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709022466906904e-05,
+      "loss": 2.6849,
+      "step": 27740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708918216945169e-05,
+      "loss": 2.5886,
+      "step": 27745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708813948871594e-05,
+      "loss": 2.5954,
+      "step": 27750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708709662686581e-05,
+      "loss": 2.5761,
+      "step": 27755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708605358390535e-05,
+      "loss": 2.7651,
+      "step": 27760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70850103598385e-05,
+      "loss": 2.7417,
+      "step": 27765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708396695466933e-05,
+      "loss": 2.4761,
+      "step": 27770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708292336840183e-05,
+      "loss": 2.3805,
+      "step": 27775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708187960104002e-05,
+      "loss": 2.5187,
+      "step": 27780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708083565258792e-05,
+      "loss": 2.5536,
+      "step": 27785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707979152304953e-05,
+      "loss": 2.4969,
+      "step": 27790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707874721242889e-05,
+      "loss": 2.5799,
+      "step": 27795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707770272072999e-05,
+      "loss": 2.5818,
+      "step": 27800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707665804795685e-05,
+      "loss": 2.5215,
+      "step": 27805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707561319411352e-05,
+      "loss": 2.4981,
+      "step": 27810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707456815920397e-05,
+      "loss": 2.6562,
+      "step": 27815
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707352294323227e-05,
+      "loss": 2.5675,
+      "step": 27820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707247754620239e-05,
+      "loss": 2.5424,
+      "step": 27825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707143196811839e-05,
+      "loss": 2.5807,
+      "step": 27830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707038620898427e-05,
+      "loss": 2.4855,
+      "step": 27835
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706934026880406e-05,
+      "loss": 2.6906,
+      "step": 27840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70682941475818e-05,
+      "loss": 2.459,
+      "step": 27845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706724784532148e-05,
+      "loss": 2.568,
+      "step": 27850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706620136202715e-05,
+      "loss": 2.5223,
+      "step": 27855
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706515469770282e-05,
+      "loss": 2.4719,
+      "step": 27860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706410785235251e-05,
+      "loss": 2.4949,
+      "step": 27865
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706306082598028e-05,
+      "loss": 2.5985,
+      "step": 27870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706201361859013e-05,
+      "loss": 2.6554,
+      "step": 27875
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70609662301861e-05,
+      "loss": 2.4876,
+      "step": 27880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70599186607722e-05,
+      "loss": 2.6392,
+      "step": 27885
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705887091035247e-05,
+      "loss": 2.6554,
+      "step": 27890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705782297893094e-05,
+      "loss": 2.5558,
+      "step": 27895
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705677486651164e-05,
+      "loss": 2.6366,
+      "step": 27900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705572657309861e-05,
+      "loss": 2.4822,
+      "step": 27905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705467809869588e-05,
+      "loss": 2.5245,
+      "step": 27910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705362944330747e-05,
+      "loss": 2.6077,
+      "step": 27915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705258060693742e-05,
+      "loss": 2.5903,
+      "step": 27920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705153158958977e-05,
+      "loss": 2.3601,
+      "step": 27925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705048239126855e-05,
+      "loss": 2.5422,
+      "step": 27930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70494330119778e-05,
+      "loss": 2.5782,
+      "step": 27935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.704838345172154e-05,
+      "loss": 2.5522,
+      "step": 27940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.704733371050383e-05,
+      "loss": 2.5931,
+      "step": 27945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70462837883287e-05,
+      "loss": 2.6078,
+      "step": 27950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.704523368520018e-05,
+      "loss": 2.7386,
+      "step": 27955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70441834011223e-05,
+      "loss": 2.4374,
+      "step": 27960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.704313293609913e-05,
+      "loss": 2.4768,
+      "step": 27965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.704208229013468e-05,
+      "loss": 2.3498,
+      "step": 27970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.704103146323303e-05,
+      "loss": 2.5571,
+      "step": 27975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.703998045539817e-05,
+      "loss": 2.7033,
+      "step": 27980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.703892926663418e-05,
+      "loss": 2.6189,
+      "step": 27985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70378778969451e-05,
+      "loss": 2.5381,
+      "step": 27990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.703682634633496e-05,
+      "loss": 2.6705,
+      "step": 27995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70357746148078e-05,
+      "loss": 2.6164,
+      "step": 28000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.703472270236768e-05,
+      "loss": 2.4933,
+      "step": 28005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.703367060901865e-05,
+      "loss": 2.5924,
+      "step": 28010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.703261833476474e-05,
+      "loss": 2.593,
+      "step": 28015
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.703156587961002e-05,
+      "loss": 2.5085,
+      "step": 28020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70305132435585e-05,
+      "loss": 2.6028,
+      "step": 28025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702946042661427e-05,
+      "loss": 2.5673,
+      "step": 28030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702840742878137e-05,
+      "loss": 2.491,
+      "step": 28035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702735425006383e-05,
+      "loss": 2.5583,
+      "step": 28040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702630089046572e-05,
+      "loss": 2.6267,
+      "step": 28045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702524734999108e-05,
+      "loss": 2.4207,
+      "step": 28050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702419362864397e-05,
+      "loss": 2.5949,
+      "step": 28055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702313972642842e-05,
+      "loss": 2.6342,
+      "step": 28060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702208564334853e-05,
+      "loss": 2.6437,
+      "step": 28065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702103137940831e-05,
+      "loss": 2.535,
+      "step": 28070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701997693461185e-05,
+      "loss": 2.446,
+      "step": 28075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701892230896317e-05,
+      "loss": 2.5435,
+      "step": 28080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701786750246636e-05,
+      "loss": 2.4587,
+      "step": 28085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701681251512546e-05,
+      "loss": 2.5963,
+      "step": 28090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701575734694452e-05,
+      "loss": 2.4797,
+      "step": 28095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701470199792762e-05,
+      "loss": 2.6325,
+      "step": 28100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70136464680788e-05,
+      "loss": 2.5533,
+      "step": 28105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701259075740212e-05,
+      "loss": 2.6801,
+      "step": 28110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701153486590166e-05,
+      "loss": 2.5549,
+      "step": 28115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701047879358146e-05,
+      "loss": 2.4747,
+      "step": 28120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70094225404456e-05,
+      "loss": 2.6237,
+      "step": 28125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700836610649812e-05,
+      "loss": 2.6456,
+      "step": 28130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700730949174312e-05,
+      "loss": 2.5645,
+      "step": 28135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700625269618461e-05,
+      "loss": 2.5182,
+      "step": 28140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700519571982671e-05,
+      "loss": 2.6716,
+      "step": 28145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700413856267346e-05,
+      "loss": 2.5421,
+      "step": 28150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700308122472892e-05,
+      "loss": 2.5272,
+      "step": 28155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700202370599716e-05,
+      "loss": 2.6617,
+      "step": 28160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700096600648226e-05,
+      "loss": 2.4392,
+      "step": 28165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699990812618827e-05,
+      "loss": 2.5022,
+      "step": 28170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699885006511928e-05,
+      "loss": 2.5212,
+      "step": 28175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699779182327936e-05,
+      "loss": 2.6325,
+      "step": 28180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699673340067254e-05,
+      "loss": 2.4275,
+      "step": 28185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699567479730294e-05,
+      "loss": 2.5104,
+      "step": 28190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699461601317459e-05,
+      "loss": 2.6159,
+      "step": 28195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69935570482916e-05,
+      "loss": 2.3828,
+      "step": 28200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699249790265804e-05,
+      "loss": 2.3771,
+      "step": 28205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699143857627795e-05,
+      "loss": 2.4873,
+      "step": 28210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699037906915542e-05,
+      "loss": 2.5011,
+      "step": 28215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.698931938129454e-05,
+      "loss": 2.6119,
+      "step": 28220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.698825951269939e-05,
+      "loss": 2.3541,
+      "step": 28225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.6987199463374e-05,
+      "loss": 2.5495,
+      "step": 28230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69861392333225e-05,
+      "loss": 2.5042,
+      "step": 28235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.698507882254895e-05,
+      "loss": 2.6664,
+      "step": 28240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.698401823105743e-05,
+      "loss": 2.6134,
+      "step": 28245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.6982957458852e-05,
+      "loss": 2.6797,
+      "step": 28250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.698189650593676e-05,
+      "loss": 2.6521,
+      "step": 28255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.698083537231579e-05,
+      "loss": 2.5689,
+      "step": 28260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697977405799318e-05,
+      "loss": 2.4208,
+      "step": 28265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697871256297299e-05,
+      "loss": 2.5874,
+      "step": 28270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697765088725931e-05,
+      "loss": 2.4603,
+      "step": 28275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697658903085622e-05,
+      "loss": 2.5582,
+      "step": 28280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697552699376782e-05,
+      "loss": 2.6524,
+      "step": 28285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697446477599819e-05,
+      "loss": 2.5144,
+      "step": 28290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697340237755141e-05,
+      "loss": 2.5951,
+      "step": 28295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697233979843157e-05,
+      "loss": 2.7007,
+      "step": 28300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697127703864274e-05,
+      "loss": 2.4882,
+      "step": 28305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697021409818902e-05,
+      "loss": 2.5304,
+      "step": 28310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696915097707452e-05,
+      "loss": 2.613,
+      "step": 28315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69680876753033e-05,
+      "loss": 2.4711,
+      "step": 28320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696702419287945e-05,
+      "loss": 2.5379,
+      "step": 28325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696596052980709e-05,
+      "loss": 2.6808,
+      "step": 28330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696489668609028e-05,
+      "loss": 2.5064,
+      "step": 28335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696383266173312e-05,
+      "loss": 2.6791,
+      "step": 28340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69627684567397e-05,
+      "loss": 2.3921,
+      "step": 28345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696170407111412e-05,
+      "loss": 2.5988,
+      "step": 28350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696063950486048e-05,
+      "loss": 2.5868,
+      "step": 28355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.695957475798287e-05,
+      "loss": 2.4523,
+      "step": 28360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.695850983048538e-05,
+      "loss": 2.4448,
+      "step": 28365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69574447223721e-05,
+      "loss": 2.49,
+      "step": 28370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.695637943364715e-05,
+      "loss": 2.4992,
+      "step": 28375
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69553139643146e-05,
+      "loss": 2.636,
+      "step": 28380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.695424831437856e-05,
+      "loss": 2.4752,
+      "step": 28385
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.695318248384313e-05,
+      "loss": 2.5209,
+      "step": 28390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.695211647271242e-05,
+      "loss": 2.5371,
+      "step": 28395
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.695105028099051e-05,
+      "loss": 2.6156,
+      "step": 28400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694998390868151e-05,
+      "loss": 2.5485,
+      "step": 28405
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694891735578952e-05,
+      "loss": 2.5513,
+      "step": 28410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694785062231865e-05,
+      "loss": 2.4939,
+      "step": 28415
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694678370827299e-05,
+      "loss": 2.5466,
+      "step": 28420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694571661365665e-05,
+      "loss": 2.4893,
+      "step": 28425
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694464933847374e-05,
+      "loss": 2.3631,
+      "step": 28430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694358188272836e-05,
+      "loss": 2.6135,
+      "step": 28435
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694251424642461e-05,
+      "loss": 2.5557,
+      "step": 28440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69414464295666e-05,
+      "loss": 2.5237,
+      "step": 28445
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694037843215846e-05,
+      "loss": 2.4958,
+      "step": 28450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693931025420425e-05,
+      "loss": 2.5189,
+      "step": 28455
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693824189570812e-05,
+      "loss": 2.6569,
+      "step": 28460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693717335667416e-05,
+      "loss": 2.5081,
+      "step": 28465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693610463710648e-05,
+      "loss": 2.5536,
+      "step": 28470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693503573700921e-05,
+      "loss": 2.6322,
+      "step": 28475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693396665638643e-05,
+      "loss": 2.3695,
+      "step": 28480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693289739524228e-05,
+      "loss": 2.4769,
+      "step": 28485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693182795358086e-05,
+      "loss": 2.6581,
+      "step": 28490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693075833140628e-05,
+      "loss": 2.5724,
+      "step": 28495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692968852872266e-05,
+      "loss": 2.5792,
+      "step": 28500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69286185455341e-05,
+      "loss": 2.4188,
+      "step": 28505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692754838184475e-05,
+      "loss": 2.6969,
+      "step": 28510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692647803765869e-05,
+      "loss": 2.4759,
+      "step": 28515
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692540751298006e-05,
+      "loss": 2.5534,
+      "step": 28520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692433680781297e-05,
+      "loss": 2.5505,
+      "step": 28525
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692326592216155e-05,
+      "loss": 2.5613,
+      "step": 28530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692219485602988e-05,
+      "loss": 2.5003,
+      "step": 28535
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692112360942211e-05,
+      "loss": 2.6163,
+      "step": 28540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692005218234236e-05,
+      "loss": 2.5159,
+      "step": 28545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691898057479477e-05,
+      "loss": 2.4881,
+      "step": 28550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691790878678342e-05,
+      "loss": 2.5473,
+      "step": 28555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691683681831246e-05,
+      "loss": 2.7125,
+      "step": 28560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691576466938597e-05,
+      "loss": 2.428,
+      "step": 28565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691469234000815e-05,
+      "loss": 2.5548,
+      "step": 28570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691361983018305e-05,
+      "loss": 2.3861,
+      "step": 28575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691254713991486e-05,
+      "loss": 2.4041,
+      "step": 28580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691147426920765e-05,
+      "loss": 2.5614,
+      "step": 28585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691040121806559e-05,
+      "loss": 2.703,
+      "step": 28590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690932798649277e-05,
+      "loss": 2.5885,
+      "step": 28595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690825457449335e-05,
+      "loss": 2.4556,
+      "step": 28600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690718098207144e-05,
+      "loss": 2.4777,
+      "step": 28605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690610720923117e-05,
+      "loss": 2.5793,
+      "step": 28610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690503325597669e-05,
+      "loss": 2.5807,
+      "step": 28615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69039591223121e-05,
+      "loss": 2.524,
+      "step": 28620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690288480824153e-05,
+      "loss": 2.4774,
+      "step": 28625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690181031376915e-05,
+      "loss": 2.4824,
+      "step": 28630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690073563889906e-05,
+      "loss": 2.6634,
+      "step": 28635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689966078363542e-05,
+      "loss": 2.6984,
+      "step": 28640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689858574798232e-05,
+      "loss": 2.3485,
+      "step": 28645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689751053194395e-05,
+      "loss": 2.6124,
+      "step": 28650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689643513552442e-05,
+      "loss": 2.5793,
+      "step": 28655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689535955872784e-05,
+      "loss": 2.5009,
+      "step": 28660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68942838015584e-05,
+      "loss": 2.553,
+      "step": 28665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689320786402018e-05,
+      "loss": 2.5435,
+      "step": 28670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689213174611737e-05,
+      "loss": 2.6363,
+      "step": 28675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689105544785409e-05,
+      "loss": 2.5951,
+      "step": 28680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688997896923447e-05,
+      "loss": 2.3088,
+      "step": 28685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688890231026267e-05,
+      "loss": 2.6746,
+      "step": 28690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688782547094279e-05,
+      "loss": 2.4802,
+      "step": 28695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688674845127903e-05,
+      "loss": 2.5468,
+      "step": 28700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688567125127548e-05,
+      "loss": 2.4512,
+      "step": 28705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688459387093632e-05,
+      "loss": 2.5504,
+      "step": 28710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688351631026567e-05,
+      "loss": 2.4132,
+      "step": 28715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688243856926769e-05,
+      "loss": 2.5486,
+      "step": 28720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688136064794653e-05,
+      "loss": 2.5663,
+      "step": 28725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688028254630632e-05,
+      "loss": 2.5832,
+      "step": 28730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687920426435121e-05,
+      "loss": 2.4338,
+      "step": 28735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687812580208535e-05,
+      "loss": 2.4504,
+      "step": 28740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68770471595129e-05,
+      "loss": 2.5459,
+      "step": 28745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687596833663796e-05,
+      "loss": 2.59,
+      "step": 28750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687488933346476e-05,
+      "loss": 2.5265,
+      "step": 28755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687381014999737e-05,
+      "loss": 2.565,
+      "step": 28760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687273078624001e-05,
+      "loss": 2.4623,
+      "step": 28765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687165124219678e-05,
+      "loss": 2.4865,
+      "step": 28770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687057151787187e-05,
+      "loss": 2.6003,
+      "step": 28775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686949161326939e-05,
+      "loss": 2.5959,
+      "step": 28780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686841152839352e-05,
+      "loss": 2.4919,
+      "step": 28785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686733126324843e-05,
+      "loss": 2.5087,
+      "step": 28790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686625081783825e-05,
+      "loss": 2.5826,
+      "step": 28795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686517019216714e-05,
+      "loss": 2.5154,
+      "step": 28800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686408938623925e-05,
+      "loss": 2.7014,
+      "step": 28805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686300840005875e-05,
+      "loss": 2.5091,
+      "step": 28810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686192723362981e-05,
+      "loss": 2.6724,
+      "step": 28815
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686084588695657e-05,
+      "loss": 2.3988,
+      "step": 28820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685976436004317e-05,
+      "loss": 2.59,
+      "step": 28825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685868265289381e-05,
+      "loss": 2.4987,
+      "step": 28830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685760076551263e-05,
+      "loss": 2.6223,
+      "step": 28835
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685651869790378e-05,
+      "loss": 2.6489,
+      "step": 28840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685543645007144e-05,
+      "loss": 2.5034,
+      "step": 28845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68543540220198e-05,
+      "loss": 2.4956,
+      "step": 28850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685327141375295e-05,
+      "loss": 2.7308,
+      "step": 28855
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685218862527511e-05,
+      "loss": 2.4408,
+      "step": 28860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685110565659042e-05,
+      "loss": 2.5503,
+      "step": 28865
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685002250770308e-05,
+      "loss": 2.5884,
+      "step": 28870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684893917861722e-05,
+      "loss": 2.6814,
+      "step": 28875
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684785566933701e-05,
+      "loss": 2.5693,
+      "step": 28880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684677197986663e-05,
+      "loss": 2.6142,
+      "step": 28885
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684568811021024e-05,
+      "loss": 2.5288,
+      "step": 28890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684460406037202e-05,
+      "loss": 2.567,
+      "step": 28895
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684351983035614e-05,
+      "loss": 2.4838,
+      "step": 28900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684243542016674e-05,
+      "loss": 2.4779,
+      "step": 28905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684135082980802e-05,
+      "loss": 2.4555,
+      "step": 28910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684026605928416e-05,
+      "loss": 2.4815,
+      "step": 28915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.683918110859931e-05,
+      "loss": 2.4054,
+      "step": 28920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.683809597775764e-05,
+      "loss": 2.7233,
+      "step": 28925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.683701066676334e-05,
+      "loss": 2.5278,
+      "step": 28930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68359251756206e-05,
+      "loss": 2.5411,
+      "step": 28935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.683483950433354e-05,
+      "loss": 2.584,
+      "step": 28940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.683375365290638e-05,
+      "loss": 2.7222,
+      "step": 28945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68326676213433e-05,
+      "loss": 2.4927,
+      "step": 28950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.683158140964845e-05,
+      "loss": 2.6297,
+      "step": 28955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.683049501782603e-05,
+      "loss": 2.6805,
+      "step": 28960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68294084458802e-05,
+      "loss": 2.5565,
+      "step": 28965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.682832169381516e-05,
+      "loss": 2.5946,
+      "step": 28970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.682723476163507e-05,
+      "loss": 2.5432,
+      "step": 28975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.682614764934412e-05,
+      "loss": 2.4696,
+      "step": 28980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68250603569465e-05,
+      "loss": 2.657,
+      "step": 28985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.682397288444638e-05,
+      "loss": 2.4999,
+      "step": 28990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.682288523184795e-05,
+      "loss": 2.7348,
+      "step": 28995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.682179739915537e-05,
+      "loss": 2.5409,
+      "step": 29000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.682070938637287e-05,
+      "loss": 2.6483,
+      "step": 29005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.681962119350459e-05,
+      "loss": 2.5241,
+      "step": 29010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.681853282055473e-05,
+      "loss": 2.4072,
+      "step": 29015
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.681744426752748e-05,
+      "loss": 2.5082,
+      "step": 29020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.681635553442704e-05,
+      "loss": 2.3296,
+      "step": 29025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.681526662125757e-05,
+      "loss": 2.5557,
+      "step": 29030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.681417752802329e-05,
+      "loss": 2.449,
+      "step": 29035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.681308825472835e-05,
+      "loss": 2.4485,
+      "step": 29040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.681199880137698e-05,
+      "loss": 2.4795,
+      "step": 29045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.681090916797334e-05,
+      "loss": 2.6037,
+      "step": 29050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.680981935452165e-05,
+      "loss": 2.6304,
+      "step": 29055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.680872936102607e-05,
+      "loss": 2.6861,
+      "step": 29060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68076391874908e-05,
+      "loss": 2.6031,
+      "step": 29065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.680654883392004e-05,
+      "loss": 2.5662,
+      "step": 29070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.6805458300318e-05,
+      "loss": 2.6184,
+      "step": 29075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.680436758668885e-05,
+      "loss": 2.3567,
+      "step": 29080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68032766930368e-05,
+      "loss": 2.5337,
+      "step": 29085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.680218561936601e-05,
+      "loss": 2.3669,
+      "step": 29090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.680109436568074e-05,
+      "loss": 2.6362,
+      "step": 29095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.680000293198513e-05,
+      "loss": 2.5609,
+      "step": 29100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.679891131828341e-05,
+      "loss": 2.4845,
+      "step": 29105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.679781952457978e-05,
+      "loss": 2.6976,
+      "step": 29110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.67967275508784e-05,
+      "loss": 2.5232,
+      "step": 29115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.679563539718353e-05,
+      "loss": 2.6067,
+      "step": 29120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.679454306349933e-05,
+      "loss": 2.5659,
+      "step": 29125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.679345054983001e-05,
+      "loss": 2.3724,
+      "step": 29130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.679235785617977e-05,
+      "loss": 2.4945,
+      "step": 29135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.679126498255282e-05,
+      "loss": 2.5606,
+      "step": 29140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.679017192895336e-05,
+      "loss": 2.5823,
+      "step": 29145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678907869538559e-05,
+      "loss": 2.6668,
+      "step": 29150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678798528185372e-05,
+      "loss": 2.5594,
+      "step": 29155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678689168836196e-05,
+      "loss": 2.5738,
+      "step": 29160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67857979149145e-05,
+      "loss": 2.5451,
+      "step": 29165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678470396151557e-05,
+      "loss": 2.5521,
+      "step": 29170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678360982816935e-05,
+      "loss": 2.5669,
+      "step": 29175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678251551488006e-05,
+      "loss": 2.7129,
+      "step": 29180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678142102165194e-05,
+      "loss": 2.455,
+      "step": 29185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678032634848914e-05,
+      "loss": 2.5161,
+      "step": 29190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677923149539593e-05,
+      "loss": 2.4847,
+      "step": 29195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677813646237649e-05,
+      "loss": 2.6087,
+      "step": 29200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677704124943503e-05,
+      "loss": 2.5421,
+      "step": 29205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677594585657577e-05,
+      "loss": 2.5204,
+      "step": 29210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677485028380291e-05,
+      "loss": 2.6027,
+      "step": 29215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677375453112069e-05,
+      "loss": 2.5519,
+      "step": 29220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677265859853331e-05,
+      "loss": 2.5749,
+      "step": 29225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677156248604496e-05,
+      "loss": 2.47,
+      "step": 29230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67704661936599e-05,
+      "loss": 2.6524,
+      "step": 29235
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676936972138233e-05,
+      "loss": 2.6072,
+      "step": 29240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676827306921647e-05,
+      "loss": 2.5607,
+      "step": 29245
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67671762371665e-05,
+      "loss": 2.6123,
+      "step": 29250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676607922523671e-05,
+      "loss": 2.5793,
+      "step": 29255
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676498203343127e-05,
+      "loss": 2.6266,
+      "step": 29260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676388466175441e-05,
+      "loss": 2.4182,
+      "step": 29265
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676278711021035e-05,
+      "loss": 2.5089,
+      "step": 29270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676168937880333e-05,
+      "loss": 2.618,
+      "step": 29275
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676059146753753e-05,
+      "loss": 2.592,
+      "step": 29280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675949337641721e-05,
+      "loss": 2.5814,
+      "step": 29285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675839510544658e-05,
+      "loss": 2.6283,
+      "step": 29290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675729665462987e-05,
+      "loss": 2.4622,
+      "step": 29295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675619802397129e-05,
+      "loss": 2.5439,
+      "step": 29300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67550992134751e-05,
+      "loss": 2.6079,
+      "step": 29305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675400022314549e-05,
+      "loss": 2.618,
+      "step": 29310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675290105298668e-05,
+      "loss": 2.4624,
+      "step": 29315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675180170300296e-05,
+      "loss": 2.401,
+      "step": 29320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675070217319849e-05,
+      "loss": 2.5662,
+      "step": 29325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.674960246357753e-05,
+      "loss": 2.5318,
+      "step": 29330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67485025741443e-05,
+      "loss": 2.3792,
+      "step": 29335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.674740250490304e-05,
+      "loss": 2.5693,
+      "step": 29340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6746302255858e-05,
+      "loss": 2.4584,
+      "step": 29345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.674520182701335e-05,
+      "loss": 2.6055,
+      "step": 29350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.674410121837338e-05,
+      "loss": 2.4587,
+      "step": 29355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.674300042994231e-05,
+      "loss": 2.6592,
+      "step": 29360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.674189946172435e-05,
+      "loss": 2.4481,
+      "step": 29365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.674079831372377e-05,
+      "loss": 2.5255,
+      "step": 29370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67396969859448e-05,
+      "loss": 2.4014,
+      "step": 29375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673859547839165e-05,
+      "loss": 2.5457,
+      "step": 29380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673749379106857e-05,
+      "loss": 2.5743,
+      "step": 29385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67363919239798e-05,
+      "loss": 2.4526,
+      "step": 29390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673528987712958e-05,
+      "loss": 2.4358,
+      "step": 29395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673418765052214e-05,
+      "loss": 2.5978,
+      "step": 29400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673308524416174e-05,
+      "loss": 2.4338,
+      "step": 29405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673198265805259e-05,
+      "loss": 2.5789,
+      "step": 29410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673087989219895e-05,
+      "loss": 2.5106,
+      "step": 29415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672977694660505e-05,
+      "loss": 2.5799,
+      "step": 29420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672867382127516e-05,
+      "loss": 2.5806,
+      "step": 29425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672757051621348e-05,
+      "loss": 2.5911,
+      "step": 29430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67264670314243e-05,
+      "loss": 2.4844,
+      "step": 29435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672536336691183e-05,
+      "loss": 2.4605,
+      "step": 29440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672425952268031e-05,
+      "loss": 2.5592,
+      "step": 29445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672315549873404e-05,
+      "loss": 2.4789,
+      "step": 29450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672205129507718e-05,
+      "loss": 2.5231,
+      "step": 29455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672094691171406e-05,
+      "loss": 2.6784,
+      "step": 29460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.671984234864887e-05,
+      "loss": 2.4308,
+      "step": 29465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.671873760588591e-05,
+      "loss": 2.5331,
+      "step": 29470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67176326834294e-05,
+      "loss": 2.5634,
+      "step": 29475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.671652758128355e-05,
+      "loss": 2.592,
+      "step": 29480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67154222994527e-05,
+      "loss": 2.5857,
+      "step": 29485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.671431683794101e-05,
+      "loss": 2.5323,
+      "step": 29490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67132111967528e-05,
+      "loss": 2.6059,
+      "step": 29495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.671210537589228e-05,
+      "loss": 2.5236,
+      "step": 29500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.671099937536374e-05,
+      "loss": 2.5882,
+      "step": 29505
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67098931951714e-05,
+      "loss": 2.5015,
+      "step": 29510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.670878683531952e-05,
+      "loss": 2.6074,
+      "step": 29515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.670768029581238e-05,
+      "loss": 2.4648,
+      "step": 29520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.670657357665422e-05,
+      "loss": 2.5934,
+      "step": 29525
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67054666778493e-05,
+      "loss": 2.4489,
+      "step": 29530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.670435959940187e-05,
+      "loss": 2.4758,
+      "step": 29535
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.670325234131619e-05,
+      "loss": 2.6017,
+      "step": 29540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67021449035965e-05,
+      "loss": 2.5435,
+      "step": 29545
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.670103728624711e-05,
+      "loss": 2.5287,
+      "step": 29550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669992948927224e-05,
+      "loss": 2.7144,
+      "step": 29555
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669882151267618e-05,
+      "loss": 2.6015,
+      "step": 29560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669771335646315e-05,
+      "loss": 2.4577,
+      "step": 29565
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669660502063744e-05,
+      "loss": 2.5639,
+      "step": 29570
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66954965052033e-05,
+      "loss": 2.5247,
+      "step": 29575
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669438781016502e-05,
+      "loss": 2.6654,
+      "step": 29580
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669327893552685e-05,
+      "loss": 2.7039,
+      "step": 29585
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669216988129304e-05,
+      "loss": 2.6378,
+      "step": 29590
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669106064746787e-05,
+      "loss": 2.5541,
+      "step": 29595
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66899512340556e-05,
+      "loss": 2.6024,
+      "step": 29600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.668884164106051e-05,
+      "loss": 2.4948,
+      "step": 29605
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.668773186848685e-05,
+      "loss": 2.5845,
+      "step": 29610
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66866219163389e-05,
+      "loss": 2.4889,
+      "step": 29615
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.668551178462094e-05,
+      "loss": 2.6217,
+      "step": 29620
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.668440147333721e-05,
+      "loss": 2.5002,
+      "step": 29625
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6683290982492e-05,
+      "loss": 2.4149,
+      "step": 29630
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.668218031208958e-05,
+      "loss": 2.536,
+      "step": 29635
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.668106946213422e-05,
+      "loss": 2.5723,
+      "step": 29640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66799584326302e-05,
+      "loss": 2.3448,
+      "step": 29645
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667884722358178e-05,
+      "loss": 2.535,
+      "step": 29650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667773583499325e-05,
+      "loss": 2.5016,
+      "step": 29655
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667662426686885e-05,
+      "loss": 2.4795,
+      "step": 29660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667551251921289e-05,
+      "loss": 2.5356,
+      "step": 29665
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667440059202966e-05,
+      "loss": 2.5754,
+      "step": 29670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667328848532339e-05,
+      "loss": 2.5769,
+      "step": 29675
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667217619909839e-05,
+      "loss": 2.5764,
+      "step": 29680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667106373335892e-05,
+      "loss": 2.647,
+      "step": 29685
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666995108810926e-05,
+      "loss": 2.5139,
+      "step": 29690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666883826335371e-05,
+      "loss": 2.5586,
+      "step": 29695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666772525909655e-05,
+      "loss": 2.7269,
+      "step": 29700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666661207534203e-05,
+      "loss": 2.6335,
+      "step": 29705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666549871209446e-05,
+      "loss": 2.5918,
+      "step": 29710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666438516935811e-05,
+      "loss": 2.4308,
+      "step": 29715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666327144713726e-05,
+      "loss": 2.5575,
+      "step": 29720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66621575454362e-05,
+      "loss": 2.5975,
+      "step": 29725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666104346425921e-05,
+      "loss": 2.5388,
+      "step": 29730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66599292036106e-05,
+      "loss": 2.5585,
+      "step": 29735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.665881476349461e-05,
+      "loss": 2.5725,
+      "step": 29740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.665770014391557e-05,
+      "loss": 2.6391,
+      "step": 29745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.665658534487774e-05,
+      "loss": 2.5175,
+      "step": 29750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66554703663854e-05,
+      "loss": 2.515,
+      "step": 29755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.665435520844289e-05,
+      "loss": 2.6382,
+      "step": 29760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.665323987105444e-05,
+      "loss": 2.6566,
+      "step": 29765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.665212435422436e-05,
+      "loss": 2.5816,
+      "step": 29770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.665100865795695e-05,
+      "loss": 2.613,
+      "step": 29775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664989278225649e-05,
+      "loss": 2.5321,
+      "step": 29780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664877672712729e-05,
+      "loss": 2.4519,
+      "step": 29785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664766049257361e-05,
+      "loss": 2.5407,
+      "step": 29790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664654407859978e-05,
+      "loss": 2.5165,
+      "step": 29795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664542748521007e-05,
+      "loss": 2.5655,
+      "step": 29800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664431071240878e-05,
+      "loss": 2.5562,
+      "step": 29805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664319376020022e-05,
+      "loss": 2.485,
+      "step": 29810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664207662858866e-05,
+      "loss": 2.6848,
+      "step": 29815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664095931757841e-05,
+      "loss": 2.4748,
+      "step": 29820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663984182717379e-05,
+      "loss": 2.5814,
+      "step": 29825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663872415737906e-05,
+      "loss": 2.4376,
+      "step": 29830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663760630819854e-05,
+      "loss": 2.479,
+      "step": 29835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663648827963652e-05,
+      "loss": 2.6479,
+      "step": 29840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66353700716973e-05,
+      "loss": 2.5035,
+      "step": 29845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663425168438519e-05,
+      "loss": 2.6232,
+      "step": 29850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663313311770449e-05,
+      "loss": 2.6753,
+      "step": 29855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66320143716595e-05,
+      "loss": 2.5726,
+      "step": 29860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663089544625453e-05,
+      "loss": 2.6273,
+      "step": 29865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662977634149384e-05,
+      "loss": 2.4634,
+      "step": 29870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662865705738181e-05,
+      "loss": 2.4998,
+      "step": 29875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662753759392269e-05,
+      "loss": 2.4769,
+      "step": 29880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662641795112081e-05,
+      "loss": 2.5721,
+      "step": 29885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662529812898046e-05,
+      "loss": 2.5023,
+      "step": 29890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662417812750597e-05,
+      "loss": 2.4792,
+      "step": 29895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662305794670161e-05,
+      "loss": 2.7016,
+      "step": 29900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662193758657172e-05,
+      "loss": 2.5137,
+      "step": 29905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66208170471206e-05,
+      "loss": 2.6062,
+      "step": 29910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661969632835257e-05,
+      "loss": 2.4445,
+      "step": 29915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661857543027193e-05,
+      "loss": 2.5443,
+      "step": 29920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661745435288298e-05,
+      "loss": 2.5529,
+      "step": 29925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661633309619007e-05,
+      "loss": 2.601,
+      "step": 29930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661521166019748e-05,
+      "loss": 2.6069,
+      "step": 29935
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661409004490951e-05,
+      "loss": 2.5238,
+      "step": 29940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.66129682503305e-05,
+      "loss": 2.5492,
+      "step": 29945
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661184627646479e-05,
+      "loss": 2.5825,
+      "step": 29950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661072412331662e-05,
+      "loss": 2.6473,
+      "step": 29955
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660960179089038e-05,
+      "loss": 2.4284,
+      "step": 29960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660847927919034e-05,
+      "loss": 2.5957,
+      "step": 29965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660735658822086e-05,
+      "loss": 2.5832,
+      "step": 29970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660623371798621e-05,
+      "loss": 2.4778,
+      "step": 29975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660511066849076e-05,
+      "loss": 2.5329,
+      "step": 29980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660398743973878e-05,
+      "loss": 2.5387,
+      "step": 29985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660286403173461e-05,
+      "loss": 2.5376,
+      "step": 29990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660174044448258e-05,
+      "loss": 2.5452,
+      "step": 29995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660061667798701e-05,
+      "loss": 2.5482,
+      "step": 30000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65994927322522e-05,
+      "loss": 2.4926,
+      "step": 30005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.659836860728251e-05,
+      "loss": 2.5917,
+      "step": 30010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.659724430308224e-05,
+      "loss": 2.4884,
+      "step": 30015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.659611981965571e-05,
+      "loss": 2.4775,
+      "step": 30020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.659499515700725e-05,
+      "loss": 2.5963,
+      "step": 30025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65938703151412e-05,
+      "loss": 2.5808,
+      "step": 30030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.659274529406187e-05,
+      "loss": 2.4858,
+      "step": 30035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.659162009377361e-05,
+      "loss": 2.4698,
+      "step": 30040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65904947142807e-05,
+      "loss": 2.5467,
+      "step": 30045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658936915558752e-05,
+      "loss": 2.5393,
+      "step": 30050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658824341769837e-05,
+      "loss": 2.6458,
+      "step": 30055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65871175006176e-05,
+      "loss": 2.6302,
+      "step": 30060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658599140434952e-05,
+      "loss": 2.5773,
+      "step": 30065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658486512889846e-05,
+      "loss": 2.4708,
+      "step": 30070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658373867426877e-05,
+      "loss": 2.5941,
+      "step": 30075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658261204046478e-05,
+      "loss": 2.6088,
+      "step": 30080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658148522749083e-05,
+      "loss": 2.4769,
+      "step": 30085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658035823535121e-05,
+      "loss": 2.5956,
+      "step": 30090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657923106405031e-05,
+      "loss": 2.5574,
+      "step": 30095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657810371359244e-05,
+      "loss": 2.5768,
+      "step": 30100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657697618398193e-05,
+      "loss": 2.5307,
+      "step": 30105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657584847522313e-05,
+      "loss": 2.5416,
+      "step": 30110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657472058732038e-05,
+      "loss": 2.4565,
+      "step": 30115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6573592520278e-05,
+      "loss": 2.4343,
+      "step": 30120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657246427410035e-05,
+      "loss": 2.5589,
+      "step": 30125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657133584879175e-05,
+      "loss": 2.6533,
+      "step": 30130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657020724435654e-05,
+      "loss": 2.5789,
+      "step": 30135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65690784607991e-05,
+      "loss": 2.5339,
+      "step": 30140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.656794949812373e-05,
+      "loss": 2.5295,
+      "step": 30145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.656682035633478e-05,
+      "loss": 2.4158,
+      "step": 30150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.656569103543659e-05,
+      "loss": 2.6598,
+      "step": 30155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.656456153543351e-05,
+      "loss": 2.567,
+      "step": 30160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65634318563299e-05,
+      "loss": 2.4027,
+      "step": 30165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.656230199813009e-05,
+      "loss": 2.5543,
+      "step": 30170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.656117196083843e-05,
+      "loss": 2.5498,
+      "step": 30175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.656004174445925e-05,
+      "loss": 2.5181,
+      "step": 30180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.655891134899691e-05,
+      "loss": 2.5344,
+      "step": 30185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.655778077445577e-05,
+      "loss": 2.5701,
+      "step": 30190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.655665002084016e-05,
+      "loss": 2.5534,
+      "step": 30195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.655551908815442e-05,
+      "loss": 2.5952,
+      "step": 30200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.655438797640293e-05,
+      "loss": 2.4004,
+      "step": 30205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.655325668559003e-05,
+      "loss": 2.4913,
+      "step": 30210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.655212521572006e-05,
+      "loss": 2.5435,
+      "step": 30215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.655099356679737e-05,
+      "loss": 2.669,
+      "step": 30220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.654986173882635e-05,
+      "loss": 2.6197,
+      "step": 30225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65487297318113e-05,
+      "loss": 2.3392,
+      "step": 30230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65475975457566e-05,
+      "loss": 2.5483,
+      "step": 30235
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.654646518066662e-05,
+      "loss": 2.4442,
+      "step": 30240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.654533263654569e-05,
+      "loss": 2.5887,
+      "step": 30245
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.654419991339818e-05,
+      "loss": 2.515,
+      "step": 30250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.654306701122844e-05,
+      "loss": 2.541,
+      "step": 30255
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.654193393004083e-05,
+      "loss": 2.4442,
+      "step": 30260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.654080066983971e-05,
+      "loss": 2.5774,
+      "step": 30265
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653966723062942e-05,
+      "loss": 2.5603,
+      "step": 30270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653853361241434e-05,
+      "loss": 2.5675,
+      "step": 30275
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653739981519884e-05,
+      "loss": 2.4847,
+      "step": 30280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653626583898726e-05,
+      "loss": 2.5322,
+      "step": 30285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653513168378396e-05,
+      "loss": 2.4997,
+      "step": 30290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653399734959333e-05,
+      "loss": 2.4566,
+      "step": 30295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653286283641971e-05,
+      "loss": 2.3668,
+      "step": 30300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653172814426746e-05,
+      "loss": 2.5454,
+      "step": 30305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653059327314095e-05,
+      "loss": 2.5418,
+      "step": 30310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652945822304456e-05,
+      "loss": 2.56,
+      "step": 30315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652832299398263e-05,
+      "loss": 2.4448,
+      "step": 30320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652718758595956e-05,
+      "loss": 2.5564,
+      "step": 30325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652605199897966e-05,
+      "loss": 2.5649,
+      "step": 30330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652491623304736e-05,
+      "loss": 2.5522,
+      "step": 30335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6523780288167e-05,
+      "loss": 2.6524,
+      "step": 30340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652264416434296e-05,
+      "loss": 2.5896,
+      "step": 30345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652150786157958e-05,
+      "loss": 2.5209,
+      "step": 30350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652037137988125e-05,
+      "loss": 2.5191,
+      "step": 30355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.651923471925238e-05,
+      "loss": 2.5563,
+      "step": 30360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.651809787969728e-05,
+      "loss": 2.7175,
+      "step": 30365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.651696086122034e-05,
+      "loss": 2.5108,
+      "step": 30370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.651582366382595e-05,
+      "loss": 2.4702,
+      "step": 30375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.651468628751848e-05,
+      "loss": 2.4058,
+      "step": 30380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.651354873230229e-05,
+      "loss": 2.6388,
+      "step": 30385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.651241099818178e-05,
+      "loss": 2.574,
+      "step": 30390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.651127308516131e-05,
+      "loss": 2.5319,
+      "step": 30395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.651013499324524e-05,
+      "loss": 2.5399,
+      "step": 30400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650899672243799e-05,
+      "loss": 2.5742,
+      "step": 30405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65078582727439e-05,
+      "loss": 2.4096,
+      "step": 30410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650671964416735e-05,
+      "loss": 2.5196,
+      "step": 30415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650558083671275e-05,
+      "loss": 2.6254,
+      "step": 30420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650444185038446e-05,
+      "loss": 2.3923,
+      "step": 30425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650330268518686e-05,
+      "loss": 2.2647,
+      "step": 30430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650216334112432e-05,
+      "loss": 2.4753,
+      "step": 30435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650102381820125e-05,
+      "loss": 2.4699,
+      "step": 30440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.649988411642203e-05,
+      "loss": 2.4936,
+      "step": 30445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.649874423579102e-05,
+      "loss": 2.6065,
+      "step": 30450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.649760417631261e-05,
+      "loss": 2.6781,
+      "step": 30455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64964639379912e-05,
+      "loss": 2.4622,
+      "step": 30460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.649532352083118e-05,
+      "loss": 2.5014,
+      "step": 30465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.649418292483691e-05,
+      "loss": 2.5349,
+      "step": 30470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.649304215001281e-05,
+      "loss": 2.4375,
+      "step": 30475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.649190119636324e-05,
+      "loss": 2.6211,
+      "step": 30480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64907600638926e-05,
+      "loss": 2.5228,
+      "step": 30485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648961875260527e-05,
+      "loss": 2.5122,
+      "step": 30490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648847726250565e-05,
+      "loss": 2.4821,
+      "step": 30495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648733559359813e-05,
+      "loss": 2.5937,
+      "step": 30500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648619374588707e-05,
+      "loss": 2.5886,
+      "step": 30505
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648505171937693e-05,
+      "loss": 2.6812,
+      "step": 30510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648390951407205e-05,
+      "loss": 2.5331,
+      "step": 30515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648276712997684e-05,
+      "loss": 2.6043,
+      "step": 30520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648162456709567e-05,
+      "loss": 2.6409,
+      "step": 30525
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648048182543297e-05,
+      "loss": 2.5141,
+      "step": 30530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647933890499312e-05,
+      "loss": 2.502,
+      "step": 30535
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647819580578052e-05,
+      "loss": 2.5837,
+      "step": 30540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647705252779955e-05,
+      "loss": 2.493,
+      "step": 30545
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647590907105464e-05,
+      "loss": 2.5032,
+      "step": 30550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647476543555015e-05,
+      "loss": 2.3877,
+      "step": 30555
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64736216212905e-05,
+      "loss": 2.4334,
+      "step": 30560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647247762828009e-05,
+      "loss": 2.5589,
+      "step": 30565
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647133345652332e-05,
+      "loss": 2.5432,
+      "step": 30570
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647018910602458e-05,
+      "loss": 2.4914,
+      "step": 30575
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646904457678828e-05,
+      "loss": 2.6036,
+      "step": 30580
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646789986881882e-05,
+      "loss": 2.6322,
+      "step": 30585
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646675498212061e-05,
+      "loss": 2.5316,
+      "step": 30590
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646560991669804e-05,
+      "loss": 2.5915,
+      "step": 30595
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646446467255552e-05,
+      "loss": 2.5767,
+      "step": 30600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646331924969747e-05,
+      "loss": 2.6198,
+      "step": 30605
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646217364812828e-05,
+      "loss": 2.5182,
+      "step": 30610
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646102786785234e-05,
+      "loss": 2.5196,
+      "step": 30615
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64598819088741e-05,
+      "loss": 2.575,
+      "step": 30620
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.645873577119793e-05,
+      "loss": 2.5079,
+      "step": 30625
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.645758945482825e-05,
+      "loss": 2.561,
+      "step": 30630
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.645644295976946e-05,
+      "loss": 2.5843,
+      "step": 30635
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6455296286026e-05,
+      "loss": 2.4166,
+      "step": 30640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.645414943360225e-05,
+      "loss": 2.5323,
+      "step": 30645
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.645300240250263e-05,
+      "loss": 2.3703,
+      "step": 30650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.645185519273156e-05,
+      "loss": 2.492,
+      "step": 30655
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.645070780429346e-05,
+      "loss": 2.3828,
+      "step": 30660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644956023719272e-05,
+      "loss": 2.4669,
+      "step": 30665
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644841249143375e-05,
+      "loss": 2.641,
+      "step": 30670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644726456702098e-05,
+      "loss": 2.5743,
+      "step": 30675
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644611646395883e-05,
+      "loss": 2.5305,
+      "step": 30680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644496818225171e-05,
+      "loss": 2.5677,
+      "step": 30685
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644381972190404e-05,
+      "loss": 2.3926,
+      "step": 30690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644267108292023e-05,
+      "loss": 2.6153,
+      "step": 30695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644152226530469e-05,
+      "loss": 2.518,
+      "step": 30700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644037326906185e-05,
+      "loss": 2.3664,
+      "step": 30705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643922409419614e-05,
+      "loss": 2.5032,
+      "step": 30710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643807474071197e-05,
+      "loss": 2.5821,
+      "step": 30715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643692520861375e-05,
+      "loss": 2.5674,
+      "step": 30720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643577549790592e-05,
+      "loss": 2.4727,
+      "step": 30725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643462560859288e-05,
+      "loss": 2.5609,
+      "step": 30730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643347554067908e-05,
+      "loss": 2.5963,
+      "step": 30735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643232529416892e-05,
+      "loss": 2.6699,
+      "step": 30740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643117486906683e-05,
+      "loss": 2.608,
+      "step": 30745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643002426537726e-05,
+      "loss": 2.4928,
+      "step": 30750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.642887348310459e-05,
+      "loss": 2.5869,
+      "step": 30755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.642772252225328e-05,
+      "loss": 2.5973,
+      "step": 30760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.642657138282776e-05,
+      "loss": 2.6293,
+      "step": 30765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.642542006483243e-05,
+      "loss": 2.4934,
+      "step": 30770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.642426856827173e-05,
+      "loss": 2.6539,
+      "step": 30775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64231168931501e-05,
+      "loss": 2.5436,
+      "step": 30780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.642196503947196e-05,
+      "loss": 2.4164,
+      "step": 30785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.642081300724174e-05,
+      "loss": 2.5663,
+      "step": 30790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.641966079646388e-05,
+      "loss": 2.5581,
+      "step": 30795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64185084071428e-05,
+      "loss": 2.5817,
+      "step": 30800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.641735583928293e-05,
+      "loss": 2.5581,
+      "step": 30805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.641620309288872e-05,
+      "loss": 2.5931,
+      "step": 30810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64150501679646e-05,
+      "loss": 2.4316,
+      "step": 30815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6413897064515e-05,
+      "loss": 2.5661,
+      "step": 30820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.641274378254434e-05,
+      "loss": 2.4338,
+      "step": 30825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.641159032205706e-05,
+      "loss": 2.4666,
+      "step": 30830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.641043668305763e-05,
+      "loss": 2.567,
+      "step": 30835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640928286555046e-05,
+      "loss": 2.6047,
+      "step": 30840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640812886953999e-05,
+      "loss": 2.5631,
+      "step": 30845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640697469503064e-05,
+      "loss": 2.4868,
+      "step": 30850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64058203420269e-05,
+      "loss": 2.4614,
+      "step": 30855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640466581053315e-05,
+      "loss": 2.4329,
+      "step": 30860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640351110055389e-05,
+      "loss": 2.4341,
+      "step": 30865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640235621209352e-05,
+      "loss": 2.599,
+      "step": 30870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640120114515648e-05,
+      "loss": 2.5516,
+      "step": 30875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640004589974723e-05,
+      "loss": 2.548,
+      "step": 30880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.639889047587023e-05,
+      "loss": 2.4488,
+      "step": 30885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.639773487352987e-05,
+      "loss": 2.4927,
+      "step": 30890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.639657909273065e-05,
+      "loss": 2.3843,
+      "step": 30895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6395423133477e-05,
+      "loss": 2.4877,
+      "step": 30900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.639426699577335e-05,
+      "loss": 2.48,
+      "step": 30905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.639311067962415e-05,
+      "loss": 2.489,
+      "step": 30910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.639195418503388e-05,
+      "loss": 2.4764,
+      "step": 30915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.639079751200693e-05,
+      "loss": 2.5629,
+      "step": 30920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638964066054781e-05,
+      "loss": 2.6427,
+      "step": 30925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638848363066092e-05,
+      "loss": 2.549,
+      "step": 30930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638732642235074e-05,
+      "loss": 2.5515,
+      "step": 30935
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63861690356217e-05,
+      "loss": 2.5769,
+      "step": 30940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638501147047829e-05,
+      "loss": 2.4265,
+      "step": 30945
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638385372692492e-05,
+      "loss": 2.4187,
+      "step": 30950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638269580496605e-05,
+      "loss": 2.5245,
+      "step": 30955
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638153770460615e-05,
+      "loss": 2.5277,
+      "step": 30960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638037942584968e-05,
+      "loss": 2.519,
+      "step": 30965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.637922096870107e-05,
+      "loss": 2.5913,
+      "step": 30970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.637806233316477e-05,
+      "loss": 2.3762,
+      "step": 30975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.637690351924529e-05,
+      "loss": 2.5686,
+      "step": 30980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.637574452694704e-05,
+      "loss": 2.6698,
+      "step": 30985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.637458535627449e-05,
+      "loss": 2.6621,
+      "step": 30990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.637342600723207e-05,
+      "loss": 2.5524,
+      "step": 30995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63722664798243e-05,
+      "loss": 2.5811,
+      "step": 31000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63711067740556e-05,
+      "loss": 2.4906,
+      "step": 31005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636994688993044e-05,
+      "loss": 2.4809,
+      "step": 31010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636878682745326e-05,
+      "loss": 2.522,
+      "step": 31015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636762658662855e-05,
+      "loss": 2.4263,
+      "step": 31020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636646616746077e-05,
+      "loss": 2.5725,
+      "step": 31025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636530556995438e-05,
+      "loss": 2.5089,
+      "step": 31030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636414479411385e-05,
+      "loss": 2.4784,
+      "step": 31035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63629838399436e-05,
+      "loss": 2.718,
+      "step": 31040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636182270744815e-05,
+      "loss": 2.5565,
+      "step": 31045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636066139663195e-05,
+      "loss": 2.4283,
+      "step": 31050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635949990749946e-05,
+      "loss": 2.5504,
+      "step": 31055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635833824005516e-05,
+      "loss": 2.5293,
+      "step": 31060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63571763943035e-05,
+      "loss": 2.4102,
+      "step": 31065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635601437024895e-05,
+      "loss": 2.4327,
+      "step": 31070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635485216789599e-05,
+      "loss": 2.5566,
+      "step": 31075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63536897872491e-05,
+      "loss": 2.4624,
+      "step": 31080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635252722831274e-05,
+      "loss": 2.5552,
+      "step": 31085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635136449109136e-05,
+      "loss": 2.5228,
+      "step": 31090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635020157558947e-05,
+      "loss": 2.634,
+      "step": 31095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.634903848181151e-05,
+      "loss": 2.5718,
+      "step": 31100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.634787520976198e-05,
+      "loss": 2.5853,
+      "step": 31105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.634671175944534e-05,
+      "loss": 2.5032,
+      "step": 31110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.634554813086607e-05,
+      "loss": 2.5781,
+      "step": 31115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.634438432402864e-05,
+      "loss": 2.5412,
+      "step": 31120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.634322033893752e-05,
+      "loss": 2.5434,
+      "step": 31125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.634205617559722e-05,
+      "loss": 2.6481,
+      "step": 31130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.634089183401219e-05,
+      "loss": 2.653,
+      "step": 31135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633972731418689e-05,
+      "loss": 2.5402,
+      "step": 31140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633856261612583e-05,
+      "loss": 2.5237,
+      "step": 31145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63373977398335e-05,
+      "loss": 2.5659,
+      "step": 31150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633623268531434e-05,
+      "loss": 2.501,
+      "step": 31155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633506745257286e-05,
+      "loss": 2.719,
+      "step": 31160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633390204161354e-05,
+      "loss": 2.5714,
+      "step": 31165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633273645244086e-05,
+      "loss": 2.5119,
+      "step": 31170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633157068505929e-05,
+      "loss": 2.4214,
+      "step": 31175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633040473947332e-05,
+      "loss": 2.4485,
+      "step": 31180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.632923861568745e-05,
+      "loss": 2.405,
+      "step": 31185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.632807231370616e-05,
+      "loss": 2.5264,
+      "step": 31190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.632690583353393e-05,
+      "loss": 2.5621,
+      "step": 31195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.632573917517525e-05,
+      "loss": 2.5428,
+      "step": 31200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.632457233863458e-05,
+      "loss": 2.5378,
+      "step": 31205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.632340532391646e-05,
+      "loss": 2.6231,
+      "step": 31210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.632223813102533e-05,
+      "loss": 2.5086,
+      "step": 31215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63210707599657e-05,
+      "loss": 2.5392,
+      "step": 31220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.631990321074208e-05,
+      "loss": 2.5668,
+      "step": 31225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.631873548335893e-05,
+      "loss": 2.5007,
+      "step": 31230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.631756757782075e-05,
+      "loss": 2.672,
+      "step": 31235
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.631639949413203e-05,
+      "loss": 2.5016,
+      "step": 31240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.631523123229728e-05,
+      "loss": 2.5515,
+      "step": 31245
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.631406279232098e-05,
+      "loss": 2.4661,
+      "step": 31250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.631289417420761e-05,
+      "loss": 2.3928,
+      "step": 31255
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63117253779617e-05,
+      "loss": 2.3589,
+      "step": 31260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63105564035877e-05,
+      "loss": 2.5847,
+      "step": 31265
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630938725109015e-05,
+      "loss": 2.5574,
+      "step": 31270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630821792047354e-05,
+      "loss": 2.5099,
+      "step": 31275
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630704841174235e-05,
+      "loss": 2.5084,
+      "step": 31280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630587872490106e-05,
+      "loss": 2.6287,
+      "step": 31285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63047088599542e-05,
+      "loss": 2.6229,
+      "step": 31290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630353881690628e-05,
+      "loss": 2.2685,
+      "step": 31295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630236859576177e-05,
+      "loss": 2.5092,
+      "step": 31300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630119819652519e-05,
+      "loss": 2.5601,
+      "step": 31305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630002761920103e-05,
+      "loss": 2.5083,
+      "step": 31310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62988568637938e-05,
+      "loss": 2.5162,
+      "step": 31315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629768593030801e-05,
+      "loss": 2.6807,
+      "step": 31320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629651481874815e-05,
+      "loss": 2.6508,
+      "step": 31325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629534352911872e-05,
+      "loss": 2.6384,
+      "step": 31330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629417206142425e-05,
+      "loss": 2.444,
+      "step": 31335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629300041566922e-05,
+      "loss": 2.5468,
+      "step": 31340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629182859185815e-05,
+      "loss": 2.5533,
+      "step": 31345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629065658999554e-05,
+      "loss": 2.5551,
+      "step": 31350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628948441008592e-05,
+      "loss": 2.5353,
+      "step": 31355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628831205213375e-05,
+      "loss": 2.521,
+      "step": 31360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628713951614359e-05,
+      "loss": 2.548,
+      "step": 31365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628596680211992e-05,
+      "loss": 2.4706,
+      "step": 31370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628479391006727e-05,
+      "loss": 2.5514,
+      "step": 31375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628362083999013e-05,
+      "loss": 2.5214,
+      "step": 31380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628244759189304e-05,
+      "loss": 2.5996,
+      "step": 31385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628127416578047e-05,
+      "loss": 2.5395,
+      "step": 31390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628010056165699e-05,
+      "loss": 2.4609,
+      "step": 31395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627892677952707e-05,
+      "loss": 2.4908,
+      "step": 31400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627775281939524e-05,
+      "loss": 2.4544,
+      "step": 31405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6276578681266e-05,
+      "loss": 2.5557,
+      "step": 31410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627540436514389e-05,
+      "loss": 2.3888,
+      "step": 31415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627422987103342e-05,
+      "loss": 2.5476,
+      "step": 31420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627305519893909e-05,
+      "loss": 2.5321,
+      "step": 31425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627188034886545e-05,
+      "loss": 2.6123,
+      "step": 31430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6270705320817e-05,
+      "loss": 2.5614,
+      "step": 31435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626953011479825e-05,
+      "loss": 2.4246,
+      "step": 31440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626835473081373e-05,
+      "loss": 2.5291,
+      "step": 31445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626717916886797e-05,
+      "loss": 2.5374,
+      "step": 31450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626600342896546e-05,
+      "loss": 2.3887,
+      "step": 31455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626482751111077e-05,
+      "loss": 2.5809,
+      "step": 31460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626365141530837e-05,
+      "loss": 2.442,
+      "step": 31465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626247514156283e-05,
+      "loss": 2.4939,
+      "step": 31470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626129868987866e-05,
+      "loss": 2.5934,
+      "step": 31475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626012206026038e-05,
+      "loss": 2.593,
+      "step": 31480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62589452527125e-05,
+      "loss": 2.5225,
+      "step": 31485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625776826723959e-05,
+      "loss": 2.4252,
+      "step": 31490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625659110384612e-05,
+      "loss": 2.5384,
+      "step": 31495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625541376253668e-05,
+      "loss": 2.6142,
+      "step": 31500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625423624331574e-05,
+      "loss": 2.6337,
+      "step": 31505
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625305854618786e-05,
+      "loss": 2.5774,
+      "step": 31510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625188067115758e-05,
+      "loss": 2.582,
+      "step": 31515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62507026182294e-05,
+      "loss": 2.4456,
+      "step": 31520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624952438740787e-05,
+      "loss": 2.5055,
+      "step": 31525
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62483459786975e-05,
+      "loss": 2.5635,
+      "step": 31530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624716739210287e-05,
+      "loss": 2.6117,
+      "step": 31535
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624598862762846e-05,
+      "loss": 2.6021,
+      "step": 31540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624480968527885e-05,
+      "loss": 2.4968,
+      "step": 31545
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624363056505854e-05,
+      "loss": 2.5482,
+      "step": 31550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624245126697206e-05,
+      "loss": 2.4556,
+      "step": 31555
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624127179102399e-05,
+      "loss": 2.4429,
+      "step": 31560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624009213721883e-05,
+      "loss": 2.566,
+      "step": 31565
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623891230556113e-05,
+      "loss": 2.4621,
+      "step": 31570
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623773229605541e-05,
+      "loss": 2.493,
+      "step": 31575
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623655210870625e-05,
+      "loss": 2.5257,
+      "step": 31580
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623537174351816e-05,
+      "loss": 2.5659,
+      "step": 31585
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623419120049566e-05,
+      "loss": 2.4591,
+      "step": 31590
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623301047964333e-05,
+      "loss": 2.5779,
+      "step": 31595
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623182958096568e-05,
+      "loss": 2.7083,
+      "step": 31600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623064850446728e-05,
+      "loss": 2.5768,
+      "step": 31605
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622946725015268e-05,
+      "loss": 2.6529,
+      "step": 31610
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622828581802637e-05,
+      "loss": 2.4532,
+      "step": 31615
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622710420809293e-05,
+      "loss": 2.5326,
+      "step": 31620
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622592242035692e-05,
+      "loss": 2.5259,
+      "step": 31625
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622474045482286e-05,
+      "loss": 2.5244,
+      "step": 31630
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622355831149529e-05,
+      "loss": 2.6289,
+      "step": 31635
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622237599037879e-05,
+      "loss": 2.5836,
+      "step": 31640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622119349147788e-05,
+      "loss": 2.4694,
+      "step": 31645
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622001081479712e-05,
+      "loss": 2.5303,
+      "step": 31650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621882796034107e-05,
+      "loss": 2.3478,
+      "step": 31655
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621764492811425e-05,
+      "loss": 2.685,
+      "step": 31660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621646171812122e-05,
+      "loss": 2.5228,
+      "step": 31665
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621527833036654e-05,
+      "loss": 2.5563,
+      "step": 31670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.621409476485477e-05,
+      "loss": 2.3505,
+      "step": 31675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.621291102159043e-05,
+      "loss": 2.533,
+      "step": 31680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.621172710057811e-05,
+      "loss": 2.4962,
+      "step": 31685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.621054300182235e-05,
+      "loss": 2.452,
+      "step": 31690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620935872532768e-05,
+      "loss": 2.5541,
+      "step": 31695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620817427109869e-05,
+      "loss": 2.6241,
+      "step": 31700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620698963913992e-05,
+      "loss": 2.4756,
+      "step": 31705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620580482945593e-05,
+      "loss": 2.5509,
+      "step": 31710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620461984205128e-05,
+      "loss": 2.5368,
+      "step": 31715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620343467693052e-05,
+      "loss": 2.5918,
+      "step": 31720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620224933409822e-05,
+      "loss": 2.5535,
+      "step": 31725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620106381355891e-05,
+      "loss": 2.6617,
+      "step": 31730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.619987811531719e-05,
+      "loss": 2.4909,
+      "step": 31735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61986922393776e-05,
+      "loss": 2.4304,
+      "step": 31740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61975061857447e-05,
+      "loss": 2.6401,
+      "step": 31745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.619631995442305e-05,
+      "loss": 2.4133,
+      "step": 31750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.619513354541722e-05,
+      "loss": 2.4535,
+      "step": 31755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.619394695873177e-05,
+      "loss": 2.5125,
+      "step": 31760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.619276019437127e-05,
+      "loss": 2.4793,
+      "step": 31765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.619157325234028e-05,
+      "loss": 2.4467,
+      "step": 31770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.619038613264336e-05,
+      "loss": 2.395,
+      "step": 31775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.618919883528508e-05,
+      "loss": 2.5893,
+      "step": 31780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.618801136027e-05,
+      "loss": 2.4316,
+      "step": 31785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61868237076027e-05,
+      "loss": 2.4307,
+      "step": 31790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.618563587728776e-05,
+      "loss": 2.5868,
+      "step": 31795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61844478693297e-05,
+      "loss": 2.5544,
+      "step": 31800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.618325968373315e-05,
+      "loss": 2.5936,
+      "step": 31805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.618207132050263e-05,
+      "loss": 2.5069,
+      "step": 31810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.618088277964274e-05,
+      "loss": 2.4767,
+      "step": 31815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617969406115805e-05,
+      "loss": 2.6167,
+      "step": 31820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617850516505311e-05,
+      "loss": 2.4489,
+      "step": 31825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617731609133251e-05,
+      "loss": 2.5142,
+      "step": 31830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617612684000083e-05,
+      "loss": 2.5208,
+      "step": 31835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617493741106264e-05,
+      "loss": 2.5107,
+      "step": 31840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61737478045225e-05,
+      "loss": 2.4206,
+      "step": 31845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.6172558020385e-05,
+      "loss": 2.517,
+      "step": 31850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617136805865472e-05,
+      "loss": 2.475,
+      "step": 31855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617017791933623e-05,
+      "loss": 2.3893,
+      "step": 31860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61689876024341e-05,
+      "loss": 2.4686,
+      "step": 31865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.616779710795289e-05,
+      "loss": 2.5531,
+      "step": 31870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.616660643589723e-05,
+      "loss": 2.6864,
+      "step": 31875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.616541558627168e-05,
+      "loss": 2.5456,
+      "step": 31880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.616422455908081e-05,
+      "loss": 2.5935,
+      "step": 31885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.616303335432921e-05,
+      "loss": 2.6451,
+      "step": 31890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.616184197202144e-05,
+      "loss": 2.5081,
+      "step": 31895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.616065041216212e-05,
+      "loss": 2.5139,
+      "step": 31900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61594586747558e-05,
+      "loss": 2.5636,
+      "step": 31905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.615826675980707e-05,
+      "loss": 2.4268,
+      "step": 31910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.615707466732054e-05,
+      "loss": 2.5418,
+      "step": 31915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.615588239730075e-05,
+      "loss": 2.5858,
+      "step": 31920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.615468994975233e-05,
+      "loss": 2.4846,
+      "step": 31925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.615349732467984e-05,
+      "loss": 2.6822,
+      "step": 31930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.615230452208788e-05,
+      "loss": 2.4819,
+      "step": 31935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.615111154198104e-05,
+      "loss": 2.5959,
+      "step": 31940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61499183843639e-05,
+      "loss": 2.4246,
+      "step": 31945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.614872504924104e-05,
+      "loss": 2.4582,
+      "step": 31950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.614753153661706e-05,
+      "loss": 2.4284,
+      "step": 31955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.614633784649656e-05,
+      "loss": 2.511,
+      "step": 31960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.614514397888412e-05,
+      "loss": 2.5193,
+      "step": 31965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.614394993378435e-05,
+      "loss": 2.6188,
+      "step": 31970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.614275571120182e-05,
+      "loss": 2.5992,
+      "step": 31975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.614156131114111e-05,
+      "loss": 2.5388,
+      "step": 31980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.614036673360685e-05,
+      "loss": 2.4432,
+      "step": 31985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.613917197860361e-05,
+      "loss": 2.5605,
+      "step": 31990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.6137977046136e-05,
+      "loss": 2.6291,
+      "step": 31995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.613678193620863e-05,
+      "loss": 2.5248,
+      "step": 32000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.613558664882605e-05,
+      "loss": 2.5251,
+      "step": 32005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61343911839929e-05,
+      "loss": 2.3874,
+      "step": 32010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.613319554171376e-05,
+      "loss": 2.5619,
+      "step": 32015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.613199972199322e-05,
+      "loss": 2.5631,
+      "step": 32020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.613080372483591e-05,
+      "loss": 2.4725,
+      "step": 32025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61296075502464e-05,
+      "loss": 2.4903,
+      "step": 32030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61284111982293e-05,
+      "loss": 2.5359,
+      "step": 32035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.612721466878922e-05,
+      "loss": 2.5744,
+      "step": 32040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.612601796193076e-05,
+      "loss": 2.4526,
+      "step": 32045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.612482107765852e-05,
+      "loss": 2.4138,
+      "step": 32050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61236240159771e-05,
+      "loss": 2.5409,
+      "step": 32055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.612242677689109e-05,
+      "loss": 2.5719,
+      "step": 32060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.612122936040514e-05,
+      "loss": 2.4552,
+      "step": 32065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61200317665238e-05,
+      "loss": 2.5165,
+      "step": 32070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.611883399525173e-05,
+      "loss": 2.5465,
+      "step": 32075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61176360465935e-05,
+      "loss": 2.4323,
+      "step": 32080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.611643792055371e-05,
+      "loss": 2.4868,
+      "step": 32085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.6115239617137e-05,
+      "loss": 2.5341,
+      "step": 32090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.611404113634797e-05,
+      "loss": 2.3633,
+      "step": 32095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.611284247819122e-05,
+      "loss": 2.4821,
+      "step": 32100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.611164364267136e-05,
+      "loss": 2.542,
+      "step": 32105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.611044462979301e-05,
+      "loss": 2.4951,
+      "step": 32110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.610924543956079e-05,
+      "loss": 2.583,
+      "step": 32115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.610804607197929e-05,
+      "loss": 2.6453,
+      "step": 32120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.610684652705313e-05,
+      "loss": 2.4146,
+      "step": 32125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.610564680478692e-05,
+      "loss": 2.5057,
+      "step": 32130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61044469051853e-05,
+      "loss": 2.5202,
+      "step": 32135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.610324682825286e-05,
+      "loss": 2.653,
+      "step": 32140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.610204657399424e-05,
+      "loss": 2.3828,
+      "step": 32145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.610084614241402e-05,
+      "loss": 2.4863,
+      "step": 32150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609964553351685e-05,
+      "loss": 2.4919,
+      "step": 32155
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609844474730731e-05,
+      "loss": 2.5851,
+      "step": 32160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609724378379007e-05,
+      "loss": 2.592,
+      "step": 32165
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609604264296972e-05,
+      "loss": 2.5412,
+      "step": 32170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609484132485087e-05,
+      "loss": 2.5371,
+      "step": 32175
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609363982943816e-05,
+      "loss": 2.5423,
+      "step": 32180
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60924381567362e-05,
+      "loss": 2.5373,
+      "step": 32185
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609123630674961e-05,
+      "loss": 2.5981,
+      "step": 32190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609003427948303e-05,
+      "loss": 2.5746,
+      "step": 32195
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608883207494107e-05,
+      "loss": 2.4172,
+      "step": 32200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608762969312836e-05,
+      "loss": 2.5796,
+      "step": 32205
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608642713404951e-05,
+      "loss": 2.7337,
+      "step": 32210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608522439770917e-05,
+      "loss": 2.599,
+      "step": 32215
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608402148411193e-05,
+      "loss": 2.5853,
+      "step": 32220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608281839326247e-05,
+      "loss": 2.6256,
+      "step": 32225
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608161512516535e-05,
+      "loss": 2.5796,
+      "step": 32230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608041167982527e-05,
+      "loss": 2.5816,
+      "step": 32235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607920805724682e-05,
+      "loss": 2.5307,
+      "step": 32240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607800425743461e-05,
+      "loss": 2.4833,
+      "step": 32245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607680028039331e-05,
+      "loss": 2.4847,
+      "step": 32250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607559612612754e-05,
+      "loss": 2.4242,
+      "step": 32255
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607439179464191e-05,
+      "loss": 2.5128,
+      "step": 32260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607318728594107e-05,
+      "loss": 2.5143,
+      "step": 32265
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607198260002966e-05,
+      "loss": 2.6076,
+      "step": 32270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607077773691229e-05,
+      "loss": 2.6322,
+      "step": 32275
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606957269659362e-05,
+      "loss": 2.6249,
+      "step": 32280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606836747907826e-05,
+      "loss": 2.4712,
+      "step": 32285
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606716208437088e-05,
+      "loss": 2.4575,
+      "step": 32290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606595651247607e-05,
+      "loss": 2.5737,
+      "step": 32295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606475076339852e-05,
+      "loss": 2.5395,
+      "step": 32300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606354483714281e-05,
+      "loss": 2.4192,
+      "step": 32305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606233873371363e-05,
+      "loss": 2.4789,
+      "step": 32310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606113245311559e-05,
+      "loss": 2.5076,
+      "step": 32315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605992599535334e-05,
+      "loss": 2.6794,
+      "step": 32320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605871936043151e-05,
+      "loss": 2.559,
+      "step": 32325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605751254835476e-05,
+      "loss": 2.5816,
+      "step": 32330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605630555912771e-05,
+      "loss": 2.5331,
+      "step": 32335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605509839275502e-05,
+      "loss": 2.4502,
+      "step": 32340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605389104924132e-05,
+      "loss": 2.4951,
+      "step": 32345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605268352859127e-05,
+      "loss": 2.4757,
+      "step": 32350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60514758308095e-05,
+      "loss": 2.6318,
+      "step": 32355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605026795590064e-05,
+      "loss": 2.535,
+      "step": 32360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.604905990386938e-05,
+      "loss": 2.5097,
+      "step": 32365
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.604785167472033e-05,
+      "loss": 2.456,
+      "step": 32370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.604664326845813e-05,
+      "loss": 2.5687,
+      "step": 32375
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.604543468508749e-05,
+      "loss": 2.6408,
+      "step": 32380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.604422592461298e-05,
+      "loss": 2.3598,
+      "step": 32385
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60430169870393e-05,
+      "loss": 2.5563,
+      "step": 32390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.604180787237108e-05,
+      "loss": 2.6002,
+      "step": 32395
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.604059858061295e-05,
+      "loss": 2.4756,
+      "step": 32400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.603938911176963e-05,
+      "loss": 2.4336,
+      "step": 32405
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60381794658457e-05,
+      "loss": 2.478,
+      "step": 32410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.603696964284585e-05,
+      "loss": 2.5631,
+      "step": 32415
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.603575964277473e-05,
+      "loss": 2.497,
+      "step": 32420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.603454946563698e-05,
+      "loss": 2.6258,
+      "step": 32425
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.603333911143726e-05,
+      "loss": 2.4367,
+      "step": 32430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.603212858018023e-05,
+      "loss": 2.5311,
+      "step": 32435
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.603091787187055e-05,
+      "loss": 2.4903,
+      "step": 32440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.602970698651286e-05,
+      "loss": 2.6925,
+      "step": 32445
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.602849592411184e-05,
+      "loss": 2.57,
+      "step": 32450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.602728468467212e-05,
+      "loss": 2.5095,
+      "step": 32455
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.602607326819838e-05,
+      "loss": 2.519,
+      "step": 32460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.602486167469529e-05,
+      "loss": 2.5332,
+      "step": 32465
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60236499041675e-05,
+      "loss": 2.5493,
+      "step": 32470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.602243795661964e-05,
+      "loss": 2.5321,
+      "step": 32475
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.602122583205642e-05,
+      "loss": 2.375,
+      "step": 32480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.602001353048246e-05,
+      "loss": 2.572,
+      "step": 32485
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.601880105190244e-05,
+      "loss": 2.5476,
+      "step": 32490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.601758839632103e-05,
+      "loss": 2.5488,
+      "step": 32495
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60163755637429e-05,
+      "loss": 2.5547,
+      "step": 32500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60151625541727e-05,
+      "loss": 2.5166,
+      "step": 32505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60139493676151e-05,
+      "loss": 2.5224,
+      "step": 32510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.601273600407477e-05,
+      "loss": 2.5721,
+      "step": 32515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.601152246355636e-05,
+      "loss": 2.4815,
+      "step": 32520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.601030874606456e-05,
+      "loss": 2.4863,
+      "step": 32525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600909485160404e-05,
+      "loss": 2.6164,
+      "step": 32530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600788078017944e-05,
+      "loss": 2.5789,
+      "step": 32535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600666653179546e-05,
+      "loss": 2.5168,
+      "step": 32540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600545210645675e-05,
+      "loss": 2.4436,
+      "step": 32545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600423750416798e-05,
+      "loss": 2.5571,
+      "step": 32550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600302272493386e-05,
+      "loss": 2.5217,
+      "step": 32555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600180776875902e-05,
+      "loss": 2.5696,
+      "step": 32560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600059263564813e-05,
+      "loss": 2.4946,
+      "step": 32565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59993773256059e-05,
+      "loss": 2.6487,
+      "step": 32570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.599816183863699e-05,
+      "loss": 2.3857,
+      "step": 32575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.599694617474604e-05,
+      "loss": 2.6136,
+      "step": 32580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.599573033393778e-05,
+      "loss": 2.4539,
+      "step": 32585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.599451431621687e-05,
+      "loss": 2.3949,
+      "step": 32590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.599329812158795e-05,
+      "loss": 2.4773,
+      "step": 32595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.599208175005574e-05,
+      "loss": 2.6349,
+      "step": 32600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.599086520162491e-05,
+      "loss": 2.5411,
+      "step": 32605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.598964847630014e-05,
+      "loss": 2.4277,
+      "step": 32610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59884315740861e-05,
+      "loss": 2.5342,
+      "step": 32615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.598721449498746e-05,
+      "loss": 2.5894,
+      "step": 32620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.598599723900894e-05,
+      "loss": 2.6513,
+      "step": 32625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.598477980615519e-05,
+      "loss": 2.7136,
+      "step": 32630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59835621964309e-05,
+      "loss": 2.6896,
+      "step": 32635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.598234440984075e-05,
+      "loss": 2.4221,
+      "step": 32640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.598112644638944e-05,
+      "loss": 2.4949,
+      "step": 32645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.597990830608162e-05,
+      "loss": 2.5023,
+      "step": 32650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5978689988922e-05,
+      "loss": 2.3252,
+      "step": 32655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.597747149491529e-05,
+      "loss": 2.5636,
+      "step": 32660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.597625282406614e-05,
+      "loss": 2.4593,
+      "step": 32665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.597503397637923e-05,
+      "loss": 2.4105,
+      "step": 32670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.597381495185929e-05,
+      "loss": 2.4447,
+      "step": 32675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.597259575051098e-05,
+      "loss": 2.6904,
+      "step": 32680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5971376372339e-05,
+      "loss": 2.537,
+      "step": 32685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.597015681734801e-05,
+      "loss": 2.4153,
+      "step": 32690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596893708554275e-05,
+      "loss": 2.53,
+      "step": 32695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596771717692787e-05,
+      "loss": 2.4994,
+      "step": 32700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596649709150809e-05,
+      "loss": 2.6399,
+      "step": 32705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596527682928808e-05,
+      "loss": 2.4899,
+      "step": 32710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596405639027256e-05,
+      "loss": 2.5215,
+      "step": 32715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596283577446618e-05,
+      "loss": 2.5599,
+      "step": 32720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596161498187369e-05,
+      "loss": 2.6243,
+      "step": 32725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596039401249976e-05,
+      "loss": 2.5778,
+      "step": 32730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595917286634908e-05,
+      "loss": 2.4254,
+      "step": 32735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595795154342634e-05,
+      "loss": 2.447,
+      "step": 32740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595673004373627e-05,
+      "loss": 2.468,
+      "step": 32745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595550836728354e-05,
+      "loss": 2.6479,
+      "step": 32750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595428651407285e-05,
+      "loss": 2.4536,
+      "step": 32755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595306448410892e-05,
+      "loss": 2.4017,
+      "step": 32760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595184227739643e-05,
+      "loss": 2.3995,
+      "step": 32765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59506198939401e-05,
+      "loss": 2.5064,
+      "step": 32770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59493973337446e-05,
+      "loss": 2.4124,
+      "step": 32775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594817459681466e-05,
+      "loss": 2.4363,
+      "step": 32780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594695168315498e-05,
+      "loss": 2.5017,
+      "step": 32785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594572859277024e-05,
+      "loss": 2.5267,
+      "step": 32790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594450532566518e-05,
+      "loss": 2.532,
+      "step": 32795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59432818818445e-05,
+      "loss": 2.4974,
+      "step": 32800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594205826131287e-05,
+      "loss": 2.5699,
+      "step": 32805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594083446407502e-05,
+      "loss": 2.5474,
+      "step": 32810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593961049013568e-05,
+      "loss": 2.5548,
+      "step": 32815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593838633949952e-05,
+      "loss": 2.4929,
+      "step": 32820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593716201217127e-05,
+      "loss": 2.4629,
+      "step": 32825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593593750815562e-05,
+      "loss": 2.4638,
+      "step": 32830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593471282745729e-05,
+      "loss": 2.573,
+      "step": 32835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593348797008101e-05,
+      "loss": 2.5752,
+      "step": 32840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593226293603147e-05,
+      "loss": 2.3576,
+      "step": 32845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593103772531338e-05,
+      "loss": 2.6218,
+      "step": 32850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592981233793146e-05,
+      "loss": 2.5053,
+      "step": 32855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592858677389043e-05,
+      "loss": 2.4996,
+      "step": 32860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592736103319498e-05,
+      "loss": 2.4933,
+      "step": 32865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592613511584984e-05,
+      "loss": 2.5003,
+      "step": 32870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592490902185974e-05,
+      "loss": 2.5529,
+      "step": 32875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592368275122936e-05,
+      "loss": 2.5575,
+      "step": 32880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592245630396345e-05,
+      "loss": 2.5346,
+      "step": 32885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592122968006673e-05,
+      "loss": 2.484,
+      "step": 32890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592000287954388e-05,
+      "loss": 2.54,
+      "step": 32895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591877590239965e-05,
+      "loss": 2.3718,
+      "step": 32900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591754874863876e-05,
+      "loss": 2.6477,
+      "step": 32905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59163214182659e-05,
+      "loss": 2.5531,
+      "step": 32910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591509391128583e-05,
+      "loss": 2.4718,
+      "step": 32915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591386622770324e-05,
+      "loss": 2.5383,
+      "step": 32920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591263836752286e-05,
+      "loss": 2.5342,
+      "step": 32925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591141033074944e-05,
+      "loss": 2.5511,
+      "step": 32930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591018211738765e-05,
+      "loss": 2.5754,
+      "step": 32935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590895372744228e-05,
+      "loss": 2.499,
+      "step": 32940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5907725160918e-05,
+      "loss": 2.7261,
+      "step": 32945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590649641781954e-05,
+      "loss": 2.436,
+      "step": 32950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590526749815168e-05,
+      "loss": 2.4422,
+      "step": 32955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590403840191908e-05,
+      "loss": 2.5071,
+      "step": 32960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59028091291265e-05,
+      "loss": 2.5172,
+      "step": 32965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590157967977867e-05,
+      "loss": 2.3954,
+      "step": 32970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590035005388029e-05,
+      "loss": 2.5182,
+      "step": 32975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589912025143614e-05,
+      "loss": 2.5306,
+      "step": 32980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589789027245091e-05,
+      "loss": 2.5572,
+      "step": 32985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589666011692934e-05,
+      "loss": 2.4914,
+      "step": 32990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589542978487616e-05,
+      "loss": 2.5265,
+      "step": 32995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589419927629612e-05,
+      "loss": 2.5782,
+      "step": 33000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589296859119394e-05,
+      "loss": 2.4341,
+      "step": 33005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589173772957434e-05,
+      "loss": 2.508,
+      "step": 33010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589050669144208e-05,
+      "loss": 2.5866,
+      "step": 33015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588927547680188e-05,
+      "loss": 2.5262,
+      "step": 33020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588804408565846e-05,
+      "loss": 2.5101,
+      "step": 33025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588681251801658e-05,
+      "loss": 2.6936,
+      "step": 33030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5885580773881e-05,
+      "loss": 2.6161,
+      "step": 33035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58843488532564e-05,
+      "loss": 2.3369,
+      "step": 33040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588311675614756e-05,
+      "loss": 2.5334,
+      "step": 33045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588188448255919e-05,
+      "loss": 2.6222,
+      "step": 33050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588065203249605e-05,
+      "loss": 2.4745,
+      "step": 33055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58794194059629e-05,
+      "loss": 2.668,
+      "step": 33060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587818660296444e-05,
+      "loss": 2.4302,
+      "step": 33065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587695362350541e-05,
+      "loss": 2.4449,
+      "step": 33070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587572046759059e-05,
+      "loss": 2.4785,
+      "step": 33075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58744871352247e-05,
+      "loss": 2.487,
+      "step": 33080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587325362641251e-05,
+      "loss": 2.5493,
+      "step": 33085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58720199411587e-05,
+      "loss": 2.4592,
+      "step": 33090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58707860794681e-05,
+      "loss": 2.6218,
+      "step": 33095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586955204134538e-05,
+      "loss": 2.5289,
+      "step": 33100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586831782679534e-05,
+      "loss": 2.4169,
+      "step": 33105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586708343582269e-05,
+      "loss": 2.4728,
+      "step": 33110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586584886843222e-05,
+      "loss": 2.5236,
+      "step": 33115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586461412462864e-05,
+      "loss": 2.4913,
+      "step": 33120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58633792044167e-05,
+      "loss": 2.6203,
+      "step": 33125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586214410780117e-05,
+      "loss": 2.5727,
+      "step": 33130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58609088347868e-05,
+      "loss": 2.4951,
+      "step": 33135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585967338537832e-05,
+      "loss": 2.5226,
+      "step": 33140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585843775958051e-05,
+      "loss": 2.4975,
+      "step": 33145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58572019573981e-05,
+      "loss": 2.5744,
+      "step": 33150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585596597883586e-05,
+      "loss": 2.5294,
+      "step": 33155
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585472982389852e-05,
+      "loss": 2.5122,
+      "step": 33160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585349349259086e-05,
+      "loss": 2.4553,
+      "step": 33165
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585225698491763e-05,
+      "loss": 2.5901,
+      "step": 33170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585102030088357e-05,
+      "loss": 2.6037,
+      "step": 33175
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584978344049347e-05,
+      "loss": 2.5531,
+      "step": 33180
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584854640375205e-05,
+      "loss": 2.4932,
+      "step": 33185
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584730919066407e-05,
+      "loss": 2.501,
+      "step": 33190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584607180123433e-05,
+      "loss": 2.5696,
+      "step": 33195
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584483423546754e-05,
+      "loss": 2.5131,
+      "step": 33200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584359649336849e-05,
+      "loss": 2.5413,
+      "step": 33205
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584235857494192e-05,
+      "loss": 2.4692,
+      "step": 33210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584112048019263e-05,
+      "loss": 2.5598,
+      "step": 33215
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583988220912533e-05,
+      "loss": 2.4635,
+      "step": 33220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583864376174481e-05,
+      "loss": 2.5122,
+      "step": 33225
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583740513805585e-05,
+      "loss": 2.5471,
+      "step": 33230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583616633806318e-05,
+      "loss": 2.6487,
+      "step": 33235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58349273617716e-05,
+      "loss": 2.5559,
+      "step": 33240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583368820918584e-05,
+      "loss": 2.5458,
+      "step": 33245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583244888031066e-05,
+      "loss": 2.4843,
+      "step": 33250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583120937515088e-05,
+      "loss": 2.4596,
+      "step": 33255
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582996969371123e-05,
+      "loss": 2.731,
+      "step": 33260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582872983599649e-05,
+      "loss": 2.4293,
+      "step": 33265
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582748980201141e-05,
+      "loss": 2.4784,
+      "step": 33270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582624959176078e-05,
+      "loss": 2.3965,
+      "step": 33275
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582500920524937e-05,
+      "loss": 2.4616,
+      "step": 33280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582376864248193e-05,
+      "loss": 2.4566,
+      "step": 33285
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582252790346327e-05,
+      "loss": 2.5035,
+      "step": 33290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582128698819812e-05,
+      "loss": 2.6556,
+      "step": 33295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582004589669127e-05,
+      "loss": 2.609,
+      "step": 33300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581880462894749e-05,
+      "loss": 2.5216,
+      "step": 33305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581756318497157e-05,
+      "loss": 2.4292,
+      "step": 33310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581632156476827e-05,
+      "loss": 2.559,
+      "step": 33315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581507976834237e-05,
+      "loss": 2.4361,
+      "step": 33320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581383779569863e-05,
+      "loss": 2.5049,
+      "step": 33325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581259564684186e-05,
+      "loss": 2.6416,
+      "step": 33330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581135332177682e-05,
+      "loss": 2.4276,
+      "step": 33335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581011082050828e-05,
+      "loss": 2.442,
+      "step": 33340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580886814304103e-05,
+      "loss": 2.4109,
+      "step": 33345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580762528937986e-05,
+      "loss": 2.5519,
+      "step": 33350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580638225952951e-05,
+      "loss": 2.4824,
+      "step": 33355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580513905349482e-05,
+      "loss": 2.5456,
+      "step": 33360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580389567128051e-05,
+      "loss": 2.4796,
+      "step": 33365
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580265211289141e-05,
+      "loss": 2.5831,
+      "step": 33370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580140837833229e-05,
+      "loss": 2.4521,
+      "step": 33375
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580016446760792e-05,
+      "loss": 2.5586,
+      "step": 33380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57989203807231e-05,
+      "loss": 2.5247,
+      "step": 33385
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579767611768261e-05,
+      "loss": 2.5198,
+      "step": 33390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579643167849124e-05,
+      "loss": 2.5129,
+      "step": 33395
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579518706315376e-05,
+      "loss": 2.6402,
+      "step": 33400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579394227167496e-05,
+      "loss": 2.5138,
+      "step": 33405
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579269730405965e-05,
+      "loss": 2.5663,
+      "step": 33410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57914521603126e-05,
+      "loss": 2.4811,
+      "step": 33415
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57902068404386e-05,
+      "loss": 2.5166,
+      "step": 33420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578896134444246e-05,
+      "loss": 2.4933,
+      "step": 33425
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578771567232894e-05,
+      "loss": 2.6553,
+      "step": 33430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578646982410286e-05,
+      "loss": 2.5589,
+      "step": 33435
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578522379976899e-05,
+      "loss": 2.5225,
+      "step": 33440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578397759933213e-05,
+      "loss": 2.4933,
+      "step": 33445
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578273122279706e-05,
+      "loss": 2.4345,
+      "step": 33450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57814846701686e-05,
+      "loss": 2.5612,
+      "step": 33455
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578023794145153e-05,
+      "loss": 2.4941,
+      "step": 33460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577899103665066e-05,
+      "loss": 2.6184,
+      "step": 33465
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577774395577075e-05,
+      "loss": 2.4918,
+      "step": 33470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577649669881664e-05,
+      "loss": 2.7023,
+      "step": 33475
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577524926579309e-05,
+      "loss": 2.6777,
+      "step": 33480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577400165670492e-05,
+      "loss": 2.4469,
+      "step": 33485
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577275387155693e-05,
+      "loss": 2.5682,
+      "step": 33490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577150591035391e-05,
+      "loss": 2.5282,
+      "step": 33495
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577025777310068e-05,
+      "loss": 2.4773,
+      "step": 33500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576900945980198e-05,
+      "loss": 2.6085,
+      "step": 33505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576776097046269e-05,
+      "loss": 2.5425,
+      "step": 33510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576651230508757e-05,
+      "loss": 2.551,
+      "step": 33515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576526346368142e-05,
+      "loss": 2.5772,
+      "step": 33520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576401444624907e-05,
+      "loss": 2.6514,
+      "step": 33525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576276525279529e-05,
+      "loss": 2.5656,
+      "step": 33530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576151588332492e-05,
+      "loss": 2.6089,
+      "step": 33535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576026633784274e-05,
+      "loss": 2.6693,
+      "step": 33540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575901661635356e-05,
+      "loss": 2.419,
+      "step": 33545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575776671886219e-05,
+      "loss": 2.5835,
+      "step": 33550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575651664537344e-05,
+      "loss": 2.3831,
+      "step": 33555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57552663958921e-05,
+      "loss": 2.5464,
+      "step": 33560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575401597042301e-05,
+      "loss": 2.4655,
+      "step": 33565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575276536897097e-05,
+      "loss": 2.4508,
+      "step": 33570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575151459154078e-05,
+      "loss": 2.4455,
+      "step": 33575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575026363813726e-05,
+      "loss": 2.5305,
+      "step": 33580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57490125087652e-05,
+      "loss": 2.4678,
+      "step": 33585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574776120342945e-05,
+      "loss": 2.5712,
+      "step": 33590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57465097221348e-05,
+      "loss": 2.5805,
+      "step": 33595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574525806488603e-05,
+      "loss": 2.6661,
+      "step": 33600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574400623168803e-05,
+      "loss": 2.5355,
+      "step": 33605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574275422254556e-05,
+      "loss": 2.5994,
+      "step": 33610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574150203746344e-05,
+      "loss": 2.5134,
+      "step": 33615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574024967644652e-05,
+      "loss": 2.5513,
+      "step": 33620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573899713949959e-05,
+      "loss": 2.4955,
+      "step": 33625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573774442662745e-05,
+      "loss": 2.5997,
+      "step": 33630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573649153783495e-05,
+      "loss": 2.5756,
+      "step": 33635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573523847312691e-05,
+      "loss": 2.6623,
+      "step": 33640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573398523250811e-05,
+      "loss": 2.6193,
+      "step": 33645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573273181598342e-05,
+      "loss": 2.5875,
+      "step": 33650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573147822355763e-05,
+      "loss": 2.4624,
+      "step": 33655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573022445523559e-05,
+      "loss": 2.4945,
+      "step": 33660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572897051102208e-05,
+      "loss": 2.5804,
+      "step": 33665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572771639092195e-05,
+      "loss": 2.4333,
+      "step": 33670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572646209494003e-05,
+      "loss": 2.5159,
+      "step": 33675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572520762308113e-05,
+      "loss": 2.4534,
+      "step": 33680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572395297535009e-05,
+      "loss": 2.6278,
+      "step": 33685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57226981517517e-05,
+      "loss": 2.5813,
+      "step": 33690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572144315229082e-05,
+      "loss": 2.5173,
+      "step": 33695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572018797697228e-05,
+      "loss": 2.5691,
+      "step": 33700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57189326258009e-05,
+      "loss": 2.4948,
+      "step": 33705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571767709878148e-05,
+      "loss": 2.5529,
+      "step": 33710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57164213959189e-05,
+      "loss": 2.5217,
+      "step": 33715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571516551721795e-05,
+      "loss": 2.5755,
+      "step": 33720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571390946268348e-05,
+      "loss": 2.6766,
+      "step": 33725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571265323232031e-05,
+      "loss": 2.4553,
+      "step": 33730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571139682613327e-05,
+      "loss": 2.4489,
+      "step": 33735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571014024412721e-05,
+      "loss": 2.4885,
+      "step": 33740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570888348630695e-05,
+      "loss": 2.5995,
+      "step": 33745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570762655267733e-05,
+      "loss": 2.5969,
+      "step": 33750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570636944324318e-05,
+      "loss": 2.4996,
+      "step": 33755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570511215800933e-05,
+      "loss": 2.4951,
+      "step": 33760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570385469698062e-05,
+      "loss": 2.6612,
+      "step": 33765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57025970601619e-05,
+      "loss": 2.5695,
+      "step": 33770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5701339247558e-05,
+      "loss": 2.4721,
+      "step": 33775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570008125917373e-05,
+      "loss": 2.4822,
+      "step": 33780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569882309501399e-05,
+      "loss": 2.5464,
+      "step": 33785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569756475508356e-05,
+      "loss": 2.5976,
+      "step": 33790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56963062393873e-05,
+      "loss": 2.5114,
+      "step": 33795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569504754793004e-05,
+      "loss": 2.4311,
+      "step": 33800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569378868071666e-05,
+      "loss": 2.504,
+      "step": 33805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569252963775195e-05,
+      "loss": 2.3728,
+      "step": 33810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56912704190408e-05,
+      "loss": 2.429,
+      "step": 33815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569001102458801e-05,
+      "loss": 2.5532,
+      "step": 33820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568875145439846e-05,
+      "loss": 2.5097,
+      "step": 33825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568749170847699e-05,
+      "loss": 2.4556,
+      "step": 33830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568623178682842e-05,
+      "loss": 2.5361,
+      "step": 33835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568497168945762e-05,
+      "loss": 2.619,
+      "step": 33840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568371141636943e-05,
+      "loss": 2.4573,
+      "step": 33845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568245096756868e-05,
+      "loss": 2.5262,
+      "step": 33850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568119034306025e-05,
+      "loss": 2.4824,
+      "step": 33855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567992954284896e-05,
+      "loss": 2.484,
+      "step": 33860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567866856693967e-05,
+      "loss": 2.7021,
+      "step": 33865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567740741533724e-05,
+      "loss": 2.546,
+      "step": 33870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567614608804651e-05,
+      "loss": 2.5501,
+      "step": 33875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567488458507234e-05,
+      "loss": 2.567,
+      "step": 33880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567362290641956e-05,
+      "loss": 2.4709,
+      "step": 33885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567236105209305e-05,
+      "loss": 2.6048,
+      "step": 33890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567109902209765e-05,
+      "loss": 2.3941,
+      "step": 33895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566983681643823e-05,
+      "loss": 2.671,
+      "step": 33900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566857443511962e-05,
+      "loss": 2.4639,
+      "step": 33905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566731187814668e-05,
+      "loss": 2.6829,
+      "step": 33910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566604914552428e-05,
+      "loss": 2.3939,
+      "step": 33915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566478623725728e-05,
+      "loss": 2.6377,
+      "step": 33920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566352315335051e-05,
+      "loss": 2.4911,
+      "step": 33925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566225989380886e-05,
+      "loss": 2.5692,
+      "step": 33930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566099645863715e-05,
+      "loss": 2.5239,
+      "step": 33935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565973284784029e-05,
+      "loss": 2.4658,
+      "step": 33940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565846906142311e-05,
+      "loss": 2.5394,
+      "step": 33945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565720509939047e-05,
+      "loss": 2.5964,
+      "step": 33950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565594096174723e-05,
+      "loss": 2.4361,
+      "step": 33955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565467664849825e-05,
+      "loss": 2.5537,
+      "step": 33960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565341215964843e-05,
+      "loss": 2.4915,
+      "step": 33965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565214749520259e-05,
+      "loss": 2.5972,
+      "step": 33970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56508826551656e-05,
+      "loss": 2.6437,
+      "step": 33975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564961763954234e-05,
+      "loss": 2.5002,
+      "step": 33980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564835244833768e-05,
+      "loss": 2.5168,
+      "step": 33985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564708708155647e-05,
+      "loss": 2.5963,
+      "step": 33990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564582153920358e-05,
+      "loss": 2.7201,
+      "step": 33995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564455582128389e-05,
+      "loss": 2.5335,
+      "step": 34000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564328992780224e-05,
+      "loss": 2.5398,
+      "step": 34005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564202385876355e-05,
+      "loss": 2.5465,
+      "step": 34010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564075761417262e-05,
+      "loss": 2.591,
+      "step": 34015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563949119403439e-05,
+      "loss": 2.4584,
+      "step": 34020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563822459835368e-05,
+      "loss": 2.5353,
+      "step": 34025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563695782713539e-05,
+      "loss": 2.4843,
+      "step": 34030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563569088038437e-05,
+      "loss": 2.5045,
+      "step": 34035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563442375810552e-05,
+      "loss": 2.5586,
+      "step": 34040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56331564603037e-05,
+      "loss": 2.4634,
+      "step": 34045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563188898698379e-05,
+      "loss": 2.5353,
+      "step": 34050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563062133815064e-05,
+      "loss": 2.5312,
+      "step": 34055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.562935351380916e-05,
+      "loss": 2.5898,
+      "step": 34060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.562808551396421e-05,
+      "loss": 2.5885,
+      "step": 34065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.562681733862066e-05,
+      "loss": 2.5378,
+      "step": 34070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56255489877834e-05,
+      "loss": 2.5508,
+      "step": 34075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56242804614573e-05,
+      "loss": 2.472,
+      "step": 34080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.562301175964725e-05,
+      "loss": 2.4383,
+      "step": 34085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.562174288235813e-05,
+      "loss": 2.4766,
+      "step": 34090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56204738295948e-05,
+      "loss": 2.4889,
+      "step": 34095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.561920460136217e-05,
+      "loss": 2.5012,
+      "step": 34100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.561793519766509e-05,
+      "loss": 2.5167,
+      "step": 34105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.561666561850847e-05,
+      "loss": 2.5685,
+      "step": 34110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.561539586389719e-05,
+      "loss": 2.5943,
+      "step": 34115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.561412593383612e-05,
+      "loss": 2.508,
+      "step": 34120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.561285582833015e-05,
+      "loss": 2.4067,
+      "step": 34125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.561158554738417e-05,
+      "loss": 2.5599,
+      "step": 34130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.561031509100304e-05,
+      "loss": 2.3525,
+      "step": 34135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56090444591917e-05,
+      "loss": 2.4476,
+      "step": 34140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5607773651955e-05,
+      "loss": 2.4026,
+      "step": 34145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.560650266929782e-05,
+      "loss": 2.5362,
+      "step": 34150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.560523151122507e-05,
+      "loss": 2.6208,
+      "step": 34155
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.560396017774162e-05,
+      "loss": 2.636,
+      "step": 34160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56026886688524e-05,
+      "loss": 2.6457,
+      "step": 34165
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.560141698456224e-05,
+      "loss": 2.367,
+      "step": 34170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.560014512487608e-05,
+      "loss": 2.5002,
+      "step": 34175
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.55988730897988e-05,
+      "loss": 2.5754,
+      "step": 34180
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.559760087933527e-05,
+      "loss": 2.3752,
+      "step": 34185
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.559632849349042e-05,
+      "loss": 2.47,
+      "step": 34190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.559505593226912e-05,
+      "loss": 2.5284,
+      "step": 34195
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.559378319567627e-05,
+      "loss": 2.4637,
+      "step": 34200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.559251028371676e-05,
+      "loss": 2.5687,
+      "step": 34205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.559123719639549e-05,
+      "loss": 2.426,
+      "step": 34210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558996393371736e-05,
+      "loss": 2.3723,
+      "step": 34215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558869049568726e-05,
+      "loss": 2.4725,
+      "step": 34220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55874168823101e-05,
+      "loss": 2.5996,
+      "step": 34225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558614309359077e-05,
+      "loss": 2.2429,
+      "step": 34230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558486912953417e-05,
+      "loss": 2.523,
+      "step": 34235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55835949901452e-05,
+      "loss": 2.5418,
+      "step": 34240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558232067542878e-05,
+      "loss": 2.4149,
+      "step": 34245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558104618538976e-05,
+      "loss": 2.5665,
+      "step": 34250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55797715200331e-05,
+      "loss": 2.5225,
+      "step": 34255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557849667936367e-05,
+      "loss": 2.4218,
+      "step": 34260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557722166338639e-05,
+      "loss": 2.526,
+      "step": 34265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557594647210615e-05,
+      "loss": 2.3974,
+      "step": 34270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557467110552786e-05,
+      "loss": 2.7069,
+      "step": 34275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557339556365642e-05,
+      "loss": 2.555,
+      "step": 34280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557211984649674e-05,
+      "loss": 2.4549,
+      "step": 34285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557084395405373e-05,
+      "loss": 2.6548,
+      "step": 34290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556956788633231e-05,
+      "loss": 2.5317,
+      "step": 34295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556829164333736e-05,
+      "loss": 2.548,
+      "step": 34300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556701522507381e-05,
+      "loss": 2.7076,
+      "step": 34305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556573863154656e-05,
+      "loss": 2.5798,
+      "step": 34310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556446186276052e-05,
+      "loss": 2.4052,
+      "step": 34315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55631849187206e-05,
+      "loss": 2.6016,
+      "step": 34320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556190779943172e-05,
+      "loss": 2.5009,
+      "step": 34325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556063050489878e-05,
+      "loss": 2.5657,
+      "step": 34330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555935303512671e-05,
+      "loss": 2.6323,
+      "step": 34335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555807539012041e-05,
+      "loss": 2.5474,
+      "step": 34340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55567975698848e-05,
+      "loss": 2.6209,
+      "step": 34345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555551957442479e-05,
+      "loss": 2.5623,
+      "step": 34350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555424140374529e-05,
+      "loss": 2.5608,
+      "step": 34355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555296305785124e-05,
+      "loss": 2.5482,
+      "step": 34360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555168453674752e-05,
+      "loss": 2.4883,
+      "step": 34365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555040584043909e-05,
+      "loss": 2.3594,
+      "step": 34370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554912696893084e-05,
+      "loss": 2.6206,
+      "step": 34375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554784792222768e-05,
+      "loss": 2.4753,
+      "step": 34380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554656870033456e-05,
+      "loss": 2.4399,
+      "step": 34385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55452893032564e-05,
+      "loss": 2.537,
+      "step": 34390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554400973099808e-05,
+      "loss": 2.5919,
+      "step": 34395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554272998356455e-05,
+      "loss": 2.4542,
+      "step": 34400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554145006096073e-05,
+      "loss": 2.451,
+      "step": 34405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554016996319156e-05,
+      "loss": 2.5577,
+      "step": 34410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553888969026195e-05,
+      "loss": 2.3944,
+      "step": 34415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55376092421768e-05,
+      "loss": 2.4447,
+      "step": 34420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553632861894106e-05,
+      "loss": 2.3309,
+      "step": 34425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553504782055966e-05,
+      "loss": 2.5386,
+      "step": 34430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553376684703752e-05,
+      "loss": 2.58,
+      "step": 34435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553248569837957e-05,
+      "loss": 2.4968,
+      "step": 34440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553120437459071e-05,
+      "loss": 2.6145,
+      "step": 34445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55299228756759e-05,
+      "loss": 2.5357,
+      "step": 34450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552864120164007e-05,
+      "loss": 2.5636,
+      "step": 34455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552735935248812e-05,
+      "loss": 2.4766,
+      "step": 34460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552607732822502e-05,
+      "loss": 2.4598,
+      "step": 34465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552479512885568e-05,
+      "loss": 2.4723,
+      "step": 34470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552351275438501e-05,
+      "loss": 2.5095,
+      "step": 34475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552223020481799e-05,
+      "loss": 2.4476,
+      "step": 34480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552094748015952e-05,
+      "loss": 2.5712,
+      "step": 34485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551966458041455e-05,
+      "loss": 2.4456,
+      "step": 34490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551838150558799e-05,
+      "loss": 2.5159,
+      "step": 34495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55170982556848e-05,
+      "loss": 2.5419,
+      "step": 34500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55158148307099e-05,
+      "loss": 2.5473,
+      "step": 34505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551453123066824e-05,
+      "loss": 2.5536,
+      "step": 34510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551324745556475e-05,
+      "loss": 2.6113,
+      "step": 34515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551196350540438e-05,
+      "loss": 2.4781,
+      "step": 34520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551067938019205e-05,
+      "loss": 2.4994,
+      "step": 34525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55093950799327e-05,
+      "loss": 2.4656,
+      "step": 34530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550811060463127e-05,
+      "loss": 2.5464,
+      "step": 34535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550682595429272e-05,
+      "loss": 2.4747,
+      "step": 34540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550554112892197e-05,
+      "loss": 2.4909,
+      "step": 34545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550425612852397e-05,
+      "loss": 2.5324,
+      "step": 34550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550297095310367e-05,
+      "loss": 2.4655,
+      "step": 34555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5501685602666e-05,
+      "loss": 2.5468,
+      "step": 34560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550040007721591e-05,
+      "loss": 2.4563,
+      "step": 34565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549911437675832e-05,
+      "loss": 2.5342,
+      "step": 34570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549782850129823e-05,
+      "loss": 2.4512,
+      "step": 34575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549654245084053e-05,
+      "loss": 2.4509,
+      "step": 34580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549525622539021e-05,
+      "loss": 2.4866,
+      "step": 34585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549396982495217e-05,
+      "loss": 2.5777,
+      "step": 34590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549268324953143e-05,
+      "loss": 2.4159,
+      "step": 34595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549139649913285e-05,
+      "loss": 2.6276,
+      "step": 34600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549010957376145e-05,
+      "loss": 2.5629,
+      "step": 34605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548882247342215e-05,
+      "loss": 2.6124,
+      "step": 34610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54875351981199e-05,
+      "loss": 2.4016,
+      "step": 34615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548624774785966e-05,
+      "loss": 2.5584,
+      "step": 34620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548496012264635e-05,
+      "loss": 2.6858,
+      "step": 34625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548367232248499e-05,
+      "loss": 2.5944,
+      "step": 34630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548238434738047e-05,
+      "loss": 2.4098,
+      "step": 34635
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548109619733778e-05,
+      "loss": 2.5174,
+      "step": 34640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547980787236185e-05,
+      "loss": 2.6559,
+      "step": 34645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547851937245765e-05,
+      "loss": 2.607,
+      "step": 34650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547723069763013e-05,
+      "loss": 2.5934,
+      "step": 34655
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547594184788425e-05,
+      "loss": 2.5142,
+      "step": 34660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547465282322496e-05,
+      "loss": 2.593,
+      "step": 34665
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547336362365724e-05,
+      "loss": 2.53,
+      "step": 34670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547207424918603e-05,
+      "loss": 2.5466,
+      "step": 34675
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547078469981628e-05,
+      "loss": 2.5016,
+      "step": 34680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546949497555298e-05,
+      "loss": 2.4658,
+      "step": 34685
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546820507640105e-05,
+      "loss": 2.4993,
+      "step": 34690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546691500236549e-05,
+      "loss": 2.5982,
+      "step": 34695
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546562475345124e-05,
+      "loss": 2.5294,
+      "step": 34700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546433432966326e-05,
+      "loss": 2.5063,
+      "step": 34705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546304373100655e-05,
+      "loss": 2.6588,
+      "step": 34710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546175295748603e-05,
+      "loss": 2.6176,
+      "step": 34715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546046200910668e-05,
+      "loss": 2.3544,
+      "step": 34720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545917088587347e-05,
+      "loss": 2.602,
+      "step": 34725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545787958779136e-05,
+      "loss": 2.5402,
+      "step": 34730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545658811486531e-05,
+      "loss": 2.4705,
+      "step": 34735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54552964671003e-05,
+      "loss": 2.5518,
+      "step": 34740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54540046445013e-05,
+      "loss": 2.439,
+      "step": 34745
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545271264707328e-05,
+      "loss": 2.4982,
+      "step": 34750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545142047482119e-05,
+      "loss": 2.5065,
+      "step": 34755
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545012812775002e-05,
+      "loss": 2.525,
+      "step": 34760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544883560586473e-05,
+      "loss": 2.4787,
+      "step": 34765
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544754290917028e-05,
+      "loss": 2.5502,
+      "step": 34770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544625003767169e-05,
+      "loss": 2.5972,
+      "step": 34775
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544495699137389e-05,
+      "loss": 2.6562,
+      "step": 34780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544366377028185e-05,
+      "loss": 2.5274,
+      "step": 34785
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544237037440055e-05,
+      "loss": 2.5532,
+      "step": 34790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544107680373499e-05,
+      "loss": 2.3846,
+      "step": 34795
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54397830582901e-05,
+      "loss": 2.5794,
+      "step": 34800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543848913807093e-05,
+      "loss": 2.594,
+      "step": 34805
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543719504308237e-05,
+      "loss": 2.4625,
+      "step": 34810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543590077332944e-05,
+      "loss": 2.5789,
+      "step": 34815
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543460632881713e-05,
+      "loss": 2.5904,
+      "step": 34820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543331170955042e-05,
+      "loss": 2.5434,
+      "step": 34825
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543201691553424e-05,
+      "loss": 2.5078,
+      "step": 34830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543072194677363e-05,
+      "loss": 2.5761,
+      "step": 34835
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542942680327353e-05,
+      "loss": 2.5805,
+      "step": 34840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542813148503896e-05,
+      "loss": 2.4407,
+      "step": 34845
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542683599207485e-05,
+      "loss": 2.5465,
+      "step": 34850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542554032438621e-05,
+      "loss": 2.5657,
+      "step": 34855
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542424448197804e-05,
+      "loss": 2.411,
+      "step": 34860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542294846485532e-05,
+      "loss": 2.4925,
+      "step": 34865
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5421652273023e-05,
+      "loss": 2.5262,
+      "step": 34870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54203559064861e-05,
+      "loss": 2.4889,
+      "step": 34875
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54190593652496e-05,
+      "loss": 2.5169,
+      "step": 34880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541776264931848e-05,
+      "loss": 2.6552,
+      "step": 34885
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541646575869772e-05,
+      "loss": 2.4946,
+      "step": 34890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541516869339234e-05,
+      "loss": 2.5388,
+      "step": 34895
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541387145340728e-05,
+      "loss": 2.5845,
+      "step": 34900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541257403874757e-05,
+      "loss": 2.4161,
+      "step": 34905
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541127644941818e-05,
+      "loss": 2.6053,
+      "step": 34910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540997868542412e-05,
+      "loss": 2.4636,
+      "step": 34915
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540868074677036e-05,
+      "loss": 2.3481,
+      "step": 34920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540738263346192e-05,
+      "loss": 2.596,
+      "step": 34925
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540608434550375e-05,
+      "loss": 2.3919,
+      "step": 34930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540478588290088e-05,
+      "loss": 2.5666,
+      "step": 34935
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54034872456583e-05,
+      "loss": 2.4012,
+      "step": 34940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5402188433781e-05,
+      "loss": 2.6734,
+      "step": 34945
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540088944727395e-05,
+      "loss": 2.4445,
+      "step": 34950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539959028614218e-05,
+      "loss": 2.5387,
+      "step": 34955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539829095039068e-05,
+      "loss": 2.537,
+      "step": 34960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539699144002444e-05,
+      "loss": 2.3768,
+      "step": 34965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539569175504846e-05,
+      "loss": 2.4671,
+      "step": 34970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539439189546775e-05,
+      "loss": 2.4923,
+      "step": 34975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539309186128728e-05,
+      "loss": 2.456,
+      "step": 34980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53917916525121e-05,
+      "loss": 2.6238,
+      "step": 34985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539049126914717e-05,
+      "loss": 2.5484,
+      "step": 34990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538919071119749e-05,
+      "loss": 2.5443,
+      "step": 34995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53878899786681e-05,
+      "loss": 2.4648,
+      "step": 35000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538658907156397e-05,
+      "loss": 2.7069,
+      "step": 35005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538528798989011e-05,
+      "loss": 2.5098,
+      "step": 35010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538398673365152e-05,
+      "loss": 2.4971,
+      "step": 35015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538268530285322e-05,
+      "loss": 2.574,
+      "step": 35020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53813836975002e-05,
+      "loss": 2.5771,
+      "step": 35025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538008191759749e-05,
+      "loss": 2.4779,
+      "step": 35030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537877996315006e-05,
+      "loss": 2.6624,
+      "step": 35035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537747783416295e-05,
+      "loss": 2.418,
+      "step": 35040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537617553064116e-05,
+      "loss": 2.581,
+      "step": 35045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53748730525897e-05,
+      "loss": 2.3577,
+      "step": 35050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537357040001355e-05,
+      "loss": 2.5307,
+      "step": 35055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537226757291777e-05,
+      "loss": 2.4972,
+      "step": 35060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537096457130734e-05,
+      "loss": 2.3487,
+      "step": 35065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536966139518728e-05,
+      "loss": 2.5712,
+      "step": 35070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536835804456261e-05,
+      "loss": 2.4935,
+      "step": 35075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536705451943833e-05,
+      "loss": 2.4857,
+      "step": 35080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536575081981945e-05,
+      "loss": 2.547,
+      "step": 35085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5364446945711e-05,
+      "loss": 2.5176,
+      "step": 35090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536314289711797e-05,
+      "loss": 2.4734,
+      "step": 35095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536183867404541e-05,
+      "loss": 2.3935,
+      "step": 35100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536053427649831e-05,
+      "loss": 2.5126,
+      "step": 35105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53592297044817e-05,
+      "loss": 2.4815,
+      "step": 35110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535792495800062e-05,
+      "loss": 2.556,
+      "step": 35115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535662003706002e-05,
+      "loss": 2.4346,
+      "step": 35120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535531494166499e-05,
+      "loss": 2.5004,
+      "step": 35125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535400967182051e-05,
+      "loss": 2.4231,
+      "step": 35130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535270422753161e-05,
+      "loss": 2.5482,
+      "step": 35135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535139860880333e-05,
+      "loss": 2.6341,
+      "step": 35140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535009281564067e-05,
+      "loss": 2.583,
+      "step": 35145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534878684804865e-05,
+      "loss": 2.3845,
+      "step": 35150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53474807060323e-05,
+      "loss": 2.3674,
+      "step": 35155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534617438959665e-05,
+      "loss": 2.6432,
+      "step": 35160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534486789874673e-05,
+      "loss": 2.5504,
+      "step": 35165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534356123348754e-05,
+      "loss": 2.6201,
+      "step": 35170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534225439382412e-05,
+      "loss": 2.4867,
+      "step": 35175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534094737976149e-05,
+      "loss": 2.6437,
+      "step": 35180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533964019130469e-05,
+      "loss": 2.4942,
+      "step": 35185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533833282845874e-05,
+      "loss": 2.5212,
+      "step": 35190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533702529122868e-05,
+      "loss": 2.4349,
+      "step": 35195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533571757961953e-05,
+      "loss": 2.7052,
+      "step": 35200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53344096936363e-05,
+      "loss": 2.4931,
+      "step": 35205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533310163328404e-05,
+      "loss": 2.6061,
+      "step": 35210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533179339856779e-05,
+      "loss": 2.571,
+      "step": 35215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533048498949257e-05,
+      "loss": 2.3972,
+      "step": 35220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532917640606343e-05,
+      "loss": 2.6071,
+      "step": 35225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532786764828536e-05,
+      "loss": 2.4694,
+      "step": 35230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532655871616344e-05,
+      "loss": 2.5037,
+      "step": 35235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532524960970268e-05,
+      "loss": 2.4982,
+      "step": 35240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532394032890812e-05,
+      "loss": 2.5417,
+      "step": 35245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532263087378482e-05,
+      "loss": 2.5632,
+      "step": 35250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532132124433776e-05,
+      "loss": 2.5801,
+      "step": 35255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532001144057203e-05,
+      "loss": 2.7007,
+      "step": 35260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531870146249265e-05,
+      "loss": 2.4607,
+      "step": 35265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531739131010465e-05,
+      "loss": 2.494,
+      "step": 35270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53160809834131e-05,
+      "loss": 2.4914,
+      "step": 35275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531477048242298e-05,
+      "loss": 2.576,
+      "step": 35280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531345980713939e-05,
+      "loss": 2.5945,
+      "step": 35285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531214895756733e-05,
+      "loss": 2.6224,
+      "step": 35290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531083793371189e-05,
+      "loss": 2.6134,
+      "step": 35295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530952673557805e-05,
+      "loss": 2.4977,
+      "step": 35300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530821536317092e-05,
+      "loss": 2.6083,
+      "step": 35305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530690381649548e-05,
+      "loss": 2.5138,
+      "step": 35310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530559209555683e-05,
+      "loss": 2.5413,
+      "step": 35315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530428020035998e-05,
+      "loss": 2.2901,
+      "step": 35320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530296813090997e-05,
+      "loss": 2.4898,
+      "step": 35325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530165588721187e-05,
+      "loss": 2.582,
+      "step": 35330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530034346927074e-05,
+      "loss": 2.6647,
+      "step": 35335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529903087709158e-05,
+      "loss": 2.4287,
+      "step": 35340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529771811067949e-05,
+      "loss": 2.3678,
+      "step": 35345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529640517003948e-05,
+      "loss": 2.5981,
+      "step": 35350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529509205517662e-05,
+      "loss": 2.4306,
+      "step": 35355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529377876609596e-05,
+      "loss": 2.545,
+      "step": 35360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529246530280255e-05,
+      "loss": 2.4521,
+      "step": 35365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529115166530143e-05,
+      "loss": 2.5659,
+      "step": 35370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.528983785359766e-05,
+      "loss": 2.6603,
+      "step": 35375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52885238676963e-05,
+      "loss": 2.5092,
+      "step": 35380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.528720970760241e-05,
+      "loss": 2.5457,
+      "step": 35385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.528589537332102e-05,
+      "loss": 2.618,
+      "step": 35390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.528458086485721e-05,
+      "loss": 2.4476,
+      "step": 35395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.528326618221601e-05,
+      "loss": 2.5546,
+      "step": 35400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52819513254025e-05,
+      "loss": 2.5444,
+      "step": 35405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.528063629442173e-05,
+      "loss": 2.5386,
+      "step": 35410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.527932108927875e-05,
+      "loss": 2.4539,
+      "step": 35415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.527800570997865e-05,
+      "loss": 2.5256,
+      "step": 35420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.527669015652644e-05,
+      "loss": 2.503,
+      "step": 35425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52753744289272e-05,
+      "loss": 2.4979,
+      "step": 35430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.527405852718601e-05,
+      "loss": 2.4416,
+      "step": 35435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.527274245130793e-05,
+      "loss": 2.5219,
+      "step": 35440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.527142620129798e-05,
+      "loss": 2.4552,
+      "step": 35445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.527010977716128e-05,
+      "loss": 2.5261,
+      "step": 35450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.526879317890284e-05,
+      "loss": 2.3971,
+      "step": 35455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.526747640652776e-05,
+      "loss": 2.4981,
+      "step": 35460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52661594600411e-05,
+      "loss": 2.5027,
+      "step": 35465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.526484233944791e-05,
+      "loss": 2.6225,
+      "step": 35470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.526352504475327e-05,
+      "loss": 2.5742,
+      "step": 35475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.526220757596223e-05,
+      "loss": 2.4136,
+      "step": 35480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.526088993307988e-05,
+      "loss": 2.5028,
+      "step": 35485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.525957211611129e-05,
+      "loss": 2.4463,
+      "step": 35490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.525825412506149e-05,
+      "loss": 2.5293,
+      "step": 35495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.525693595993558e-05,
+      "loss": 2.621,
+      "step": 35500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.525561762073865e-05,
+      "loss": 2.4495,
+      "step": 35505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.525429910747572e-05,
+      "loss": 2.4527,
+      "step": 35510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52529804201519e-05,
+      "loss": 2.527,
+      "step": 35515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.525166155877225e-05,
+      "loss": 2.553,
+      "step": 35520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.525034252334184e-05,
+      "loss": 2.5463,
+      "step": 35525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.524902331386574e-05,
+      "loss": 2.6113,
+      "step": 35530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.524770393034903e-05,
+      "loss": 2.5419,
+      "step": 35535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.524638437279678e-05,
+      "loss": 2.6062,
+      "step": 35540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.524506464121408e-05,
+      "loss": 2.4672,
+      "step": 35545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5243744735606e-05,
+      "loss": 2.5313,
+      "step": 35550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52424246559776e-05,
+      "loss": 2.387,
+      "step": 35555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.524110440233398e-05,
+      "loss": 2.491,
+      "step": 35560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52397839746802e-05,
+      "loss": 2.4869,
+      "step": 35565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523846337302136e-05,
+      "loss": 2.5037,
+      "step": 35570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523714259736252e-05,
+      "loss": 2.4097,
+      "step": 35575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523582164770874e-05,
+      "loss": 2.5647,
+      "step": 35580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523450052406516e-05,
+      "loss": 2.464,
+      "step": 35585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523317922643682e-05,
+      "loss": 2.5309,
+      "step": 35590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523185775482882e-05,
+      "loss": 2.5223,
+      "step": 35595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523053610924621e-05,
+      "loss": 2.5393,
+      "step": 35600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52292142896941e-05,
+      "loss": 2.5199,
+      "step": 35605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52278922961776e-05,
+      "loss": 2.4798,
+      "step": 35610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.522657012870174e-05,
+      "loss": 2.4661,
+      "step": 35615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.522524778727161e-05,
+      "loss": 2.6886,
+      "step": 35620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.522392527189235e-05,
+      "loss": 2.6176,
+      "step": 35625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.522260258256901e-05,
+      "loss": 2.4875,
+      "step": 35630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.522127971930668e-05,
+      "loss": 2.4246,
+      "step": 35635
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.521995668211044e-05,
+      "loss": 2.6126,
+      "step": 35640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52186334709854e-05,
+      "loss": 2.4055,
+      "step": 35645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.521731008593664e-05,
+      "loss": 2.5442,
+      "step": 35650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.521598652696924e-05,
+      "loss": 2.5188,
+      "step": 35655
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.521466279408828e-05,
+      "loss": 2.437,
+      "step": 35660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52133388872989e-05,
+      "loss": 2.4981,
+      "step": 35665
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.521201480660616e-05,
+      "loss": 2.4925,
+      "step": 35670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.521069055201514e-05,
+      "loss": 2.5361,
+      "step": 35675
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.520936612353095e-05,
+      "loss": 2.5125,
+      "step": 35680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.520804152115869e-05,
+      "loss": 2.545,
+      "step": 35685
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.520671674490344e-05,
+      "loss": 2.5374,
+      "step": 35690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52053917947703e-05,
+      "loss": 2.6888,
+      "step": 35695
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.520406667076437e-05,
+      "loss": 2.5848,
+      "step": 35700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.520274137289076e-05,
+      "loss": 2.507,
+      "step": 35705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.520141590115454e-05,
+      "loss": 2.58,
+      "step": 35710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.520009025556082e-05,
+      "loss": 2.4192,
+      "step": 35715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51987644361147e-05,
+      "loss": 2.5014,
+      "step": 35720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.519743844282129e-05,
+      "loss": 2.4284,
+      "step": 35725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.519611227568568e-05,
+      "loss": 2.4656,
+      "step": 35730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.519478593471296e-05,
+      "loss": 2.5737,
+      "step": 35735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.519345941990825e-05,
+      "loss": 2.3939,
+      "step": 35740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.519213273127665e-05,
+      "loss": 2.4621,
+      "step": 35745
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.519080586882324e-05,
+      "loss": 2.5966,
+      "step": 35750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.518947883255314e-05,
+      "loss": 2.604,
+      "step": 35755
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.518815162247148e-05,
+      "loss": 2.4639,
+      "step": 35760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.518682423858332e-05,
+      "loss": 2.4356,
+      "step": 35765
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.518549668089378e-05,
+      "loss": 2.6363,
+      "step": 35770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.518416894940799e-05,
+      "loss": 2.4886,
+      "step": 35775
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.518284104413102e-05,
+      "loss": 2.5516,
+      "step": 35780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5181512965068e-05,
+      "loss": 2.4075,
+      "step": 35785
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.518018471222403e-05,
+      "loss": 2.5333,
+      "step": 35790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.517885628560423e-05,
+      "loss": 2.5219,
+      "step": 35795
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.517752768521371e-05,
+      "loss": 2.5105,
+      "step": 35800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.517619891105756e-05,
+      "loss": 2.5082,
+      "step": 35805
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51748699631409e-05,
+      "loss": 2.5242,
+      "step": 35810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.517354084146885e-05,
+      "loss": 2.5912,
+      "step": 35815
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.517221154604653e-05,
+      "loss": 2.5594,
+      "step": 35820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.517088207687901e-05,
+      "loss": 2.5238,
+      "step": 35825
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.516955243397145e-05,
+      "loss": 2.5448,
+      "step": 35830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.516822261732894e-05,
+      "loss": 2.5214,
+      "step": 35835
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51668926269566e-05,
+      "loss": 2.426,
+      "step": 35840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.516556246285957e-05,
+      "loss": 2.4939,
+      "step": 35845
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.516423212504293e-05,
+      "loss": 2.5371,
+      "step": 35850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51629016135118e-05,
+      "loss": 2.5529,
+      "step": 35855
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.516157092827131e-05,
+      "loss": 2.4347,
+      "step": 35860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51602400693266e-05,
+      "loss": 2.5206,
+      "step": 35865
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.515890903668275e-05,
+      "loss": 2.5217,
+      "step": 35870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.515757783034489e-05,
+      "loss": 2.5084,
+      "step": 35875
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.515624645031816e-05,
+      "loss": 2.6023,
+      "step": 35880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.515491489660764e-05,
+      "loss": 2.481,
+      "step": 35885
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51535831692185e-05,
+      "loss": 2.5969,
+      "step": 35890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.515225126815582e-05,
+      "loss": 2.5565,
+      "step": 35895
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.515091919342475e-05,
+      "loss": 2.5082,
+      "step": 35900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.514958694503043e-05,
+      "loss": 2.5804,
+      "step": 35905
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.514825452297793e-05,
+      "loss": 2.6428,
+      "step": 35910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.514692192727241e-05,
+      "loss": 2.6039,
+      "step": 35915
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5145589157919e-05,
+      "loss": 2.4239,
+      "step": 35920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.514425621492281e-05,
+      "loss": 2.4023,
+      "step": 35925
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.514292309828899e-05,
+      "loss": 2.5398,
+      "step": 35930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.514158980802263e-05,
+      "loss": 2.5603,
+      "step": 35935
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.514025634412888e-05,
+      "loss": 2.4663,
+      "step": 35940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.513892270661288e-05,
+      "loss": 2.7128,
+      "step": 35945
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.513758889547975e-05,
+      "loss": 2.5169,
+      "step": 35950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51362549107346e-05,
+      "loss": 2.5596,
+      "step": 35955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51349207523826e-05,
+      "loss": 2.5635,
+      "step": 35960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.513358642042884e-05,
+      "loss": 2.5715,
+      "step": 35965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51322519148785e-05,
+      "loss": 2.5201,
+      "step": 35970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.513091723573666e-05,
+      "loss": 2.4093,
+      "step": 35975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.512958238300848e-05,
+      "loss": 2.6607,
+      "step": 35980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.512824735669909e-05,
+      "loss": 2.5198,
+      "step": 35985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.512691215681365e-05,
+      "loss": 2.5214,
+      "step": 35990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.512557678335728e-05,
+      "loss": 2.5347,
+      "step": 35995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.512424123633507e-05,
+      "loss": 2.5734,
+      "step": 36000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.512290551575223e-05,
+      "loss": 2.6398,
+      "step": 36005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.512156962161385e-05,
+      "loss": 2.5012,
+      "step": 36010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51202335539251e-05,
+      "loss": 2.5838,
+      "step": 36015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.511889731269108e-05,
+      "loss": 2.4978,
+      "step": 36020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.511756089791697e-05,
+      "loss": 2.6038,
+      "step": 36025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.511622430960788e-05,
+      "loss": 2.5199,
+      "step": 36030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.511488754776896e-05,
+      "loss": 2.6872,
+      "step": 36035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.511355061240537e-05,
+      "loss": 2.5401,
+      "step": 36040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.511221350352221e-05,
+      "loss": 2.4486,
+      "step": 36045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.511087622112467e-05,
+      "loss": 2.5328,
+      "step": 36050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510953876521786e-05,
+      "loss": 2.473,
+      "step": 36055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510820113580694e-05,
+      "loss": 2.4063,
+      "step": 36060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510686333289707e-05,
+      "loss": 2.6282,
+      "step": 36065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510552535649335e-05,
+      "loss": 2.5092,
+      "step": 36070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510418720660097e-05,
+      "loss": 2.4837,
+      "step": 36075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510284888322505e-05,
+      "loss": 2.5705,
+      "step": 36080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510151038637076e-05,
+      "loss": 2.5578,
+      "step": 36085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510017171604323e-05,
+      "loss": 2.4532,
+      "step": 36090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.509883287224761e-05,
+      "loss": 2.5394,
+      "step": 36095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.509749385498906e-05,
+      "loss": 2.5453,
+      "step": 36100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.509615466427273e-05,
+      "loss": 2.5714,
+      "step": 36105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.509481530010375e-05,
+      "loss": 2.6217,
+      "step": 36110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.50934757624873e-05,
+      "loss": 2.6101,
+      "step": 36115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.509213605142852e-05,
+      "loss": 2.3929,
+      "step": 36120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.509079616693257e-05,
+      "loss": 2.5262,
+      "step": 36125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508945610900458e-05,
+      "loss": 2.5445,
+      "step": 36130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508811587764973e-05,
+      "loss": 2.4057,
+      "step": 36135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508677547287316e-05,
+      "loss": 2.5214,
+      "step": 36140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508543489468003e-05,
+      "loss": 2.459,
+      "step": 36145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508409414307551e-05,
+      "loss": 2.3755,
+      "step": 36150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508275321806474e-05,
+      "loss": 2.5164,
+      "step": 36155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508141211965286e-05,
+      "loss": 2.4316,
+      "step": 36160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508007084784506e-05,
+      "loss": 2.5545,
+      "step": 36165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.507872940264652e-05,
+      "loss": 2.3799,
+      "step": 36170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.507738778406235e-05,
+      "loss": 2.5556,
+      "step": 36175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.507604599209771e-05,
+      "loss": 2.5996,
+      "step": 36180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.507470402675778e-05,
+      "loss": 2.6529,
+      "step": 36185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.507336188804774e-05,
+      "loss": 2.5249,
+      "step": 36190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.507201957597273e-05,
+      "loss": 2.5549,
+      "step": 36195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.507067709053792e-05,
+      "loss": 2.7389,
+      "step": 36200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.506933443174846e-05,
+      "loss": 2.585,
+      "step": 36205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.506799159960953e-05,
+      "loss": 2.5029,
+      "step": 36210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.506664859412628e-05,
+      "loss": 2.606,
+      "step": 36215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.50653054153039e-05,
+      "loss": 2.5725,
+      "step": 36220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.506396206314751e-05,
+      "loss": 2.4694,
+      "step": 36225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.506261853766234e-05,
+      "loss": 2.5759,
+      "step": 36230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.506127483885352e-05,
+      "loss": 2.4366,
+      "step": 36235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505993096672621e-05,
+      "loss": 2.5532,
+      "step": 36240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.50585869212856e-05,
+      "loss": 2.6383,
+      "step": 36245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505724270253687e-05,
+      "loss": 2.381,
+      "step": 36250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505589831048514e-05,
+      "loss": 2.5466,
+      "step": 36255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505455374513563e-05,
+      "loss": 2.5005,
+      "step": 36260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505320900649351e-05,
+      "loss": 2.3867,
+      "step": 36265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505186409456391e-05,
+      "loss": 2.4043,
+      "step": 36270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505051900935205e-05,
+      "loss": 2.4527,
+      "step": 36275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.504917375086307e-05,
+      "loss": 2.7119,
+      "step": 36280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.504782831910215e-05,
+      "loss": 2.5346,
+      "step": 36285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.50464827140745e-05,
+      "loss": 2.4854,
+      "step": 36290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.504513693578526e-05,
+      "loss": 2.5236,
+      "step": 36295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.504379098423959e-05,
+      "loss": 2.435,
+      "step": 36300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.504244485944272e-05,
+      "loss": 2.4579,
+      "step": 36305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.504109856139978e-05,
+      "loss": 2.5156,
+      "step": 36310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.503975209011597e-05,
+      "loss": 2.4475,
+      "step": 36315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.503840544559648e-05,
+      "loss": 2.58,
+      "step": 36320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.503705862784647e-05,
+      "loss": 2.3576,
+      "step": 36325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.503571163687112e-05,
+      "loss": 2.6467,
+      "step": 36330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.503436447267562e-05,
+      "loss": 2.4774,
+      "step": 36335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.503301713526515e-05,
+      "loss": 2.6066,
+      "step": 36340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.50316696246449e-05,
+      "loss": 2.5801,
+      "step": 36345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.503032194082002e-05,
+      "loss": 2.6215,
+      "step": 36350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.502897408379575e-05,
+      "loss": 2.5461,
+      "step": 36355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.50276260535772e-05,
+      "loss": 2.6598,
+      "step": 36360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.502627785016962e-05,
+      "loss": 2.5912,
+      "step": 36365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.502492947357817e-05,
+      "loss": 2.5394,
+      "step": 36370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.502358092380804e-05,
+      "loss": 2.5505,
+      "step": 36375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.502223220086441e-05,
+      "loss": 2.448,
+      "step": 36380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.50208833047525e-05,
+      "loss": 2.4772,
+      "step": 36385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501953423547744e-05,
+      "loss": 2.5276,
+      "step": 36390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501818499304446e-05,
+      "loss": 2.5389,
+      "step": 36395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501683557745873e-05,
+      "loss": 2.5256,
+      "step": 36400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501548598872546e-05,
+      "loss": 2.494,
+      "step": 36405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501413622684982e-05,
+      "loss": 2.5917,
+      "step": 36410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501278629183702e-05,
+      "loss": 2.5164,
+      "step": 36415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501143618369225e-05,
+      "loss": 2.5096,
+      "step": 36420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501008590242069e-05,
+      "loss": 2.4542,
+      "step": 36425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.500873544802755e-05,
+      "loss": 2.4341,
+      "step": 36430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5007384820518e-05,
+      "loss": 2.6332,
+      "step": 36435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.500603401989726e-05,
+      "loss": 2.435,
+      "step": 36440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.500468304617051e-05,
+      "loss": 2.3664,
+      "step": 36445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.500333189934297e-05,
+      "loss": 2.4144,
+      "step": 36450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.500198057941978e-05,
+      "loss": 2.3938,
+      "step": 36455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.500062908640621e-05,
+      "loss": 2.5468,
+      "step": 36460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49992774203074e-05,
+      "loss": 2.5018,
+      "step": 36465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49979255811286e-05,
+      "loss": 2.4551,
+      "step": 36470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.499657356887497e-05,
+      "loss": 2.4994,
+      "step": 36475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.499522138355173e-05,
+      "loss": 2.5162,
+      "step": 36480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.499386902516405e-05,
+      "loss": 2.601,
+      "step": 36485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.499251649371718e-05,
+      "loss": 2.6464,
+      "step": 36490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.499116378921629e-05,
+      "loss": 2.5516,
+      "step": 36495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.498981091166659e-05,
+      "loss": 2.5597,
+      "step": 36500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49884578610733e-05,
+      "loss": 2.4987,
+      "step": 36505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.498710463744158e-05,
+      "loss": 2.4147,
+      "step": 36510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.498575124077667e-05,
+      "loss": 2.5002,
+      "step": 36515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.498439767108378e-05,
+      "loss": 2.5771,
+      "step": 36520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49830439283681e-05,
+      "loss": 2.5664,
+      "step": 36525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.498169001263484e-05,
+      "loss": 2.4981,
+      "step": 36530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.498033592388923e-05,
+      "loss": 2.4366,
+      "step": 36535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.497898166213644e-05,
+      "loss": 2.4167,
+      "step": 36540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.497762722738168e-05,
+      "loss": 2.6866,
+      "step": 36545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49762726196302e-05,
+      "loss": 2.4653,
+      "step": 36550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.497491783888718e-05,
+      "loss": 2.5107,
+      "step": 36555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.497356288515784e-05,
+      "loss": 2.5848,
+      "step": 36560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.497220775844738e-05,
+      "loss": 2.5121,
+      "step": 36565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.497085245876103e-05,
+      "loss": 2.524,
+      "step": 36570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.4969496986104e-05,
+      "loss": 2.3889,
+      "step": 36575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.496814134048148e-05,
+      "loss": 2.5307,
+      "step": 36580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49667855218987e-05,
+      "loss": 2.358,
+      "step": 36585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.496542953036089e-05,
+      "loss": 2.5244,
+      "step": 36590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.496407336587325e-05,
+      "loss": 2.596,
+      "step": 36595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.496271702844099e-05,
+      "loss": 2.356,
+      "step": 36600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.496136051806934e-05,
+      "loss": 2.5973,
+      "step": 36605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49600038347635e-05,
+      "loss": 2.5834,
+      "step": 36610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.495864697852872e-05,
+      "loss": 2.4947,
+      "step": 36615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49572899493702e-05,
+      "loss": 2.5952,
+      "step": 36620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.495593274729316e-05,
+      "loss": 2.5686,
+      "step": 36625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.495457537230281e-05,
+      "loss": 2.4983,
+      "step": 36630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.495321782440437e-05,
+      "loss": 2.5527,
+      "step": 36635
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.495186010360308e-05,
+      "loss": 2.5991,
+      "step": 36640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.495050220990416e-05,
+      "loss": 2.5564,
+      "step": 36645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.494914414331283e-05,
+      "loss": 2.5889,
+      "step": 36650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49477859038343e-05,
+      "loss": 2.4665,
+      "step": 36655
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.494642749147382e-05,
+      "loss": 2.5849,
+      "step": 36660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.494506890623657e-05,
+      "loss": 2.3493,
+      "step": 36665
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.494371014812783e-05,
+      "loss": 2.5992,
+      "step": 36670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.494235121715279e-05,
+      "loss": 2.556,
+      "step": 36675
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.494099211331668e-05,
+      "loss": 2.474,
+      "step": 36680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.493963283662476e-05,
+      "loss": 2.4379,
+      "step": 36685
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.493827338708221e-05,
+      "loss": 2.5895,
+      "step": 36690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49369137646943e-05,
+      "loss": 2.2309,
+      "step": 36695
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.493555396946623e-05,
+      "loss": 2.4927,
+      "step": 36700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.493419400140325e-05,
+      "loss": 2.4011,
+      "step": 36705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.493283386051057e-05,
+      "loss": 2.5354,
+      "step": 36710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.493147354679344e-05,
+      "loss": 2.6104,
+      "step": 36715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.49301130602571e-05,
+      "loss": 2.3456,
+      "step": 36720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.492875240090675e-05,
+      "loss": 2.545,
+      "step": 36725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.492739156874765e-05,
+      "loss": 2.5321,
+      "step": 36730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.492603056378502e-05,
+      "loss": 2.4924,
+      "step": 36735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.492466938602411e-05,
+      "loss": 2.6672,
+      "step": 36740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.492330803547015e-05,
+      "loss": 2.5396,
+      "step": 36745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.492194651212837e-05,
+      "loss": 2.6559,
+      "step": 36750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.4920584816004e-05,
+      "loss": 2.4307,
+      "step": 36755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.49192229471023e-05,
+      "loss": 2.6012,
+      "step": 36760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.491786090542849e-05,
+      "loss": 2.5425,
+      "step": 36765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.491649869098782e-05,
+      "loss": 2.5791,
+      "step": 36770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.491513630378551e-05,
+      "loss": 2.4837,
+      "step": 36775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.491377374382683e-05,
+      "loss": 2.471,
+      "step": 36780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.4912411011117e-05,
+      "loss": 2.458,
+      "step": 36785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.491104810566126e-05,
+      "loss": 2.6022,
+      "step": 36790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490968502746487e-05,
+      "loss": 2.6189,
+      "step": 36795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490832177653305e-05,
+      "loss": 2.6649,
+      "step": 36800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490695835287106e-05,
+      "loss": 2.4454,
+      "step": 36805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490559475648413e-05,
+      "loss": 2.5454,
+      "step": 36810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490423098737753e-05,
+      "loss": 2.5552,
+      "step": 36815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490286704555647e-05,
+      "loss": 2.5015,
+      "step": 36820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490150293102622e-05,
+      "loss": 2.5186,
+      "step": 36825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490013864379203e-05,
+      "loss": 2.5331,
+      "step": 36830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.489877418385914e-05,
+      "loss": 2.3853,
+      "step": 36835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.489740955123278e-05,
+      "loss": 2.54,
+      "step": 36840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.489604474591824e-05,
+      "loss": 2.5899,
+      "step": 36845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.489467976792073e-05,
+      "loss": 2.5744,
+      "step": 36850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.489331461724553e-05,
+      "loss": 2.3278,
+      "step": 36855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.489194929389787e-05,
+      "loss": 2.5741,
+      "step": 36860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.489058379788301e-05,
+      "loss": 2.5085,
+      "step": 36865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.488921812920618e-05,
+      "loss": 2.6005,
+      "step": 36870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.488785228787267e-05,
+      "loss": 2.6017,
+      "step": 36875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.488648627388772e-05,
+      "loss": 2.5686,
+      "step": 36880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.488512008725657e-05,
+      "loss": 2.5759,
+      "step": 36885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.488375372798448e-05,
+      "loss": 2.6033,
+      "step": 36890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.488238719607672e-05,
+      "loss": 2.4678,
+      "step": 36895
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.488102049153854e-05,
+      "loss": 2.5195,
+      "step": 36900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487965361437519e-05,
+      "loss": 2.5425,
+      "step": 36905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487828656459192e-05,
+      "loss": 2.4433,
+      "step": 36910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487691934219401e-05,
+      "loss": 2.5948,
+      "step": 36915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487555194718672e-05,
+      "loss": 2.4663,
+      "step": 36920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487418437957527e-05,
+      "loss": 2.4944,
+      "step": 36925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487281663936496e-05,
+      "loss": 2.4333,
+      "step": 36930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487144872656103e-05,
+      "loss": 2.6716,
+      "step": 36935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487008064116876e-05,
+      "loss": 2.4812,
+      "step": 36940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48687123831934e-05,
+      "loss": 2.6673,
+      "step": 36945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48673439526402e-05,
+      "loss": 2.4644,
+      "step": 36950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.486597534951445e-05,
+      "loss": 2.4578,
+      "step": 36955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48646065738214e-05,
+      "loss": 2.5983,
+      "step": 36960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.486323762556631e-05,
+      "loss": 2.5245,
+      "step": 36965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.486186850475444e-05,
+      "loss": 2.5731,
+      "step": 36970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.486049921139108e-05,
+      "loss": 2.5195,
+      "step": 36975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.485912974548148e-05,
+      "loss": 2.5237,
+      "step": 36980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.485776010703092e-05,
+      "loss": 2.4907,
+      "step": 36985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.485639029604466e-05,
+      "loss": 2.5893,
+      "step": 36990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.485502031252796e-05,
+      "loss": 2.5806,
+      "step": 36995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.485365015648611e-05,
+      "loss": 2.5044,
+      "step": 37000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.485227982792435e-05,
+      "loss": 2.4701,
+      "step": 37005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.485090932684797e-05,
+      "loss": 2.5605,
+      "step": 37010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.484953865326225e-05,
+      "loss": 2.4942,
+      "step": 37015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.484816780717244e-05,
+      "loss": 2.4457,
+      "step": 37020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.484679678858384e-05,
+      "loss": 2.5293,
+      "step": 37025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.484542559750169e-05,
+      "loss": 2.4397,
+      "step": 37030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.484405423393127e-05,
+      "loss": 2.5133,
+      "step": 37035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48426826978779e-05,
+      "loss": 2.5006,
+      "step": 37040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48413109893468e-05,
+      "loss": 2.6177,
+      "step": 37045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483993910834327e-05,
+      "loss": 2.5165,
+      "step": 37050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483856705487258e-05,
+      "loss": 2.4365,
+      "step": 37055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483719482894002e-05,
+      "loss": 2.4494,
+      "step": 37060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483582243055086e-05,
+      "loss": 2.4959,
+      "step": 37065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483444985971038e-05,
+      "loss": 2.443,
+      "step": 37070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483307711642385e-05,
+      "loss": 2.4588,
+      "step": 37075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483170420069655e-05,
+      "loss": 2.5718,
+      "step": 37080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483033111253377e-05,
+      "loss": 2.4205,
+      "step": 37085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48289578519408e-05,
+      "loss": 2.4688,
+      "step": 37090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48275844189229e-05,
+      "loss": 2.667,
+      "step": 37095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.482621081348538e-05,
+      "loss": 2.3519,
+      "step": 37100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.482483703563347e-05,
+      "loss": 2.5479,
+      "step": 37105
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.482346308537254e-05,
+      "loss": 2.5022,
+      "step": 37110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.482208896270779e-05,
+      "loss": 2.455,
+      "step": 37115
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.482071466764455e-05,
+      "loss": 2.5493,
+      "step": 37120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48193402001881e-05,
+      "loss": 2.5525,
+      "step": 37125
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.481796556034371e-05,
+      "loss": 2.5651,
+      "step": 37130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.481659074811669e-05,
+      "loss": 2.508,
+      "step": 37135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.481521576351233e-05,
+      "loss": 2.4558,
+      "step": 37140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.481384060653589e-05,
+      "loss": 2.6004,
+      "step": 37145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.481246527719267e-05,
+      "loss": 2.6174,
+      "step": 37150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.481108977548799e-05,
+      "loss": 2.4571,
+      "step": 37155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48097141014271e-05,
+      "loss": 2.5923,
+      "step": 37160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.480833825501531e-05,
+      "loss": 2.4686,
+      "step": 37165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.480696223625792e-05,
+      "loss": 2.5612,
+      "step": 37170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48055860451602e-05,
+      "loss": 2.5298,
+      "step": 37175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.480420968172745e-05,
+      "loss": 2.559,
+      "step": 37180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.480283314596498e-05,
+      "loss": 2.4677,
+      "step": 37185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.480145643787808e-05,
+      "loss": 2.5708,
+      "step": 37190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.480007955747203e-05,
+      "loss": 2.5817,
+      "step": 37195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.479870250475213e-05,
+      "loss": 2.6141,
+      "step": 37200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47973252797237e-05,
+      "loss": 2.4409,
+      "step": 37205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.4795947882392e-05,
+      "loss": 2.4639,
+      "step": 37210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.479457031276235e-05,
+      "loss": 2.5167,
+      "step": 37215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.479319257084005e-05,
+      "loss": 2.4972,
+      "step": 37220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.479181465663038e-05,
+      "loss": 2.4358,
+      "step": 37225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.479043657013869e-05,
+      "loss": 2.4498,
+      "step": 37230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.478905831137022e-05,
+      "loss": 2.4437,
+      "step": 37235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.478767988033028e-05,
+      "loss": 2.6784,
+      "step": 37240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.478630127702421e-05,
+      "loss": 2.4562,
+      "step": 37245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.478492250145728e-05,
+      "loss": 2.5402,
+      "step": 37250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.478354355363482e-05,
+      "loss": 2.5099,
+      "step": 37255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.478216443356209e-05,
+      "loss": 2.454,
+      "step": 37260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.478078514124444e-05,
+      "loss": 2.4862,
+      "step": 37265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.477940567668716e-05,
+      "loss": 2.6261,
+      "step": 37270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.477802603989553e-05,
+      "loss": 2.3736,
+      "step": 37275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.477664623087491e-05,
+      "loss": 2.5056,
+      "step": 37280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.477526624963055e-05,
+      "loss": 2.5638,
+      "step": 37285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.477388609616779e-05,
+      "loss": 2.462,
+      "step": 37290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.477250577049195e-05,
+      "loss": 2.5359,
+      "step": 37295
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.477112527260832e-05,
+      "loss": 2.5411,
+      "step": 37300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.476974460252219e-05,
+      "loss": 2.5577,
+      "step": 37305
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47683637602389e-05,
+      "loss": 2.5347,
+      "step": 37310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.476698274576376e-05,
+      "loss": 2.5256,
+      "step": 37315
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.476560155910207e-05,
+      "loss": 2.448,
+      "step": 37320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.476422020025915e-05,
+      "loss": 2.4249,
+      "step": 37325
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.476283866924031e-05,
+      "loss": 2.4402,
+      "step": 37330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.476145696605085e-05,
+      "loss": 2.5188,
+      "step": 37335
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47600750906961e-05,
+      "loss": 2.6209,
+      "step": 37340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47586930431814e-05,
+      "loss": 2.5184,
+      "step": 37345
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.475731082351203e-05,
+      "loss": 2.4156,
+      "step": 37350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47559284316933e-05,
+      "loss": 2.3647,
+      "step": 37355
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.475454586773056e-05,
+      "loss": 2.6102,
+      "step": 37360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47531631316291e-05,
+      "loss": 2.511,
+      "step": 37365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.475178022339424e-05,
+      "loss": 2.3647,
+      "step": 37370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.475039714303131e-05,
+      "loss": 2.5506,
+      "step": 37375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.474901389054565e-05,
+      "loss": 2.5644,
+      "step": 37380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.474763046594253e-05,
+      "loss": 2.4151,
+      "step": 37385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.474624686922733e-05,
+      "loss": 2.485,
+      "step": 37390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.474486310040531e-05,
+      "loss": 2.5126,
+      "step": 37395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.474347915948184e-05,
+      "loss": 2.5172,
+      "step": 37400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.474209504646223e-05,
+      "loss": 2.4479,
+      "step": 37405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.474071076135178e-05,
+      "loss": 2.5511,
+      "step": 37410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.473932630415586e-05,
+      "loss": 2.5501,
+      "step": 37415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.473794167487975e-05,
+      "loss": 2.3545,
+      "step": 37420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47365568735288e-05,
+      "loss": 2.5372,
+      "step": 37425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.473517190010833e-05,
+      "loss": 2.4784,
+      "step": 37430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.473378675462368e-05,
+      "loss": 2.3803,
+      "step": 37435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.473240143708015e-05,
+      "loss": 2.4238,
+      "step": 37440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.473101594748309e-05,
+      "loss": 2.5157,
+      "step": 37445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.472963028583784e-05,
+      "loss": 2.3616,
+      "step": 37450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.472824445214969e-05,
+      "loss": 2.3949,
+      "step": 37455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.4726858446424e-05,
+      "loss": 2.4412,
+      "step": 37460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.472547226866608e-05,
+      "loss": 2.4604,
+      "step": 37465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47240859188813e-05,
+      "loss": 2.5384,
+      "step": 37470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.472269939707496e-05,
+      "loss": 2.4522,
+      "step": 37475
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47213127032524e-05,
+      "loss": 2.474,
+      "step": 37480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.471992583741895e-05,
+      "loss": 2.5622,
+      "step": 37485
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.471853879957995e-05,
+      "loss": 2.5014,
+      "step": 37490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.471715158974074e-05,
+      "loss": 2.4452,
+      "step": 37495
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.471576420790665e-05,
+      "loss": 2.5459,
+      "step": 37500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.4714376654083e-05,
+      "loss": 2.4073,
+      "step": 37505
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.471298892827516e-05,
+      "loss": 2.552,
+      "step": 37510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.471160103048845e-05,
+      "loss": 2.3327,
+      "step": 37515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47102129607282e-05,
+      "loss": 2.6275,
+      "step": 37520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.470882471899976e-05,
+      "loss": 2.6339,
+      "step": 37525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.470743630530848e-05,
+      "loss": 2.4518,
+      "step": 37530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.470604771965967e-05,
+      "loss": 2.4371,
+      "step": 37535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.470465896205869e-05,
+      "loss": 2.3752,
+      "step": 37540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.470327003251088e-05,
+      "loss": 2.5746,
+      "step": 37545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47018809310216e-05,
+      "loss": 2.5628,
+      "step": 37550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.470049165759614e-05,
+      "loss": 2.4994,
+      "step": 37555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46991022122399e-05,
+      "loss": 2.4658,
+      "step": 37560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.469771259495821e-05,
+      "loss": 2.5626,
+      "step": 37565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46963228057564e-05,
+      "loss": 2.3715,
+      "step": 37570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.469493284463982e-05,
+      "loss": 2.4334,
+      "step": 37575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.469354271161382e-05,
+      "loss": 2.5374,
+      "step": 37580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.469215240668374e-05,
+      "loss": 2.5169,
+      "step": 37585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.469076192985493e-05,
+      "loss": 2.5793,
+      "step": 37590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.468937128113275e-05,
+      "loss": 2.4485,
+      "step": 37595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.468798046052255e-05,
+      "loss": 2.3833,
+      "step": 37600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.468658946802967e-05,
+      "loss": 2.4993,
+      "step": 37605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.468519830365946e-05,
+      "loss": 2.6259,
+      "step": 37610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.468380696741726e-05,
+      "loss": 2.5579,
+      "step": 37615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.468241545930842e-05,
+      "loss": 2.5133,
+      "step": 37620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.468102377933832e-05,
+      "loss": 2.6457,
+      "step": 37625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46796319275123e-05,
+      "loss": 2.486,
+      "step": 37630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46782399038357e-05,
+      "loss": 2.4616,
+      "step": 37635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46768477083139e-05,
+      "loss": 2.5381,
+      "step": 37640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.467545534095224e-05,
+      "loss": 2.5834,
+      "step": 37645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.467406280175606e-05,
+      "loss": 2.5874,
+      "step": 37650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.467267009073072e-05,
+      "loss": 2.5151,
+      "step": 37655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.467127720788162e-05,
+      "loss": 2.4178,
+      "step": 37660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.466988415321407e-05,
+      "loss": 2.4944,
+      "step": 37665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.466849092673344e-05,
+      "loss": 2.5094,
+      "step": 37670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.466709752844511e-05,
+      "loss": 2.5558,
+      "step": 37675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.466570395835441e-05,
+      "loss": 2.4359,
+      "step": 37680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.466431021646672e-05,
+      "loss": 2.6221,
+      "step": 37685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.466291630278737e-05,
+      "loss": 2.4387,
+      "step": 37690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.466152221732176e-05,
+      "loss": 2.5749,
+      "step": 37695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.466012796007524e-05,
+      "loss": 2.5507,
+      "step": 37700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.465873353105315e-05,
+      "loss": 2.493,
+      "step": 37705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.465733893026089e-05,
+      "loss": 2.5677,
+      "step": 37710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46559441577038e-05,
+      "loss": 2.5679,
+      "step": 37715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.465454921338724e-05,
+      "loss": 2.6194,
+      "step": 37720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.465315409731662e-05,
+      "loss": 2.493,
+      "step": 37725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.465175880949724e-05,
+      "loss": 2.522,
+      "step": 37730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46503633499345e-05,
+      "loss": 2.5559,
+      "step": 37735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.464896771863378e-05,
+      "loss": 2.5565,
+      "step": 37740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.464757191560043e-05,
+      "loss": 2.5859,
+      "step": 37745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46461759408398e-05,
+      "loss": 2.5661,
+      "step": 37750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.464477979435731e-05,
+      "loss": 2.5596,
+      "step": 37755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46433834761583e-05,
+      "loss": 2.3727,
+      "step": 37760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.464198698624813e-05,
+      "loss": 2.6546,
+      "step": 37765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46405903246322e-05,
+      "loss": 2.4954,
+      "step": 37770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.463919349131585e-05,
+      "loss": 2.3954,
+      "step": 37775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.463779648630448e-05,
+      "loss": 2.5085,
+      "step": 37780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.463639930960344e-05,
+      "loss": 2.6021,
+      "step": 37785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.463500196121813e-05,
+      "loss": 2.5513,
+      "step": 37790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46336044411539e-05,
+      "loss": 2.3803,
+      "step": 37795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.463220674941613e-05,
+      "loss": 2.4662,
+      "step": 37800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.463080888601021e-05,
+      "loss": 2.5271,
+      "step": 37805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.462941085094152e-05,
+      "loss": 2.4556,
+      "step": 37810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.462801264421542e-05,
+      "loss": 2.5366,
+      "step": 37815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.462661426583728e-05,
+      "loss": 2.5231,
+      "step": 37820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.462521571581249e-05,
+      "loss": 2.4651,
+      "step": 37825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.462381699414644e-05,
+      "loss": 2.4392,
+      "step": 37830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46224181008445e-05,
+      "loss": 2.489,
+      "step": 37835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.462101903591205e-05,
+      "loss": 2.5294,
+      "step": 37840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.461961979935447e-05,
+      "loss": 2.4862,
+      "step": 37845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.461822039117716e-05,
+      "loss": 2.4525,
+      "step": 37850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.461682081138547e-05,
+      "loss": 2.4114,
+      "step": 37855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.461542105998481e-05,
+      "loss": 2.5375,
+      "step": 37860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.461402113698055e-05,
+      "loss": 2.4621,
+      "step": 37865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.461262104237807e-05,
+      "loss": 2.5595,
+      "step": 37870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.461122077618278e-05,
+      "loss": 2.4211,
+      "step": 37875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.460982033840005e-05,
+      "loss": 2.4593,
+      "step": 37880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.460841972903525e-05,
+      "loss": 2.4246,
+      "step": 37885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.460701894809378e-05,
+      "loss": 2.6362,
+      "step": 37890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.460561799558105e-05,
+      "loss": 2.5713,
+      "step": 37895
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.460421687150242e-05,
+      "loss": 2.4075,
+      "step": 37900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.460281557586329e-05,
+      "loss": 2.5647,
+      "step": 37905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.460141410866905e-05,
+      "loss": 2.5511,
+      "step": 37910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46000124699251e-05,
+      "loss": 2.437,
+      "step": 37915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45986106596368e-05,
+      "loss": 2.5362,
+      "step": 37920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.459720867780957e-05,
+      "loss": 2.6318,
+      "step": 37925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.459580652444879e-05,
+      "loss": 2.297,
+      "step": 37930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.459440419955985e-05,
+      "loss": 2.4452,
+      "step": 37935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.459300170314815e-05,
+      "loss": 2.586,
+      "step": 37940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.459159903521911e-05,
+      "loss": 2.6054,
+      "step": 37945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.459019619577809e-05,
+      "loss": 2.492,
+      "step": 37950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.458879318483049e-05,
+      "loss": 2.3581,
+      "step": 37955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45873900023817e-05,
+      "loss": 2.6251,
+      "step": 37960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.458598664843715e-05,
+      "loss": 2.5323,
+      "step": 37965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.458458312300221e-05,
+      "loss": 2.488,
+      "step": 37970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45831794260823e-05,
+      "loss": 2.5343,
+      "step": 37975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.458177555768278e-05,
+      "loss": 2.614,
+      "step": 37980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.458037151780909e-05,
+      "loss": 2.3732,
+      "step": 37985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45789673064666e-05,
+      "loss": 2.4797,
+      "step": 37990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.457756292366074e-05,
+      "loss": 2.588,
+      "step": 37995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45761583693969e-05,
+      "loss": 2.3778,
+      "step": 38000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.457475364368047e-05,
+      "loss": 2.4756,
+      "step": 38005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.457334874651687e-05,
+      "loss": 2.6031,
+      "step": 38010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.457194367791149e-05,
+      "loss": 2.5353,
+      "step": 38015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.457053843786974e-05,
+      "loss": 2.4316,
+      "step": 38020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.456913302639702e-05,
+      "loss": 2.6715,
+      "step": 38025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.456772744349875e-05,
+      "loss": 2.4402,
+      "step": 38030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.456632168918033e-05,
+      "loss": 2.4239,
+      "step": 38035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.456491576344715e-05,
+      "loss": 2.5569,
+      "step": 38040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.456350966630465e-05,
+      "loss": 2.6061,
+      "step": 38045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.456210339775819e-05,
+      "loss": 2.5023,
+      "step": 38050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.456069695781324e-05,
+      "loss": 2.5391,
+      "step": 38055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.455929034647518e-05,
+      "loss": 2.6056,
+      "step": 38060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45578835637494e-05,
+      "loss": 2.5117,
+      "step": 38065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.455647660964133e-05,
+      "loss": 2.3981,
+      "step": 38070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.455506948415639e-05,
+      "loss": 2.5082,
+      "step": 38075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.455366218729999e-05,
+      "loss": 2.5114,
+      "step": 38080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.455225471907752e-05,
+      "loss": 2.6301,
+      "step": 38085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.455084707949442e-05,
+      "loss": 2.4676,
+      "step": 38090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.454943926855608e-05,
+      "loss": 2.4629,
+      "step": 38095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.454803128626795e-05,
+      "loss": 2.6391,
+      "step": 38100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45466231326354e-05,
+      "loss": 2.3861,
+      "step": 38105
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45452148076639e-05,
+      "loss": 2.4955,
+      "step": 38110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.454380631135882e-05,
+      "loss": 2.5171,
+      "step": 38115
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45423976437256e-05,
+      "loss": 2.5726,
+      "step": 38120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.454098880476964e-05,
+      "loss": 2.4767,
+      "step": 38125
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.453957979449637e-05,
+      "loss": 2.4405,
+      "step": 38130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.453817061291123e-05,
+      "loss": 2.3846,
+      "step": 38135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.453676126001963e-05,
+      "loss": 2.536,
+      "step": 38140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.453535173582695e-05,
+      "loss": 2.5005,
+      "step": 38145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.453394204033866e-05,
+      "loss": 2.6585,
+      "step": 38150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.453253217356018e-05,
+      "loss": 2.4537,
+      "step": 38155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.453112213549689e-05,
+      "loss": 2.4546,
+      "step": 38160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.452971192615425e-05,
+      "loss": 2.4668,
+      "step": 38165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.452830154553769e-05,
+      "loss": 2.517,
+      "step": 38170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.452689099365259e-05,
+      "loss": 2.632,
+      "step": 38175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.452548027050444e-05,
+      "loss": 2.4897,
+      "step": 38180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.452406937609861e-05,
+      "loss": 2.5972,
+      "step": 38185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.452265831044056e-05,
+      "loss": 2.448,
+      "step": 38190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.452124707353572e-05,
+      "loss": 2.473,
+      "step": 38195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.451983566538946e-05,
+      "loss": 2.4947,
+      "step": 38200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.451842408600728e-05,
+      "loss": 2.4889,
+      "step": 38205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.451701233539458e-05,
+      "loss": 2.5639,
+      "step": 38210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45156004135568e-05,
+      "loss": 2.5312,
+      "step": 38215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.451418832049937e-05,
+      "loss": 2.4944,
+      "step": 38220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45127760562277e-05,
+      "loss": 2.5555,
+      "step": 38225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.451136362074723e-05,
+      "loss": 2.5863,
+      "step": 38230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45099510140634e-05,
+      "loss": 2.5222,
+      "step": 38235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.450853823618164e-05,
+      "loss": 2.556,
+      "step": 38240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.450712528710738e-05,
+      "loss": 2.5241,
+      "step": 38245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.450571216684609e-05,
+      "loss": 2.6473,
+      "step": 38250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.450429887540315e-05,
+      "loss": 2.5877,
+      "step": 38255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.450288541278404e-05,
+      "loss": 2.4375,
+      "step": 38260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.450147177899415e-05,
+      "loss": 2.59,
+      "step": 38265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.450005797403897e-05,
+      "loss": 2.5026,
+      "step": 38270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.449864399792392e-05,
+      "loss": 2.4214,
+      "step": 38275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44972298506544e-05,
+      "loss": 2.611,
+      "step": 38280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.449581553223591e-05,
+      "loss": 2.5294,
+      "step": 38285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.449440104267385e-05,
+      "loss": 2.3546,
+      "step": 38290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.449298638197367e-05,
+      "loss": 2.5777,
+      "step": 38295
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.449157155014081e-05,
+      "loss": 2.6249,
+      "step": 38300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.449015654718072e-05,
+      "loss": 2.5972,
+      "step": 38305
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.448874137309884e-05,
+      "loss": 2.3356,
+      "step": 38310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44873260279006e-05,
+      "loss": 2.5129,
+      "step": 38315
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.448591051159146e-05,
+      "loss": 2.4781,
+      "step": 38320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.448449482417686e-05,
+      "loss": 2.4105,
+      "step": 38325
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.448307896566224e-05,
+      "loss": 2.5431,
+      "step": 38330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.448166293605305e-05,
+      "loss": 2.4239,
+      "step": 38335
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.448024673535474e-05,
+      "loss": 2.4766,
+      "step": 38340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.447883036357274e-05,
+      "loss": 2.5623,
+      "step": 38345
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.447741382071253e-05,
+      "loss": 2.45,
+      "step": 38350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.447599710677953e-05,
+      "loss": 2.4997,
+      "step": 38355
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44745802217792e-05,
+      "loss": 2.5674,
+      "step": 38360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.447316316571698e-05,
+      "loss": 2.6035,
+      "step": 38365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.447174593859834e-05,
+      "loss": 2.5482,
+      "step": 38370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44703285404287e-05,
+      "loss": 2.4923,
+      "step": 38375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.446891097121355e-05,
+      "loss": 2.4601,
+      "step": 38380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.446749323095832e-05,
+      "loss": 2.3669,
+      "step": 38385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.446607531966847e-05,
+      "loss": 2.5046,
+      "step": 38390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.446465723734945e-05,
+      "loss": 2.5776,
+      "step": 38395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44632389840067e-05,
+      "loss": 2.408,
+      "step": 38400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.446182055964569e-05,
+      "loss": 2.44,
+      "step": 38405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44604019642719e-05,
+      "loss": 2.6298,
+      "step": 38410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.445898319789073e-05,
+      "loss": 2.4765,
+      "step": 38415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44575642605077e-05,
+      "loss": 2.4664,
+      "step": 38420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.445614515212822e-05,
+      "loss": 2.5656,
+      "step": 38425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.445472587275776e-05,
+      "loss": 2.5686,
+      "step": 38430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.445330642240179e-05,
+      "loss": 2.5121,
+      "step": 38435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.445188680106576e-05,
+      "loss": 2.5503,
+      "step": 38440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.445046700875513e-05,
+      "loss": 2.5406,
+      "step": 38445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.444904704547537e-05,
+      "loss": 2.4936,
+      "step": 38450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.444762691123194e-05,
+      "loss": 2.5927,
+      "step": 38455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44462066060303e-05,
+      "loss": 2.5684,
+      "step": 38460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44447861298759e-05,
+      "loss": 2.5777,
+      "step": 38465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.444336548277422e-05,
+      "loss": 2.5803,
+      "step": 38470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44419446647307e-05,
+      "loss": 2.5508,
+      "step": 38475
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.444052367575084e-05,
+      "loss": 2.4357,
+      "step": 38480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.443910251584009e-05,
+      "loss": 2.4264,
+      "step": 38485
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.443768118500393e-05,
+      "loss": 2.4693,
+      "step": 38490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.443625968324778e-05,
+      "loss": 2.5556,
+      "step": 38495
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.443483801057716e-05,
+      "loss": 2.5362,
+      "step": 38500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.443341616699752e-05,
+      "loss": 2.625,
+      "step": 38505
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.443199415251432e-05,
+      "loss": 2.6553,
+      "step": 38510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.443057196713303e-05,
+      "loss": 2.4553,
+      "step": 38515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.442914961085913e-05,
+      "loss": 2.4756,
+      "step": 38520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.442772708369811e-05,
+      "loss": 2.4256,
+      "step": 38525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.442630438565539e-05,
+      "loss": 2.6349,
+      "step": 38530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.442488151673648e-05,
+      "loss": 2.5849,
+      "step": 38535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.442345847694685e-05,
+      "loss": 2.6006,
+      "step": 38540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.442203526629195e-05,
+      "loss": 2.4942,
+      "step": 38545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.442061188477728e-05,
+      "loss": 2.4797,
+      "step": 38550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.441918833240831e-05,
+      "loss": 2.4352,
+      "step": 38555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.44177646091905e-05,
+      "loss": 2.4905,
+      "step": 38560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.441634071512935e-05,
+      "loss": 2.607,
+      "step": 38565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.441491665023031e-05,
+      "loss": 2.5185,
+      "step": 38570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.441349241449888e-05,
+      "loss": 2.4801,
+      "step": 38575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.441206800794054e-05,
+      "loss": 2.5212,
+      "step": 38580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.441064343056074e-05,
+      "loss": 2.447,
+      "step": 38585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.440921868236498e-05,
+      "loss": 2.4809,
+      "step": 38590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.440779376335873e-05,
+      "loss": 2.4303,
+      "step": 38595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.440636867354749e-05,
+      "loss": 2.5008,
+      "step": 38600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.440494341293673e-05,
+      "loss": 2.372,
+      "step": 38605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.440351798153193e-05,
+      "loss": 2.4528,
+      "step": 38610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.440209237933858e-05,
+      "loss": 2.5525,
+      "step": 38615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.440066660636213e-05,
+      "loss": 2.4512,
+      "step": 38620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43992406626081e-05,
+      "loss": 2.6032,
+      "step": 38625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.439781454808199e-05,
+      "loss": 2.5713,
+      "step": 38630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.439638826278924e-05,
+      "loss": 2.3988,
+      "step": 38635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.439496180673535e-05,
+      "loss": 2.551,
+      "step": 38640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.439353517992582e-05,
+      "loss": 2.3688,
+      "step": 38645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.439210838236614e-05,
+      "loss": 2.4513,
+      "step": 38650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.439068141406178e-05,
+      "loss": 2.4704,
+      "step": 38655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.438925427501823e-05,
+      "loss": 2.4757,
+      "step": 38660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.4387826965241e-05,
+      "loss": 2.3226,
+      "step": 38665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.438639948473555e-05,
+      "loss": 2.3402,
+      "step": 38670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.438497183350741e-05,
+      "loss": 2.4618,
+      "step": 38675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.438354401156203e-05,
+      "loss": 2.523,
+      "step": 38680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.438211601890492e-05,
+      "loss": 2.5886,
+      "step": 38685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.438068785554157e-05,
+      "loss": 2.4626,
+      "step": 38690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.437925952147748e-05,
+      "loss": 2.5417,
+      "step": 38695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.437783101671813e-05,
+      "loss": 2.4562,
+      "step": 38700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.437640234126903e-05,
+      "loss": 2.4959,
+      "step": 38705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.437497349513567e-05,
+      "loss": 2.4219,
+      "step": 38710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.437354447832355e-05,
+      "loss": 2.6123,
+      "step": 38715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.437211529083815e-05,
+      "loss": 2.6134,
+      "step": 38720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.437068593268498e-05,
+      "loss": 2.4328,
+      "step": 38725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.436925640386954e-05,
+      "loss": 2.4302,
+      "step": 38730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43678267043973e-05,
+      "loss": 2.5398,
+      "step": 38735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.436639683427381e-05,
+      "loss": 2.5378,
+      "step": 38740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.436496679350453e-05,
+      "loss": 2.6488,
+      "step": 38745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.436353658209498e-05,
+      "loss": 2.5389,
+      "step": 38750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.436210620005064e-05,
+      "loss": 2.5762,
+      "step": 38755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.436067564737703e-05,
+      "loss": 2.5797,
+      "step": 38760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.435924492407964e-05,
+      "loss": 2.5168,
+      "step": 38765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.4357814030164e-05,
+      "loss": 2.5649,
+      "step": 38770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.435638296563558e-05,
+      "loss": 2.564,
+      "step": 38775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.435495173049991e-05,
+      "loss": 2.5792,
+      "step": 38780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.435352032476247e-05,
+      "loss": 2.5378,
+      "step": 38785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.435208874842878e-05,
+      "loss": 2.6461,
+      "step": 38790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.435065700150435e-05,
+      "loss": 2.5416,
+      "step": 38795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.434922508399468e-05,
+      "loss": 2.4357,
+      "step": 38800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.434779299590528e-05,
+      "loss": 2.6365,
+      "step": 38805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.434636073724165e-05,
+      "loss": 2.504,
+      "step": 38810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.434492830800932e-05,
+      "loss": 2.55,
+      "step": 38815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.434349570821379e-05,
+      "loss": 2.4766,
+      "step": 38820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.434206293786054e-05,
+      "loss": 2.6861,
+      "step": 38825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.434062999695514e-05,
+      "loss": 2.4359,
+      "step": 38830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.433919688550305e-05,
+      "loss": 2.5549,
+      "step": 38835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43377636035098e-05,
+      "loss": 2.4303,
+      "step": 38840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.433633015098091e-05,
+      "loss": 2.5374,
+      "step": 38845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.433489652792188e-05,
+      "loss": 2.5581,
+      "step": 38850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.433346273433823e-05,
+      "loss": 2.4122,
+      "step": 38855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.433202877023548e-05,
+      "loss": 2.5081,
+      "step": 38860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.433059463561915e-05,
+      "loss": 2.5633,
+      "step": 38865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.432916033049474e-05,
+      "loss": 2.6702,
+      "step": 38870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.432772585486778e-05,
+      "loss": 2.4994,
+      "step": 38875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.432629120874377e-05,
+      "loss": 2.3867,
+      "step": 38880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.432485639212825e-05,
+      "loss": 2.3685,
+      "step": 38885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.432342140502672e-05,
+      "loss": 2.4275,
+      "step": 38890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.432198624744471e-05,
+      "loss": 2.531,
+      "step": 38895
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.432055091938774e-05,
+      "loss": 2.5634,
+      "step": 38900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.431911542086132e-05,
+      "loss": 2.5922,
+      "step": 38905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.431767975187097e-05,
+      "loss": 2.458,
+      "step": 38910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.431624391242225e-05,
+      "loss": 2.5579,
+      "step": 38915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.431480790252066e-05,
+      "loss": 2.5138,
+      "step": 38920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43133717221717e-05,
+      "loss": 2.3962,
+      "step": 38925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43119353713809e-05,
+      "loss": 2.5117,
+      "step": 38930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.431049885015382e-05,
+      "loss": 2.374,
+      "step": 38935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.430906215849594e-05,
+      "loss": 2.4912,
+      "step": 38940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.430762529641281e-05,
+      "loss": 2.5023,
+      "step": 38945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.430618826390998e-05,
+      "loss": 2.5131,
+      "step": 38950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.430475106099292e-05,
+      "loss": 2.5192,
+      "step": 38955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43033136876672e-05,
+      "loss": 2.4686,
+      "step": 38960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.430187614393836e-05,
+      "loss": 2.4412,
+      "step": 38965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.430043842981188e-05,
+      "loss": 2.6123,
+      "step": 38970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.429900054529333e-05,
+      "loss": 2.4885,
+      "step": 38975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.429756249038822e-05,
+      "loss": 2.4536,
+      "step": 38980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.429612426510208e-05,
+      "loss": 2.5462,
+      "step": 38985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.429468586944045e-05,
+      "loss": 2.505,
+      "step": 38990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.429324730340888e-05,
+      "loss": 2.5369,
+      "step": 38995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.429180856701289e-05,
+      "loss": 2.3909,
+      "step": 39000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.429036966025799e-05,
+      "loss": 2.4729,
+      "step": 39005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.428893058314974e-05,
+      "loss": 2.5167,
+      "step": 39010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.428749133569368e-05,
+      "loss": 2.5607,
+      "step": 39015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.428605191789533e-05,
+      "loss": 2.5389,
+      "step": 39020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.428461232976023e-05,
+      "loss": 2.5587,
+      "step": 39025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.42831725712939e-05,
+      "loss": 2.4952,
+      "step": 39030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.428173264250192e-05,
+      "loss": 2.3987,
+      "step": 39035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.42802925433898e-05,
+      "loss": 2.5353,
+      "step": 39040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.427885227396308e-05,
+      "loss": 2.5734,
+      "step": 39045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.42774118342273e-05,
+      "loss": 2.5201,
+      "step": 39050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.4275971224188e-05,
+      "loss": 2.389,
+      "step": 39055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.427453044385075e-05,
+      "loss": 2.4983,
+      "step": 39060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.427308949322104e-05,
+      "loss": 2.5328,
+      "step": 39065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.427164837230445e-05,
+      "loss": 2.5062,
+      "step": 39070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.427020708110651e-05,
+      "loss": 2.3762,
+      "step": 39075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.426876561963275e-05,
+      "loss": 2.5295,
+      "step": 39080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.426732398788876e-05,
+      "loss": 2.4552,
+      "step": 39085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.426588218588004e-05,
+      "loss": 2.5285,
+      "step": 39090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.426444021361214e-05,
+      "loss": 2.4816,
+      "step": 39095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.426299807109063e-05,
+      "loss": 2.4802,
+      "step": 39100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.426155575832103e-05,
+      "loss": 2.4337,
+      "step": 39105
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.426011327530891e-05,
+      "loss": 2.4545,
+      "step": 39110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.425867062205981e-05,
+      "loss": 2.5306,
+      "step": 39115
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.425722779857928e-05,
+      "loss": 2.4142,
+      "step": 39120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.425578480487286e-05,
+      "loss": 2.4326,
+      "step": 39125
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.42543416409461e-05,
+      "loss": 2.6679,
+      "step": 39130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.425289830680457e-05,
+      "loss": 2.4636,
+      "step": 39135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.425145480245381e-05,
+      "loss": 2.5728,
+      "step": 39140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.425001112789937e-05,
+      "loss": 2.4968,
+      "step": 39145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.42485672831468e-05,
+      "loss": 2.5129,
+      "step": 39150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.424712326820167e-05,
+      "loss": 2.4644,
+      "step": 39155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.424567908306951e-05,
+      "loss": 2.6361,
+      "step": 39160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.424423472775588e-05,
+      "loss": 2.4565,
+      "step": 39165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.424279020226636e-05,
+      "loss": 2.5857,
+      "step": 39170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.424134550660649e-05,
+      "loss": 2.3347,
+      "step": 39175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.42399006407818e-05,
+      "loss": 2.4102,
+      "step": 39180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.42384556047979e-05,
+      "loss": 2.4846,
+      "step": 39185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.423701039866031e-05,
+      "loss": 2.6599,
+      "step": 39190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.423556502237459e-05,
+      "loss": 2.56,
+      "step": 39195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.423411947594631e-05,
+      "loss": 2.527,
+      "step": 39200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.423267375938103e-05,
+      "loss": 2.4256,
+      "step": 39205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.423122787268433e-05,
+      "loss": 2.3864,
+      "step": 39210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.422978181586173e-05,
+      "loss": 2.5093,
+      "step": 39215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.422833558891882e-05,
+      "loss": 2.4752,
+      "step": 39220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.422688919186114e-05,
+      "loss": 2.5852,
+      "step": 39225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.422544262469429e-05,
+      "loss": 2.5762,
+      "step": 39230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.422399588742379e-05,
+      "loss": 2.5206,
+      "step": 39235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.422254898005524e-05,
+      "loss": 2.5062,
+      "step": 39240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.422110190259418e-05,
+      "loss": 2.5513,
+      "step": 39245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.421965465504621e-05,
+      "loss": 2.5265,
+      "step": 39250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.421820723741685e-05,
+      "loss": 2.533,
+      "step": 39255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.42167596497117e-05,
+      "loss": 2.4062,
+      "step": 39260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.421531189193633e-05,
+      "loss": 2.637,
+      "step": 39265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.421386396409628e-05,
+      "loss": 2.4849,
+      "step": 39270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.421241586619715e-05,
+      "loss": 2.5402,
+      "step": 39275
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.42109675982445e-05,
+      "loss": 2.4098,
+      "step": 39280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.420951916024388e-05,
+      "loss": 2.5004,
+      "step": 39285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.420807055220088e-05,
+      "loss": 2.5076,
+      "step": 39290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.420662177412108e-05,
+      "loss": 2.4742,
+      "step": 39295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.420517282601005e-05,
+      "loss": 2.6155,
+      "step": 39300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.420372370787334e-05,
+      "loss": 2.52,
+      "step": 39305
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.420227441971654e-05,
+      "loss": 2.4888,
+      "step": 39310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.420082496154523e-05,
+      "loss": 2.5433,
+      "step": 39315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.419937533336498e-05,
+      "loss": 2.4901,
+      "step": 39320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.419792553518137e-05,
+      "loss": 2.4476,
+      "step": 39325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.419647556699996e-05,
+      "loss": 2.466,
+      "step": 39330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.419502542882632e-05,
+      "loss": 2.4125,
+      "step": 39335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.419357512066607e-05,
+      "loss": 2.5537,
+      "step": 39340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.419212464252476e-05,
+      "loss": 2.3686,
+      "step": 39345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.419067399440798e-05,
+      "loss": 2.4961,
+      "step": 39350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.418922317632128e-05,
+      "loss": 2.5598,
+      "step": 39355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.418777218827027e-05,
+      "loss": 2.5637,
+      "step": 39360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.418632103026053e-05,
+      "loss": 2.4047,
+      "step": 39365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.418486970229762e-05,
+      "loss": 2.564,
+      "step": 39370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.418341820438717e-05,
+      "loss": 2.5671,
+      "step": 39375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.41819665365347e-05,
+      "loss": 2.526,
+      "step": 39380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.418051469874582e-05,
+      "loss": 2.5202,
+      "step": 39385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.417906269102614e-05,
+      "loss": 2.5974,
+      "step": 39390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.417761051338121e-05,
+      "loss": 2.4736,
+      "step": 39395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.417615816581663e-05,
+      "loss": 2.4465,
+      "step": 39400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.417470564833799e-05,
+      "loss": 2.4456,
+      "step": 39405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.417325296095086e-05,
+      "loss": 2.5796,
+      "step": 39410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.417180010366085e-05,
+      "loss": 2.4952,
+      "step": 39415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.417034707647352e-05,
+      "loss": 2.5598,
+      "step": 39420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.416889387939449e-05,
+      "loss": 2.4316,
+      "step": 39425
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.416744051242933e-05,
+      "loss": 2.4458,
+      "step": 39430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.416598697558364e-05,
+      "loss": 2.478,
+      "step": 39435
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.416453326886299e-05,
+      "loss": 2.3611,
+      "step": 39440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.4163079392273e-05,
+      "loss": 2.5864,
+      "step": 39445
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.416162534581924e-05,
+      "loss": 2.4937,
+      "step": 39450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.416017112950731e-05,
+      "loss": 2.425,
+      "step": 39455
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.415871674334282e-05,
+      "loss": 2.5507,
+      "step": 39460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.415726218733133e-05,
+      "loss": 2.503,
+      "step": 39465
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.415580746147847e-05,
+      "loss": 2.5724,
+      "step": 39470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.415435256578982e-05,
+      "loss": 2.4882,
+      "step": 39475
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.415289750027095e-05,
+      "loss": 2.456,
+      "step": 39480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.415144226492751e-05,
+      "loss": 2.5327,
+      "step": 39485
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.414998685976504e-05,
+      "loss": 2.4429,
+      "step": 39490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.414853128478919e-05,
+      "loss": 2.4874,
+      "step": 39495
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.414707554000553e-05,
+      "loss": 2.3858,
+      "step": 39500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.414561962541966e-05,
+      "loss": 2.4148,
+      "step": 39505
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.414416354103718e-05,
+      "loss": 2.4908,
+      "step": 39510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.41427072868637e-05,
+      "loss": 2.4607,
+      "step": 39515
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.414125086290482e-05,
+      "loss": 2.5353,
+      "step": 39520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.413979426916613e-05,
+      "loss": 2.5688,
+      "step": 39525
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.413833750565324e-05,
+      "loss": 2.5321,
+      "step": 39530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.413688057237176e-05,
+      "loss": 2.4926,
+      "step": 39535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.413542346932728e-05,
+      "loss": 2.5797,
+      "step": 39540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.413396619652542e-05,
+      "loss": 2.5076,
+      "step": 39545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.413250875397176e-05,
+      "loss": 2.4633,
+      "step": 39550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.413105114167194e-05,
+      "loss": 2.5115,
+      "step": 39555
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.412959335963154e-05,
+      "loss": 2.6282,
+      "step": 39560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.412813540785617e-05,
+      "loss": 2.5039,
+      "step": 39565
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.412667728635145e-05,
+      "loss": 2.5288,
+      "step": 39570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.412521899512298e-05,
+      "loss": 2.6875,
+      "step": 39575
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.412376053417637e-05,
+      "loss": 2.5523,
+      "step": 39580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.412230190351723e-05,
+      "loss": 2.4543,
+      "step": 39585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.412084310315116e-05,
+      "loss": 2.3931,
+      "step": 39590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.41193841330838e-05,
+      "loss": 2.446,
+      "step": 39595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.411792499332074e-05,
+      "loss": 2.4016,
+      "step": 39600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.41164656838676e-05,
+      "loss": 2.4915,
+      "step": 39605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.411500620472997e-05,
+      "loss": 2.4635,
+      "step": 39610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.41135465559135e-05,
+      "loss": 2.3338,
+      "step": 39615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.411208673742377e-05,
+      "loss": 2.5672,
+      "step": 39620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.411062674926642e-05,
+      "loss": 2.6581,
+      "step": 39625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.410916659144706e-05,
+      "loss": 2.4089,
+      "step": 39630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.41077062639713e-05,
+      "loss": 2.4583,
+      "step": 39635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.410624576684475e-05,
+      "loss": 2.4321,
+      "step": 39640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.410478510007305e-05,
+      "loss": 2.6531,
+      "step": 39645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.41033242636618e-05,
+      "loss": 2.6066,
+      "step": 39650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.410186325761663e-05,
+      "loss": 2.7054,
+      "step": 39655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.410040208194314e-05,
+      "loss": 2.4807,
+      "step": 39660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.409894073664697e-05,
+      "loss": 2.4924,
+      "step": 39665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.409747922173372e-05,
+      "loss": 2.525,
+      "step": 39670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.409601753720905e-05,
+      "loss": 2.5626,
+      "step": 39675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.409455568307854e-05,
+      "loss": 2.3107,
+      "step": 39680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.409309365934785e-05,
+      "loss": 2.5548,
+      "step": 39685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.409163146602258e-05,
+      "loss": 2.4097,
+      "step": 39690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.409016910310835e-05,
+      "loss": 2.4679,
+      "step": 39695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40887065706108e-05,
+      "loss": 2.4505,
+      "step": 39700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.408724386853554e-05,
+      "loss": 2.5215,
+      "step": 39705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.408578099688818e-05,
+      "loss": 2.4888,
+      "step": 39710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40843179556744e-05,
+      "loss": 2.4232,
+      "step": 39715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.408285474489979e-05,
+      "loss": 2.5088,
+      "step": 39720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.408139136456999e-05,
+      "loss": 2.5436,
+      "step": 39725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.407992781469063e-05,
+      "loss": 2.539,
+      "step": 39730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.407846409526733e-05,
+      "loss": 2.5894,
+      "step": 39735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40770002063057e-05,
+      "loss": 2.4531,
+      "step": 39740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.407553614781142e-05,
+      "loss": 2.5188,
+      "step": 39745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.407407191979009e-05,
+      "loss": 2.4372,
+      "step": 39750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.407260752224734e-05,
+      "loss": 2.4591,
+      "step": 39755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.407114295518881e-05,
+      "loss": 2.5666,
+      "step": 39760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.406967821862014e-05,
+      "loss": 2.3504,
+      "step": 39765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.406821331254696e-05,
+      "loss": 2.3727,
+      "step": 39770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.406674823697488e-05,
+      "loss": 2.5559,
+      "step": 39775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.406528299190957e-05,
+      "loss": 2.5598,
+      "step": 39780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.406381757735666e-05,
+      "loss": 2.6573,
+      "step": 39785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.406235199332176e-05,
+      "loss": 2.5691,
+      "step": 39790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.406088623981053e-05,
+      "loss": 2.4625,
+      "step": 39795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.405942031682861e-05,
+      "loss": 2.6398,
+      "step": 39800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.405795422438162e-05,
+      "loss": 2.6374,
+      "step": 39805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.405648796247522e-05,
+      "loss": 2.5628,
+      "step": 39810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.405502153111503e-05,
+      "loss": 2.5997,
+      "step": 39815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40535549303067e-05,
+      "loss": 2.5666,
+      "step": 39820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.405208816005589e-05,
+      "loss": 2.3986,
+      "step": 39825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40506212203682e-05,
+      "loss": 2.5917,
+      "step": 39830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.404915411124931e-05,
+      "loss": 2.5846,
+      "step": 39835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.404768683270484e-05,
+      "loss": 2.456,
+      "step": 39840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.404621938474044e-05,
+      "loss": 2.5351,
+      "step": 39845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.404475176736176e-05,
+      "loss": 2.5582,
+      "step": 39850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.404328398057443e-05,
+      "loss": 2.5021,
+      "step": 39855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.404181602438412e-05,
+      "loss": 2.4281,
+      "step": 39860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.404034789879646e-05,
+      "loss": 2.4899,
+      "step": 39865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40388796038171e-05,
+      "loss": 2.4576,
+      "step": 39870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.403741113945167e-05,
+      "loss": 2.5332,
+      "step": 39875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.403594250570585e-05,
+      "loss": 2.3656,
+      "step": 39880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.403447370258527e-05,
+      "loss": 2.5982,
+      "step": 39885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.403300473009557e-05,
+      "loss": 2.5382,
+      "step": 39890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.403153558824244e-05,
+      "loss": 2.4763,
+      "step": 39895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.403006627703148e-05,
+      "loss": 2.457,
+      "step": 39900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.402859679646838e-05,
+      "loss": 2.4855,
+      "step": 39905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.402712714655877e-05,
+      "loss": 2.6173,
+      "step": 39910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.402565732730832e-05,
+      "loss": 2.5068,
+      "step": 39915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.402418733872265e-05,
+      "loss": 2.4417,
+      "step": 39920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.402271718080746e-05,
+      "loss": 2.5058,
+      "step": 39925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.402124685356836e-05,
+      "loss": 2.6295,
+      "step": 39930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.401977635701105e-05,
+      "loss": 2.5716,
+      "step": 39935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.401830569114116e-05,
+      "loss": 2.6652,
+      "step": 39940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.401683485596434e-05,
+      "loss": 2.5138,
+      "step": 39945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.401536385148625e-05,
+      "loss": 2.5567,
+      "step": 39950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.401389267771256e-05,
+      "loss": 2.546,
+      "step": 39955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.401242133464894e-05,
+      "loss": 2.4186,
+      "step": 39960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.401094982230101e-05,
+      "loss": 2.5081,
+      "step": 39965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.400947814067446e-05,
+      "loss": 2.4579,
+      "step": 39970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.400800628977495e-05,
+      "loss": 2.3771,
+      "step": 39975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.400653426960812e-05,
+      "loss": 2.459,
+      "step": 39980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.400506208017967e-05,
+      "loss": 2.6828,
+      "step": 39985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.400358972149522e-05,
+      "loss": 2.6346,
+      "step": 39990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.400211719356045e-05,
+      "loss": 2.517,
+      "step": 39995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.400064449638104e-05,
+      "loss": 2.4656,
+      "step": 40000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.399917162996261e-05,
+      "loss": 2.5347,
+      "step": 40005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.399769859431087e-05,
+      "loss": 2.5431,
+      "step": 40010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.399622538943148e-05,
+      "loss": 2.5999,
+      "step": 40015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.399475201533009e-05,
+      "loss": 2.5516,
+      "step": 40020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.399327847201238e-05,
+      "loss": 2.3529,
+      "step": 40025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.399180475948399e-05,
+      "loss": 2.4751,
+      "step": 40030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.399033087775064e-05,
+      "loss": 2.5139,
+      "step": 40035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.398885682681795e-05,
+      "loss": 2.4919,
+      "step": 40040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.398738260669162e-05,
+      "loss": 2.4694,
+      "step": 40045
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.39859082173773e-05,
+      "loss": 2.5231,
+      "step": 40050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.398443365888066e-05,
+      "loss": 2.5033,
+      "step": 40055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.398295893120738e-05,
+      "loss": 2.5915,
+      "step": 40060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.398148403436315e-05,
+      "loss": 2.5854,
+      "step": 40065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.398000896835361e-05,
+      "loss": 2.5949,
+      "step": 40070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.397853373318447e-05,
+      "loss": 2.453,
+      "step": 40075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.397705832886138e-05,
+      "loss": 2.3165,
+      "step": 40080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.397558275539e-05,
+      "loss": 2.5298,
+      "step": 40085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.397410701277605e-05,
+      "loss": 2.4947,
+      "step": 40090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.397263110102517e-05,
+      "loss": 2.3488,
+      "step": 40095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.397115502014303e-05,
+      "loss": 2.45,
+      "step": 40100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.396967877013533e-05,
+      "loss": 2.497,
+      "step": 40105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.396820235100777e-05,
+      "loss": 2.3571,
+      "step": 40110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.396672576276597e-05,
+      "loss": 2.5007,
+      "step": 40115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.396524900541565e-05,
+      "loss": 2.4463,
+      "step": 40120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.396377207896248e-05,
+      "loss": 2.5121,
+      "step": 40125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.396229498341216e-05,
+      "loss": 2.5504,
+      "step": 40130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.396081771877032e-05,
+      "loss": 2.6967,
+      "step": 40135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.39593402850427e-05,
+      "loss": 2.4905,
+      "step": 40140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.395786268223494e-05,
+      "loss": 2.4047,
+      "step": 40145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.395638491035276e-05,
+      "loss": 2.6764,
+      "step": 40150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.395490696940182e-05,
+      "loss": 2.5897,
+      "step": 40155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.395342885938778e-05,
+      "loss": 2.5853,
+      "step": 40160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.395195058031639e-05,
+      "loss": 2.5301,
+      "step": 40165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.395047213219328e-05,
+      "loss": 2.459,
+      "step": 40170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.394899351502417e-05,
+      "loss": 2.434,
+      "step": 40175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.394751472881473e-05,
+      "loss": 2.5059,
+      "step": 40180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.394603577357064e-05,
+      "loss": 2.5293,
+      "step": 40185
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.394455664929763e-05,
+      "loss": 2.4175,
+      "step": 40190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.394307735600132e-05,
+      "loss": 2.4936,
+      "step": 40195
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.394159789368745e-05,
+      "loss": 2.4009,
+      "step": 40200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.394011826236172e-05,
+      "loss": 2.4806,
+      "step": 40205
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.39386384620298e-05,
+      "loss": 2.433,
+      "step": 40210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.393715849269735e-05,
+      "loss": 2.4467,
+      "step": 40215
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.393567835437011e-05,
+      "loss": 2.4154,
+      "step": 40220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.393419804705377e-05,
+      "loss": 2.4665,
+      "step": 40225
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.393271757075402e-05,
+      "loss": 2.3508,
+      "step": 40230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.393123692547652e-05,
+      "loss": 2.4019,
+      "step": 40235
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.3929756111227e-05,
+      "loss": 2.4594,
+      "step": 40240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.392827512801114e-05,
+      "loss": 2.5629,
+      "step": 40245
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.392679397583466e-05,
+      "loss": 2.5584,
+      "step": 40250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.392531265470323e-05,
+      "loss": 2.5198,
+      "step": 40255
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.392383116462255e-05,
+      "loss": 2.6095,
+      "step": 40260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.392234950559833e-05,
+      "loss": 2.6023,
+      "step": 40265
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.392086767763627e-05,
+      "loss": 2.4571,
+      "step": 40270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.391938568074206e-05,
+      "loss": 2.5109,
+      "step": 40275
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.39179035149214e-05,
+      "loss": 2.5255,
+      "step": 40280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.391642118018e-05,
+      "loss": 2.3594,
+      "step": 40285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.391493867652356e-05,
+      "loss": 2.48,
+      "step": 40290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.391345600395777e-05,
+      "loss": 2.4525,
+      "step": 40295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.391197316248835e-05,
+      "loss": 2.4264,
+      "step": 40300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.391049015212099e-05,
+      "loss": 2.5073,
+      "step": 40305
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.39090069728614e-05,
+      "loss": 2.6294,
+      "step": 40310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.390752362471529e-05,
+      "loss": 2.5993,
+      "step": 40315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.390604010768835e-05,
+      "loss": 2.4522,
+      "step": 40320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.390455642178628e-05,
+      "loss": 2.5341,
+      "step": 40325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.390307256701483e-05,
+      "loss": 2.35,
+      "step": 40330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.390158854337966e-05,
+      "loss": 2.581,
+      "step": 40335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.39001043508865e-05,
+      "loss": 2.5157,
+      "step": 40340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.389861998954106e-05,
+      "loss": 2.3862,
+      "step": 40345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.389713545934905e-05,
+      "loss": 2.5062,
+      "step": 40350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.389565076031617e-05,
+      "loss": 2.5392,
+      "step": 40355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.389416589244813e-05,
+      "loss": 2.5149,
+      "step": 40360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.389268085575065e-05,
+      "loss": 2.423,
+      "step": 40365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.389119565022943e-05,
+      "loss": 2.4832,
+      "step": 40370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.388971027589021e-05,
+      "loss": 2.4617,
+      "step": 40375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.388822473273868e-05,
+      "loss": 2.5221,
+      "step": 40380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.388673902078054e-05,
+      "loss": 2.4604,
+      "step": 40385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.388525314002153e-05,
+      "loss": 2.5644,
+      "step": 40390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.388376709046737e-05,
+      "loss": 2.4696,
+      "step": 40395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.388228087212376e-05,
+      "loss": 2.314,
+      "step": 40400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.38807944849964e-05,
+      "loss": 2.4347,
+      "step": 40405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.387930792909105e-05,
+      "loss": 2.6066,
+      "step": 40410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.38778212044134e-05,
+      "loss": 2.5113,
+      "step": 40415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.387633431096916e-05,
+      "loss": 2.4633,
+      "step": 40420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.387484724876407e-05,
+      "loss": 2.5779,
+      "step": 40425
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.387336001780385e-05,
+      "loss": 2.4241,
+      "step": 40430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.387187261809421e-05,
+      "loss": 2.3461,
+      "step": 40435
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.387038504964085e-05,
+      "loss": 2.4794,
+      "step": 40440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.386889731244954e-05,
+      "loss": 2.5254,
+      "step": 40445
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.386740940652596e-05,
+      "loss": 2.5054,
+      "step": 40450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.386592133187586e-05,
+      "loss": 2.5413,
+      "step": 40455
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.386443308850494e-05,
+      "loss": 2.5139,
+      "step": 40460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.386294467641895e-05,
+      "loss": 2.6551,
+      "step": 40465
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.38614560956236e-05,
+      "loss": 2.4897,
+      "step": 40470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.385996734612461e-05,
+      "loss": 2.4511,
+      "step": 40475
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.385847842792774e-05,
+      "loss": 2.441,
+      "step": 40480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.385698934103866e-05,
+      "loss": 2.4177,
+      "step": 40485
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.385550008546313e-05,
+      "loss": 2.6122,
+      "step": 40490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.385401066120689e-05,
+      "loss": 2.5584,
+      "step": 40495
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.385252106827565e-05,
+      "loss": 2.5264,
+      "step": 40500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.385103130667515e-05,
+      "loss": 2.5173,
+      "step": 40505
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.384954137641111e-05,
+      "loss": 2.582,
+      "step": 40510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.384805127748925e-05,
+      "loss": 2.4767,
+      "step": 40515
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.384656100991534e-05,
+      "loss": 2.372,
+      "step": 40520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.384507057369509e-05,
+      "loss": 2.5543,
+      "step": 40525
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.384357996883422e-05,
+      "loss": 2.3869,
+      "step": 40530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.384208919533848e-05,
+      "loss": 2.5027,
+      "step": 40535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.384059825321358e-05,
+      "loss": 2.5074,
+      "step": 40540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.383910714246529e-05,
+      "loss": 2.4067,
+      "step": 40545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.383761586309933e-05,
+      "loss": 2.4145,
+      "step": 40550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.383612441512142e-05,
+      "loss": 2.5106,
+      "step": 40555
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.383463279853732e-05,
+      "loss": 2.3563,
+      "step": 40560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.383314101335275e-05,
+      "loss": 2.4518,
+      "step": 40565
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.383164905957347e-05,
+      "loss": 2.3353,
+      "step": 40570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.383015693720519e-05,
+      "loss": 2.5027,
+      "step": 40575
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.382866464625367e-05,
+      "loss": 2.5101,
+      "step": 40580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.382717218672466e-05,
+      "loss": 2.5305,
+      "step": 40585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.382567955862384e-05,
+      "loss": 2.5196,
+      "step": 40590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.382418676195703e-05,
+      "loss": 2.4574,
+      "step": 40595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.382269379672992e-05,
+      "loss": 2.5131,
+      "step": 40600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.382120066294829e-05,
+      "loss": 2.3319,
+      "step": 40605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.381970736061784e-05,
+      "loss": 2.5178,
+      "step": 40610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.381821388974434e-05,
+      "loss": 2.571,
+      "step": 40615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.381672025033352e-05,
+      "loss": 2.6234,
+      "step": 40620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.381522644239116e-05,
+      "loss": 2.493,
+      "step": 40625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.381373246592295e-05,
+      "loss": 2.4681,
+      "step": 40630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.381223832093469e-05,
+      "loss": 2.5523,
+      "step": 40635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.381074400743209e-05,
+      "loss": 2.6026,
+      "step": 40640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.380924952542091e-05,
+      "loss": 2.6233,
+      "step": 40645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.380775487490692e-05,
+      "loss": 2.5947,
+      "step": 40650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.380626005589583e-05,
+      "loss": 2.5048,
+      "step": 40655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.38047650683934e-05,
+      "loss": 2.4735,
+      "step": 40660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.38032699124054e-05,
+      "loss": 2.4941,
+      "step": 40665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.380177458793758e-05,
+      "loss": 2.6221,
+      "step": 40670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.380027909499565e-05,
+      "loss": 2.475,
+      "step": 40675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.379878343358542e-05,
+      "loss": 2.4701,
+      "step": 40680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37972876037126e-05,
+      "loss": 2.3401,
+      "step": 40685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.379579160538295e-05,
+      "loss": 2.5298,
+      "step": 40690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.379429543860224e-05,
+      "loss": 2.3495,
+      "step": 40695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.379279910337622e-05,
+      "loss": 2.4736,
+      "step": 40700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.379130259971065e-05,
+      "loss": 2.597,
+      "step": 40705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.378980592761126e-05,
+      "loss": 2.6106,
+      "step": 40710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.378830908708384e-05,
+      "loss": 2.5018,
+      "step": 40715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.378681207813411e-05,
+      "loss": 2.6087,
+      "step": 40720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.378531490076787e-05,
+      "loss": 2.4962,
+      "step": 40725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.378381755499086e-05,
+      "loss": 2.4984,
+      "step": 40730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.378232004080883e-05,
+      "loss": 2.4722,
+      "step": 40735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.378082235822754e-05,
+      "loss": 2.5775,
+      "step": 40740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.377932450725277e-05,
+      "loss": 2.5437,
+      "step": 40745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.377782648789026e-05,
+      "loss": 2.5157,
+      "step": 40750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37763283001458e-05,
+      "loss": 2.559,
+      "step": 40755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.377482994402511e-05,
+      "loss": 2.5399,
+      "step": 40760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.3773331419534e-05,
+      "loss": 2.49,
+      "step": 40765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.377183272667817e-05,
+      "loss": 2.6756,
+      "step": 40770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.377033386546347e-05,
+      "loss": 2.6029,
+      "step": 40775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.376883483589559e-05,
+      "loss": 2.5295,
+      "step": 40780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.376733563798035e-05,
+      "loss": 2.5178,
+      "step": 40785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.376583627172347e-05,
+      "loss": 2.3379,
+      "step": 40790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.376433673713074e-05,
+      "loss": 2.4458,
+      "step": 40795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.376283703420794e-05,
+      "loss": 2.4437,
+      "step": 40800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37613371629608e-05,
+      "loss": 2.5469,
+      "step": 40805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.375983712339515e-05,
+      "loss": 2.452,
+      "step": 40810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37583369155167e-05,
+      "loss": 2.5357,
+      "step": 40815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.375683653933126e-05,
+      "loss": 2.5986,
+      "step": 40820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.375533599484456e-05,
+      "loss": 2.4595,
+      "step": 40825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.375383528206241e-05,
+      "loss": 2.5728,
+      "step": 40830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.375233440099056e-05,
+      "loss": 2.4635,
+      "step": 40835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37508333516348e-05,
+      "loss": 2.5128,
+      "step": 40840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.374933213400088e-05,
+      "loss": 2.5329,
+      "step": 40845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37478307480946e-05,
+      "loss": 2.5702,
+      "step": 40850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.374632919392172e-05,
+      "loss": 2.594,
+      "step": 40855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.374482747148803e-05,
+      "loss": 2.3667,
+      "step": 40860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.374332558079927e-05,
+      "loss": 2.4456,
+      "step": 40865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.374182352186127e-05,
+      "loss": 2.6258,
+      "step": 40870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.374032129467975e-05,
+      "loss": 2.5147,
+      "step": 40875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.373881889926055e-05,
+      "loss": 2.5791,
+      "step": 40880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.373731633560939e-05,
+      "loss": 2.6145,
+      "step": 40885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.373581360373208e-05,
+      "loss": 2.4055,
+      "step": 40890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.373431070363439e-05,
+      "loss": 2.4718,
+      "step": 40895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.373280763532211e-05,
+      "loss": 2.5613,
+      "step": 40900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.373130439880102e-05,
+      "loss": 2.5227,
+      "step": 40905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.372980099407689e-05,
+      "loss": 2.5776,
+      "step": 40910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.372829742115552e-05,
+      "loss": 2.5905,
+      "step": 40915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.372679368004268e-05,
+      "loss": 2.4165,
+      "step": 40920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.372528977074416e-05,
+      "loss": 2.5447,
+      "step": 40925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.372378569326573e-05,
+      "loss": 2.631,
+      "step": 40930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37222814476132e-05,
+      "loss": 2.534,
+      "step": 40935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.372077703379233e-05,
+      "loss": 2.4124,
+      "step": 40940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.371927245180893e-05,
+      "loss": 2.4649,
+      "step": 40945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.371776770166878e-05,
+      "loss": 2.3506,
+      "step": 40950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.371626278337767e-05,
+      "loss": 2.5103,
+      "step": 40955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.371475769694135e-05,
+      "loss": 2.5254,
+      "step": 40960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.371325244236568e-05,
+      "loss": 2.6126,
+      "step": 40965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.371174701965639e-05,
+      "loss": 2.4662,
+      "step": 40970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.371024142881929e-05,
+      "loss": 2.5435,
+      "step": 40975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.370873566986019e-05,
+      "loss": 2.3688,
+      "step": 40980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.370722974278484e-05,
+      "loss": 2.595,
+      "step": 40985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.370572364759906e-05,
+      "loss": 2.5598,
+      "step": 40990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.370421738430866e-05,
+      "loss": 2.5526,
+      "step": 40995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37027109529194e-05,
+      "loss": 2.4316,
+      "step": 41000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.370120435343707e-05,
+      "loss": 2.5425,
+      "step": 41005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36996975858675e-05,
+      "loss": 2.5193,
+      "step": 41010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.369819065021648e-05,
+      "loss": 2.553,
+      "step": 41015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.369668354648977e-05,
+      "loss": 2.452,
+      "step": 41020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36951762746932e-05,
+      "loss": 2.3882,
+      "step": 41025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.369366883483256e-05,
+      "loss": 2.466,
+      "step": 41030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.369216122691363e-05,
+      "loss": 2.5387,
+      "step": 41035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.369065345094223e-05,
+      "loss": 2.6177,
+      "step": 41040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.368914550692415e-05,
+      "loss": 2.5773,
+      "step": 41045
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36876373948652e-05,
+      "loss": 2.5029,
+      "step": 41050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.368612911477117e-05,
+      "loss": 2.4918,
+      "step": 41055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.368462066664787e-05,
+      "loss": 2.5071,
+      "step": 41060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.368311205050111e-05,
+      "loss": 2.454,
+      "step": 41065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.368160326633667e-05,
+      "loss": 2.6068,
+      "step": 41070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.368009431416033e-05,
+      "loss": 2.4717,
+      "step": 41075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.367858519397796e-05,
+      "loss": 2.5044,
+      "step": 41080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.367707590579533e-05,
+      "loss": 2.4422,
+      "step": 41085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.367556644961825e-05,
+      "loss": 2.5004,
+      "step": 41090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36740568254525e-05,
+      "loss": 2.567,
+      "step": 41095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.367254703330393e-05,
+      "loss": 2.3051,
+      "step": 41100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.367103707317833e-05,
+      "loss": 2.5046,
+      "step": 41105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.366952694508148e-05,
+      "loss": 2.553,
+      "step": 41110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.366801664901922e-05,
+      "loss": 2.5733,
+      "step": 41115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.366650618499735e-05,
+      "loss": 2.412,
+      "step": 41120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36649955530217e-05,
+      "loss": 2.3504,
+      "step": 41125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.366348475309803e-05,
+      "loss": 2.4097,
+      "step": 41130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36619737852322e-05,
+      "loss": 2.4113,
+      "step": 41135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.366046264943e-05,
+      "loss": 2.5911,
+      "step": 41140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.365895134569724e-05,
+      "loss": 2.3593,
+      "step": 41145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.365743987403977e-05,
+      "loss": 2.5124,
+      "step": 41150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.365592823446333e-05,
+      "loss": 2.4896,
+      "step": 41155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.365441642697379e-05,
+      "loss": 2.5938,
+      "step": 41160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.365290445157696e-05,
+      "loss": 2.505,
+      "step": 41165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.365139230827864e-05,
+      "loss": 2.4951,
+      "step": 41170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.364987999708465e-05,
+      "loss": 2.5646,
+      "step": 41175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.364836751800081e-05,
+      "loss": 2.3096,
+      "step": 41180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.364685487103295e-05,
+      "loss": 2.4872,
+      "step": 41185
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.364534205618687e-05,
+      "loss": 2.5096,
+      "step": 41190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.364382907346838e-05,
+      "loss": 2.4763,
+      "step": 41195
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.364231592288332e-05,
+      "loss": 2.4896,
+      "step": 41200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36408026044375e-05,
+      "loss": 2.5995,
+      "step": 41205
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.363928911813676e-05,
+      "loss": 2.5348,
+      "step": 41210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.363777546398689e-05,
+      "loss": 2.6574,
+      "step": 41215
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.363626164199372e-05,
+      "loss": 2.3556,
+      "step": 41220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36347476521631e-05,
+      "loss": 2.4671,
+      "step": 41225
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.363323349450081e-05,
+      "loss": 2.2939,
+      "step": 41230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.363171916901272e-05,
+      "loss": 2.3552,
+      "step": 41235
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.363020467570463e-05,
+      "loss": 2.4017,
+      "step": 41240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.362869001458236e-05,
+      "loss": 2.6051,
+      "step": 41245
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.362717518565174e-05,
+      "loss": 2.4347,
+      "step": 41250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.362566018891859e-05,
+      "loss": 2.4961,
+      "step": 41255
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.362414502438875e-05,
+      "loss": 2.6,
+      "step": 41260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.362262969206805e-05,
+      "loss": 2.3838,
+      "step": 41265
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.362111419196231e-05,
+      "loss": 2.5223,
+      "step": 41270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.361959852407736e-05,
+      "loss": 2.5685,
+      "step": 41275
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.361808268841903e-05,
+      "loss": 2.4033,
+      "step": 41280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.361656668499315e-05,
+      "loss": 2.543,
+      "step": 41285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.361505051380556e-05,
+      "loss": 2.5503,
+      "step": 41290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.361353417486209e-05,
+      "loss": 2.6518,
+      "step": 41295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.361201766816853e-05,
+      "loss": 2.4542,
+      "step": 41300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.361050099373079e-05,
+      "loss": 2.6146,
+      "step": 41305
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.360898415155463e-05,
+      "loss": 2.4609,
+      "step": 41310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.360746714164595e-05,
+      "loss": 2.5336,
+      "step": 41315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.360594996401053e-05,
+      "loss": 2.5196,
+      "step": 41320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.360443261865423e-05,
+      "loss": 2.5419,
+      "step": 41325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.360291510558288e-05,
+      "loss": 2.5065,
+      "step": 41330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.360139742480232e-05,
+      "loss": 2.3613,
+      "step": 41335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.359987957631839e-05,
+      "loss": 2.5087,
+      "step": 41340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.359836156013693e-05,
+      "loss": 2.4614,
+      "step": 41345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.359684337626377e-05,
+      "loss": 2.543,
+      "step": 41350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.359532502470476e-05,
+      "loss": 2.454,
+      "step": 41355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.359380650546572e-05,
+      "loss": 2.524,
+      "step": 41360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.359228781855251e-05,
+      "loss": 2.4165,
+      "step": 41365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.359076896397097e-05,
+      "loss": 2.5016,
+      "step": 41370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.358924994172693e-05,
+      "loss": 2.4139,
+      "step": 41375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.358773075182625e-05,
+      "loss": 2.5574,
+      "step": 41380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.358621139427475e-05,
+      "loss": 2.4699,
+      "step": 41385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35846918690783e-05,
+      "loss": 2.4664,
+      "step": 41390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.358317217624272e-05,
+      "loss": 2.4418,
+      "step": 41395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.358165231577388e-05,
+      "loss": 2.4969,
+      "step": 41400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.358013228767759e-05,
+      "loss": 2.5775,
+      "step": 41405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.357861209195972e-05,
+      "loss": 2.5575,
+      "step": 41410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.357709172862615e-05,
+      "loss": 2.4272,
+      "step": 41415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.357557119768267e-05,
+      "loss": 2.4847,
+      "step": 41420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.357405049913514e-05,
+      "loss": 2.478,
+      "step": 41425
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.357252963298946e-05,
+      "loss": 2.449,
+      "step": 41430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.357100859925141e-05,
+      "loss": 2.4738,
+      "step": 41435
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.356948739792688e-05,
+      "loss": 2.4308,
+      "step": 41440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.356796602902172e-05,
+      "loss": 2.4972,
+      "step": 41445
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.356644449254175e-05,
+      "loss": 2.59,
+      "step": 41450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.356492278849288e-05,
+      "loss": 2.4246,
+      "step": 41455
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35634009168809e-05,
+      "loss": 2.445,
+      "step": 41460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35618788777117e-05,
+      "loss": 2.3661,
+      "step": 41465
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.356035667099113e-05,
+      "loss": 2.5613,
+      "step": 41470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.355883429672505e-05,
+      "loss": 2.6733,
+      "step": 41475
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35573117549193e-05,
+      "loss": 2.5236,
+      "step": 41480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.355578904557974e-05,
+      "loss": 2.4899,
+      "step": 41485
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.355426616871224e-05,
+      "loss": 2.5022,
+      "step": 41490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.355274312432263e-05,
+      "loss": 2.4712,
+      "step": 41495
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35512199124168e-05,
+      "loss": 2.6544,
+      "step": 41500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.354969653300058e-05,
+      "loss": 2.4499,
+      "step": 41505
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.354817298607987e-05,
+      "loss": 2.5528,
+      "step": 41510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.354664927166049e-05,
+      "loss": 2.5531,
+      "step": 41515
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35451253897483e-05,
+      "loss": 2.6109,
+      "step": 41520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.354360134034919e-05,
+      "loss": 2.4932,
+      "step": 41525
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.354207712346901e-05,
+      "loss": 2.6045,
+      "step": 41530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.354055273911362e-05,
+      "loss": 2.3187,
+      "step": 41535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.353902818728887e-05,
+      "loss": 2.5142,
+      "step": 41540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.353750346800064e-05,
+      "loss": 2.5731,
+      "step": 41545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35359785812548e-05,
+      "loss": 2.672,
+      "step": 41550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.353445352705721e-05,
+      "loss": 2.4504,
+      "step": 41555
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.353292830541371e-05,
+      "loss": 2.505,
+      "step": 41560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.353140291633022e-05,
+      "loss": 2.5938,
+      "step": 41565
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.352987735981256e-05,
+      "loss": 2.4732,
+      "step": 41570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.352835163586661e-05,
+      "loss": 2.5882,
+      "step": 41575
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.352682574449825e-05,
+      "loss": 2.5783,
+      "step": 41580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.352529968571333e-05,
+      "loss": 2.5839,
+      "step": 41585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.352377345951775e-05,
+      "loss": 2.4495,
+      "step": 41590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.352224706591734e-05,
+      "loss": 2.3881,
+      "step": 41595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.3520720504918e-05,
+      "loss": 2.5306,
+      "step": 41600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35191937765256e-05,
+      "loss": 2.486,
+      "step": 41605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.3517666880746e-05,
+      "loss": 2.5912,
+      "step": 41610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.351613981758509e-05,
+      "loss": 2.4519,
+      "step": 41615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.351461258704873e-05,
+      "loss": 2.43,
+      "step": 41620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.351308518914277e-05,
+      "loss": 2.4847,
+      "step": 41625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.351155762387314e-05,
+      "loss": 2.5579,
+      "step": 41630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.351002989124567e-05,
+      "loss": 2.5713,
+      "step": 41635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.350850199126626e-05,
+      "loss": 2.3565,
+      "step": 41640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.350697392394078e-05,
+      "loss": 2.442,
+      "step": 41645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.350544568927511e-05,
+      "loss": 2.438,
+      "step": 41650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.350391728727512e-05,
+      "loss": 2.5269,
+      "step": 41655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.350238871794669e-05,
+      "loss": 2.6165,
+      "step": 41660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.350085998129573e-05,
+      "loss": 2.5087,
+      "step": 41665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.349933107732806e-05,
+      "loss": 2.4378,
+      "step": 41670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.34978020060496e-05,
+      "loss": 2.576,
+      "step": 41675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.349627276746622e-05,
+      "loss": 2.538,
+      "step": 41680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.349474336158382e-05,
+      "loss": 2.4885,
+      "step": 41685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.349321378840828e-05,
+      "loss": 2.5247,
+      "step": 41690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.349168404794543e-05,
+      "loss": 2.5111,
+      "step": 41695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.349015414020124e-05,
+      "loss": 2.5206,
+      "step": 41700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.348862406518153e-05,
+      "loss": 2.4875,
+      "step": 41705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.34870938228922e-05,
+      "loss": 2.6359,
+      "step": 41710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.348556341333917e-05,
+      "loss": 2.6752,
+      "step": 41715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.348403283652827e-05,
+      "loss": 2.4166,
+      "step": 41720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.348250209246543e-05,
+      "loss": 2.5752,
+      "step": 41725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.348097118115651e-05,
+      "loss": 2.2929,
+      "step": 41730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.347944010260741e-05,
+      "loss": 2.485,
+      "step": 41735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.347790885682404e-05,
+      "loss": 2.6263,
+      "step": 41740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.347637744381226e-05,
+      "loss": 2.3551,
+      "step": 41745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.347484586357796e-05,
+      "loss": 2.4915,
+      "step": 41750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.347331411612705e-05,
+      "loss": 2.7206,
+      "step": 41755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.34717822014654e-05,
+      "loss": 2.5282,
+      "step": 41760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.347025011959893e-05,
+      "loss": 2.4641,
+      "step": 41765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.34687178705335e-05,
+      "loss": 2.5431,
+      "step": 41770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.346718545427504e-05,
+      "loss": 2.4884,
+      "step": 41775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.346565287082941e-05,
+      "loss": 2.6375,
+      "step": 41780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.346412012020252e-05,
+      "loss": 2.4553,
+      "step": 41785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.346258720240028e-05,
+      "loss": 2.4558,
+      "step": 41790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.346105411742856e-05,
+      "loss": 2.5393,
+      "step": 41795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.345952086529325e-05,
+      "loss": 2.5409,
+      "step": 41800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.345798744600029e-05,
+      "loss": 2.4342,
+      "step": 41805
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.345645385955555e-05,
+      "loss": 2.6017,
+      "step": 41810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.345492010596493e-05,
+      "loss": 2.5714,
+      "step": 41815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.345338618523433e-05,
+      "loss": 2.3389,
+      "step": 41820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.345185209736964e-05,
+      "loss": 2.5364,
+      "step": 41825
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.345031784237679e-05,
+      "loss": 2.4941,
+      "step": 41830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.344878342026164e-05,
+      "loss": 2.5243,
+      "step": 41835
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.344724883103014e-05,
+      "loss": 2.5167,
+      "step": 41840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.344571407468814e-05,
+      "loss": 2.5541,
+      "step": 41845
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.344417915124158e-05,
+      "loss": 2.481,
+      "step": 41850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.344264406069635e-05,
+      "loss": 2.4217,
+      "step": 41855
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.344110880305838e-05,
+      "loss": 2.4761,
+      "step": 41860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.343957337833354e-05,
+      "loss": 2.4835,
+      "step": 41865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.343803778652773e-05,
+      "loss": 2.4246,
+      "step": 41870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.343650202764689e-05,
+      "loss": 2.5815,
+      "step": 41875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.343496610169692e-05,
+      "loss": 2.4961,
+      "step": 41880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.343343000868372e-05,
+      "loss": 2.5043,
+      "step": 41885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.343189374861318e-05,
+      "loss": 2.5931,
+      "step": 41890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.343035732149123e-05,
+      "loss": 2.4166,
+      "step": 41895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.342882072732379e-05,
+      "loss": 2.4915,
+      "step": 41900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.342728396611675e-05,
+      "loss": 2.5074,
+      "step": 41905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.342574703787601e-05,
+      "loss": 2.589,
+      "step": 41910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.342420994260752e-05,
+      "loss": 2.6273,
+      "step": 41915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.342267268031717e-05,
+      "loss": 2.7015,
+      "step": 41920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.342113525101086e-05,
+      "loss": 2.5804,
+      "step": 41925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.341959765469453e-05,
+      "loss": 2.4721,
+      "step": 41930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.341805989137408e-05,
+      "loss": 2.4896,
+      "step": 41935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.341652196105542e-05,
+      "loss": 2.4972,
+      "step": 41940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.341498386374447e-05,
+      "loss": 2.5636,
+      "step": 41945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.341344559944715e-05,
+      "loss": 2.496,
+      "step": 41950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.341190716816936e-05,
+      "loss": 2.534,
+      "step": 41955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.341036856991705e-05,
+      "loss": 2.5392,
+      "step": 41960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.340882980469612e-05,
+      "loss": 2.4202,
+      "step": 41965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.340729087251247e-05,
+      "loss": 2.4954,
+      "step": 41970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.340575177337205e-05,
+      "loss": 2.4912,
+      "step": 41975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.340421250728075e-05,
+      "loss": 2.5246,
+      "step": 41980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.340267307424453e-05,
+      "loss": 2.5159,
+      "step": 41985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.340113347426927e-05,
+      "loss": 2.3419,
+      "step": 41990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.339959370736091e-05,
+      "loss": 2.4489,
+      "step": 41995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.339805377352539e-05,
+      "loss": 2.5573,
+      "step": 42000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33965136727686e-05,
+      "loss": 2.5656,
+      "step": 42005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.339497340509648e-05,
+      "loss": 2.4397,
+      "step": 42010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.339343297051496e-05,
+      "loss": 2.5391,
+      "step": 42015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.339189236902995e-05,
+      "loss": 2.4957,
+      "step": 42020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.339035160064737e-05,
+      "loss": 2.4707,
+      "step": 42025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.338881066537318e-05,
+      "loss": 2.4181,
+      "step": 42030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.338726956321328e-05,
+      "loss": 2.533,
+      "step": 42035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33857282941736e-05,
+      "loss": 2.5556,
+      "step": 42040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.338418685826007e-05,
+      "loss": 2.4005,
+      "step": 42045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.338264525547863e-05,
+      "loss": 2.407,
+      "step": 42050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33811034858352e-05,
+      "loss": 2.5475,
+      "step": 42055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33795615493357e-05,
+      "loss": 2.4545,
+      "step": 42060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.337801944598608e-05,
+      "loss": 2.528,
+      "step": 42065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.337647717579227e-05,
+      "loss": 2.4497,
+      "step": 42070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.337493473876018e-05,
+      "loss": 2.4399,
+      "step": 42075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.337339213489576e-05,
+      "loss": 2.454,
+      "step": 42080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.337184936420496e-05,
+      "loss": 2.5109,
+      "step": 42085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.337030642669367e-05,
+      "loss": 2.6953,
+      "step": 42090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.336876332236786e-05,
+      "loss": 2.6098,
+      "step": 42095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.336722005123344e-05,
+      "loss": 2.563,
+      "step": 42100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.336567661329637e-05,
+      "loss": 2.4541,
+      "step": 42105
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.336413300856257e-05,
+      "loss": 2.6363,
+      "step": 42110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.3362589237038e-05,
+      "loss": 2.5315,
+      "step": 42115
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.336104529872858e-05,
+      "loss": 2.6623,
+      "step": 42120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.335950119364023e-05,
+      "loss": 2.4826,
+      "step": 42125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.335795692177894e-05,
+      "loss": 2.5191,
+      "step": 42130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.335641248315059e-05,
+      "loss": 2.6156,
+      "step": 42135
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.335486787776115e-05,
+      "loss": 2.5269,
+      "step": 42140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.335332310561657e-05,
+      "loss": 2.468,
+      "step": 42145
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.335177816672279e-05,
+      "loss": 2.4572,
+      "step": 42150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.335023306108572e-05,
+      "loss": 2.5295,
+      "step": 42155
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.334868778871135e-05,
+      "loss": 2.4202,
+      "step": 42160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33471423496056e-05,
+      "loss": 2.6157,
+      "step": 42165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33455967437744e-05,
+      "loss": 2.4538,
+      "step": 42170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.334405097122372e-05,
+      "loss": 2.4015,
+      "step": 42175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.334250503195948e-05,
+      "loss": 2.6387,
+      "step": 42180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.334095892598766e-05,
+      "loss": 2.5344,
+      "step": 42185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.333941265331418e-05,
+      "loss": 2.6584,
+      "step": 42190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.333786621394499e-05,
+      "loss": 2.5706,
+      "step": 42195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.333631960788604e-05,
+      "loss": 2.4398,
+      "step": 42200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.333477283514329e-05,
+      "loss": 2.4716,
+      "step": 42205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.333322589572268e-05,
+      "loss": 2.5273,
+      "step": 42210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.333167878963017e-05,
+      "loss": 2.4565,
+      "step": 42215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.333013151687168e-05,
+      "loss": 2.4468,
+      "step": 42220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.332858407745322e-05,
+      "loss": 2.6737,
+      "step": 42225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.332703647138068e-05,
+      "loss": 2.5081,
+      "step": 42230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.332548869866004e-05,
+      "loss": 2.4708,
+      "step": 42235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.332394075929725e-05,
+      "loss": 2.5834,
+      "step": 42240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.332239265329826e-05,
+      "loss": 2.5179,
+      "step": 42245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.332084438066903e-05,
+      "loss": 2.4016,
+      "step": 42250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.331929594141553e-05,
+      "loss": 2.3628,
+      "step": 42255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.331774733554368e-05,
+      "loss": 2.5844,
+      "step": 42260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.331619856305947e-05,
+      "loss": 2.5242,
+      "step": 42265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.331464962396883e-05,
+      "loss": 2.4419,
+      "step": 42270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.331310051827774e-05,
+      "loss": 2.6128,
+      "step": 42275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.331155124599215e-05,
+      "loss": 2.5704,
+      "step": 42280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.331000180711801e-05,
+      "loss": 2.501,
+      "step": 42285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.330845220166128e-05,
+      "loss": 2.3786,
+      "step": 42290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.330690242962796e-05,
+      "loss": 2.5225,
+      "step": 42295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.330535249102396e-05,
+      "loss": 2.606,
+      "step": 42300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.330380238585525e-05,
+      "loss": 2.4908,
+      "step": 42305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.330225211412781e-05,
+      "loss": 2.6291,
+      "step": 42310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.330070167584761e-05,
+      "loss": 2.4338,
+      "step": 42315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.329915107102059e-05,
+      "loss": 2.3854,
+      "step": 42320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.32976002996527e-05,
+      "loss": 2.5663,
+      "step": 42325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.329604936174995e-05,
+      "loss": 2.4425,
+      "step": 42330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.329449825731829e-05,
+      "loss": 2.5186,
+      "step": 42335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.329294698636366e-05,
+      "loss": 2.5421,
+      "step": 42340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.329139554889206e-05,
+      "loss": 2.4254,
+      "step": 42345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.328984394490943e-05,
+      "loss": 2.5412,
+      "step": 42350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.328829217442176e-05,
+      "loss": 2.7268,
+      "step": 42355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.328674023743502e-05,
+      "loss": 2.4008,
+      "step": 42360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.328518813395515e-05,
+      "loss": 2.6125,
+      "step": 42365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.328363586398814e-05,
+      "loss": 2.4523,
+      "step": 42370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.328208342753996e-05,
+      "loss": 2.582,
+      "step": 42375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.328053082461658e-05,
+      "loss": 2.4253,
+      "step": 42380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.327897805522397e-05,
+      "loss": 2.4327,
+      "step": 42385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.32774251193681e-05,
+      "loss": 2.5585,
+      "step": 42390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.327587201705495e-05,
+      "loss": 2.5348,
+      "step": 42395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.32743187482905e-05,
+      "loss": 2.539,
+      "step": 42400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.327276531308071e-05,
+      "loss": 2.3188,
+      "step": 42405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.327121171143156e-05,
+      "loss": 2.4482,
+      "step": 42410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.326965794334902e-05,
+      "loss": 2.4314,
+      "step": 42415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.326810400883907e-05,
+      "loss": 2.4759,
+      "step": 42420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.32665499079077e-05,
+      "loss": 2.5945,
+      "step": 42425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.326499564056088e-05,
+      "loss": 2.4572,
+      "step": 42430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.326344120680458e-05,
+      "loss": 2.421,
+      "step": 42435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.326188660664476e-05,
+      "loss": 2.531,
+      "step": 42440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.326033184008744e-05,
+      "loss": 2.5078,
+      "step": 42445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.325877690713858e-05,
+      "loss": 2.54,
+      "step": 42450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.325722180780416e-05,
+      "loss": 2.5532,
+      "step": 42455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.325566654209017e-05,
+      "loss": 2.3822,
+      "step": 42460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.32541111100026e-05,
+      "loss": 2.5782,
+      "step": 42465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.325255551154739e-05,
+      "loss": 2.3655,
+      "step": 42470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.325099974673057e-05,
+      "loss": 2.506,
+      "step": 42475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.324944381555812e-05,
+      "loss": 2.6316,
+      "step": 42480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.324788771803599e-05,
+      "loss": 2.5834,
+      "step": 42485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.324633145417019e-05,
+      "loss": 2.5717,
+      "step": 42490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.324477502396669e-05,
+      "loss": 2.4477,
+      "step": 42495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.324321842743151e-05,
+      "loss": 2.5579,
+      "step": 42500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.32416616645706e-05,
+      "loss": 2.5492,
+      "step": 42505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.324010473538998e-05,
+      "loss": 2.3503,
+      "step": 42510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.323854763989561e-05,
+      "loss": 2.5256,
+      "step": 42515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.323699037809349e-05,
+      "loss": 2.5231,
+      "step": 42520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.323543294998962e-05,
+      "loss": 2.4886,
+      "step": 42525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.323387535558997e-05,
+      "loss": 2.5414,
+      "step": 42530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.323231759490054e-05,
+      "loss": 2.506,
+      "step": 42535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.323075966792733e-05,
+      "loss": 2.5454,
+      "step": 42540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.322920157467633e-05,
+      "loss": 2.5292,
+      "step": 42545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.322764331515353e-05,
+      "loss": 2.3673,
+      "step": 42550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.322608488936491e-05,
+      "loss": 2.5936,
+      "step": 42555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.322452629731648e-05,
+      "loss": 2.6118,
+      "step": 42560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.322296753901422e-05,
+      "loss": 2.2965,
+      "step": 42565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.322140861446416e-05,
+      "loss": 2.4764,
+      "step": 42570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.321984952367225e-05,
+      "loss": 2.4216,
+      "step": 42575
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.321829026664453e-05,
+      "loss": 2.4875,
+      "step": 42580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.321673084338696e-05,
+      "loss": 2.3666,
+      "step": 42585
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.321517125390555e-05,
+      "loss": 2.5126,
+      "step": 42590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.321361149820632e-05,
+      "loss": 2.5158,
+      "step": 42595
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.321205157629525e-05,
+      "loss": 2.6263,
+      "step": 42600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.321049148817833e-05,
+      "loss": 2.4519,
+      "step": 42605
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.320893123386158e-05,
+      "loss": 2.5658,
+      "step": 42610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.3207370813351e-05,
+      "loss": 2.4055,
+      "step": 42615
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.320581022665257e-05,
+      "loss": 2.4346,
+      "step": 42620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.320424947377233e-05,
+      "loss": 2.4052,
+      "step": 42625
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.320268855471625e-05,
+      "loss": 2.4016,
+      "step": 42630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.320112746949033e-05,
+      "loss": 2.4967,
+      "step": 42635
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.31995662181006e-05,
+      "loss": 2.5268,
+      "step": 42640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.319800480055306e-05,
+      "loss": 2.3687,
+      "step": 42645
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.319644321685371e-05,
+      "loss": 2.5228,
+      "step": 42650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.319488146700856e-05,
+      "loss": 2.5966,
+      "step": 42655
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.319331955102361e-05,
+      "loss": 2.3747,
+      "step": 42660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.319175746890486e-05,
+      "loss": 2.525,
+      "step": 42665
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.319019522065834e-05,
+      "loss": 2.5456,
+      "step": 42670
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.318863280629004e-05,
+      "loss": 2.2884,
+      "step": 42675
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.318707022580598e-05,
+      "loss": 2.6199,
+      "step": 42680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.318550747921217e-05,
+      "loss": 2.4277,
+      "step": 42685
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.318394456651462e-05,
+      "loss": 2.4913,
+      "step": 42690
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.318238148771933e-05,
+      "loss": 2.3645,
+      "step": 42695
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.318081824283233e-05,
+      "loss": 2.6201,
+      "step": 42700
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.317925483185963e-05,
+      "loss": 2.3998,
+      "step": 42705
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.317769125480723e-05,
+      "loss": 2.4494,
+      "step": 42710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.317612751168115e-05,
+      "loss": 2.373,
+      "step": 42715
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.317456360248742e-05,
+      "loss": 2.3992,
+      "step": 42720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.317299952723203e-05,
+      "loss": 2.6957,
+      "step": 42725
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.3171435285921e-05,
+      "loss": 2.3818,
+      "step": 42730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.316987087856038e-05,
+      "loss": 2.643,
+      "step": 42735
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.316830630515612e-05,
+      "loss": 2.4887,
+      "step": 42740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.316674156571431e-05,
+      "loss": 2.3852,
+      "step": 42745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.316517666024094e-05,
+      "loss": 2.4732,
+      "step": 42750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.3163611588742e-05,
+      "loss": 2.581,
+      "step": 42755
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.316204635122356e-05,
+      "loss": 2.5159,
+      "step": 42760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.316048094769161e-05,
+      "loss": 2.5605,
+      "step": 42765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.315891537815218e-05,
+      "loss": 2.4473,
+      "step": 42770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.315734964261128e-05,
+      "loss": 2.4381,
+      "step": 42775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.315578374107494e-05,
+      "loss": 2.646,
+      "step": 42780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.31542176735492e-05,
+      "loss": 2.6146,
+      "step": 42785
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.315265144004005e-05,
+      "loss": 2.5149,
+      "step": 42790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.315108504055354e-05,
+      "loss": 2.3845,
+      "step": 42795
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.314951847509569e-05,
+      "loss": 2.5123,
+      "step": 42800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.314795174367253e-05,
+      "loss": 2.4809,
+      "step": 42805
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.314638484629005e-05,
+      "loss": 2.403,
+      "step": 42810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.314481778295432e-05,
+      "loss": 2.3525,
+      "step": 42815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.314325055367137e-05,
+      "loss": 2.5444,
+      "step": 42820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.314168315844719e-05,
+      "loss": 2.4341,
+      "step": 42825
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.314011559728784e-05,
+      "loss": 2.6212,
+      "step": 42830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.313854787019934e-05,
+      "loss": 2.5128,
+      "step": 42835
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.31369799771877e-05,
+      "loss": 2.4857,
+      "step": 42840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.313541191825899e-05,
+      "loss": 2.5308,
+      "step": 42845
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.31338436934192e-05,
+      "loss": 2.5047,
+      "step": 42850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.313227530267439e-05,
+      "loss": 2.495,
+      "step": 42855
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.313070674603058e-05,
+      "loss": 2.3829,
+      "step": 42860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.312913802349381e-05,
+      "loss": 2.5587,
+      "step": 42865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.312756913507013e-05,
+      "loss": 2.5145,
+      "step": 42870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.312600008076553e-05,
+      "loss": 2.5957,
+      "step": 42875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.312443086058609e-05,
+      "loss": 2.5646,
+      "step": 42880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.312286147453782e-05,
+      "loss": 2.4982,
+      "step": 42885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.312129192262676e-05,
+      "loss": 2.5505,
+      "step": 42890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.311972220485895e-05,
+      "loss": 2.5683,
+      "step": 42895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.311815232124042e-05,
+      "loss": 2.4783,
+      "step": 42900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.311658227177722e-05,
+      "loss": 2.5427,
+      "step": 42905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.31150120564754e-05,
+      "loss": 2.4253,
+      "step": 42910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.311344167534096e-05,
+      "loss": 2.5249,
+      "step": 42915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.311187112837997e-05,
+      "loss": 2.4067,
+      "step": 42920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.311030041559848e-05,
+      "loss": 2.5069,
+      "step": 42925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.31087295370025e-05,
+      "loss": 2.3227,
+      "step": 42930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.310715849259808e-05,
+      "loss": 2.6479,
+      "step": 42935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.31055872823913e-05,
+      "loss": 2.3753,
+      "step": 42940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.310401590638815e-05,
+      "loss": 2.3658,
+      "step": 42945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.31024443645947e-05,
+      "loss": 2.5165,
+      "step": 42950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.3100872657017e-05,
+      "loss": 2.4594,
+      "step": 42955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.309930078366109e-05,
+      "loss": 2.5276,
+      "step": 42960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.3097728744533e-05,
+      "loss": 2.5431,
+      "step": 42965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30961565396388e-05,
+      "loss": 2.5534,
+      "step": 42970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.309458416898453e-05,
+      "loss": 2.5411,
+      "step": 42975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.309301163257623e-05,
+      "loss": 2.22,
+      "step": 42980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.309143893041996e-05,
+      "loss": 2.502,
+      "step": 42985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.308986606252176e-05,
+      "loss": 2.4758,
+      "step": 42990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.308829302888768e-05,
+      "loss": 2.587,
+      "step": 42995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.308671982952377e-05,
+      "loss": 2.4637,
+      "step": 43000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.308514646443609e-05,
+      "loss": 2.5068,
+      "step": 43005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.308357293363068e-05,
+      "loss": 2.6168,
+      "step": 43010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.308199923711361e-05,
+      "loss": 2.4886,
+      "step": 43015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30804253748909e-05,
+      "loss": 2.5225,
+      "step": 43020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.307885134696864e-05,
+      "loss": 2.5784,
+      "step": 43025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.307727715335286e-05,
+      "loss": 2.4191,
+      "step": 43030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.307570279404963e-05,
+      "loss": 2.5772,
+      "step": 43035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.307412826906499e-05,
+      "loss": 2.5932,
+      "step": 43040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.3072553578405e-05,
+      "loss": 2.4933,
+      "step": 43045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.307097872207573e-05,
+      "loss": 2.3618,
+      "step": 43050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.306940370008324e-05,
+      "loss": 2.3986,
+      "step": 43055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.306782851243357e-05,
+      "loss": 2.548,
+      "step": 43060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.306625315913278e-05,
+      "loss": 2.3601,
+      "step": 43065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.306467764018691e-05,
+      "loss": 2.4716,
+      "step": 43070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.306310195560208e-05,
+      "loss": 2.446,
+      "step": 43075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30615261053843e-05,
+      "loss": 2.5147,
+      "step": 43080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.305995008953965e-05,
+      "loss": 2.633,
+      "step": 43085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30583739080742e-05,
+      "loss": 2.478,
+      "step": 43090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.305679756099397e-05,
+      "loss": 2.4702,
+      "step": 43095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.305522104830508e-05,
+      "loss": 2.4282,
+      "step": 43100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.305364437001354e-05,
+      "loss": 2.5668,
+      "step": 43105
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.305206752612547e-05,
+      "loss": 2.4942,
+      "step": 43110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30504905166469e-05,
+      "loss": 2.6395,
+      "step": 43115
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.304891334158388e-05,
+      "loss": 2.5173,
+      "step": 43120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.304733600094252e-05,
+      "loss": 2.4969,
+      "step": 43125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.304575849472885e-05,
+      "loss": 2.4259,
+      "step": 43130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.304418082294897e-05,
+      "loss": 2.4885,
+      "step": 43135
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.304260298560892e-05,
+      "loss": 2.5076,
+      "step": 43140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.304102498271478e-05,
+      "loss": 2.4982,
+      "step": 43145
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.303944681427262e-05,
+      "loss": 2.4958,
+      "step": 43150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.303786848028851e-05,
+      "loss": 2.5916,
+      "step": 43155
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.303628998076852e-05,
+      "loss": 2.6173,
+      "step": 43160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.303471131571872e-05,
+      "loss": 2.4754,
+      "step": 43165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30331324851452e-05,
+      "loss": 2.4331,
+      "step": 43170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.303155348905398e-05,
+      "loss": 2.4417,
+      "step": 43175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.302997432745119e-05,
+      "loss": 2.5273,
+      "step": 43180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.302839500034289e-05,
+      "loss": 2.4457,
+      "step": 43185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.302681550773514e-05,
+      "loss": 2.4736,
+      "step": 43190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.302523584963402e-05,
+      "loss": 2.4336,
+      "step": 43195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.302365602604561e-05,
+      "loss": 2.5497,
+      "step": 43200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.3022076036976e-05,
+      "loss": 2.4076,
+      "step": 43205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.302049588243123e-05,
+      "loss": 2.5785,
+      "step": 43210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.301891556241741e-05,
+      "loss": 2.4463,
+      "step": 43215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.301733507694061e-05,
+      "loss": 2.4303,
+      "step": 43220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30157544260069e-05,
+      "loss": 2.438,
+      "step": 43225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.301417360962237e-05,
+      "loss": 2.6171,
+      "step": 43230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30125926277931e-05,
+      "loss": 2.4762,
+      "step": 43235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.301101148052516e-05,
+      "loss": 2.5186,
+      "step": 43240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.300943016782464e-05,
+      "loss": 2.4159,
+      "step": 43245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.300784868969763e-05,
+      "loss": 2.5814,
+      "step": 43250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30062670461502e-05,
+      "loss": 2.4173,
+      "step": 43255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.300468523718844e-05,
+      "loss": 2.5057,
+      "step": 43260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.300310326281843e-05,
+      "loss": 2.5547,
+      "step": 43265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.300152112304626e-05,
+      "loss": 2.452,
+      "step": 43270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.2999938817878e-05,
+      "loss": 2.4822,
+      "step": 43275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.299835634731977e-05,
+      "loss": 2.4139,
+      "step": 43280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.299677371137762e-05,
+      "loss": 2.4871,
+      "step": 43285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.299519091005765e-05,
+      "loss": 2.5788,
+      "step": 43290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.299360794336596e-05,
+      "loss": 2.4918,
+      "step": 43295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.299202481130861e-05,
+      "loss": 2.4048,
+      "step": 43300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.299044151389173e-05,
+      "loss": 2.4046,
+      "step": 43305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.298885805112138e-05,
+      "loss": 2.3335,
+      "step": 43310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.298727442300366e-05,
+      "loss": 2.6062,
+      "step": 43315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.298569062954465e-05,
+      "loss": 2.4236,
+      "step": 43320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.298410667075046e-05,
+      "loss": 2.334,
+      "step": 43325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.298252254662716e-05,
+      "loss": 2.4397,
+      "step": 43330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.298093825718086e-05,
+      "loss": 2.4403,
+      "step": 43335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.297935380241765e-05,
+      "loss": 2.7024,
+      "step": 43340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.297776918234365e-05,
+      "loss": 2.4951,
+      "step": 43345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29761843969649e-05,
+      "loss": 2.3586,
+      "step": 43350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.297459944628752e-05,
+      "loss": 2.6062,
+      "step": 43355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.297301433031762e-05,
+      "loss": 2.4116,
+      "step": 43360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.297142904906128e-05,
+      "loss": 2.4739,
+      "step": 43365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.296984360252463e-05,
+      "loss": 2.4652,
+      "step": 43370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.296825799071373e-05,
+      "loss": 2.3871,
+      "step": 43375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.296667221363468e-05,
+      "loss": 2.5051,
+      "step": 43380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29650862712936e-05,
+      "loss": 2.5339,
+      "step": 43385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.296350016369659e-05,
+      "loss": 2.4735,
+      "step": 43390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.296191389084972e-05,
+      "loss": 2.4479,
+      "step": 43395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.296032745275913e-05,
+      "loss": 2.4341,
+      "step": 43400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29587408494309e-05,
+      "loss": 2.4197,
+      "step": 43405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.295715408087113e-05,
+      "loss": 2.3354,
+      "step": 43410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.295556714708595e-05,
+      "loss": 2.6026,
+      "step": 43415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.295398004808143e-05,
+      "loss": 2.4846,
+      "step": 43420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.295239278386369e-05,
+      "loss": 2.5309,
+      "step": 43425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.295080535443883e-05,
+      "loss": 2.6928,
+      "step": 43430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.294921775981296e-05,
+      "loss": 2.3552,
+      "step": 43435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.294762999999219e-05,
+      "loss": 2.4372,
+      "step": 43440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.294604207498262e-05,
+      "loss": 2.5165,
+      "step": 43445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.294445398479036e-05,
+      "loss": 2.4433,
+      "step": 43450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.294286572942152e-05,
+      "loss": 2.4045,
+      "step": 43455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29412773088822e-05,
+      "loss": 2.5925,
+      "step": 43460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.293968872317853e-05,
+      "loss": 2.4992,
+      "step": 43465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.293809997231661e-05,
+      "loss": 2.5818,
+      "step": 43470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.293651105630253e-05,
+      "loss": 2.5333,
+      "step": 43475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.293492197514243e-05,
+      "loss": 2.5822,
+      "step": 43480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.293333272884242e-05,
+      "loss": 2.3772,
+      "step": 43485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29317433174086e-05,
+      "loss": 2.6064,
+      "step": 43490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.293015374084707e-05,
+      "loss": 2.3873,
+      "step": 43495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.292856399916397e-05,
+      "loss": 2.5149,
+      "step": 43500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.292697409236541e-05,
+      "loss": 2.4072,
+      "step": 43505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.292538402045749e-05,
+      "loss": 2.66,
+      "step": 43510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.292379378344635e-05,
+      "loss": 2.5791,
+      "step": 43515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29222033813381e-05,
+      "loss": 2.4354,
+      "step": 43520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.292061281413884e-05,
+      "loss": 2.3419,
+      "step": 43525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29190220818547e-05,
+      "loss": 2.5156,
+      "step": 43530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.291743118449179e-05,
+      "loss": 2.3831,
+      "step": 43535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.291584012205625e-05,
+      "loss": 2.4877,
+      "step": 43540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.291424889455418e-05,
+      "loss": 2.4758,
+      "step": 43545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.291265750199169e-05,
+      "loss": 2.5449,
+      "step": 43550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.291106594437494e-05,
+      "loss": 2.4611,
+      "step": 43555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.290947422171002e-05,
+      "loss": 2.5841,
+      "step": 43560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.290788233400306e-05,
+      "loss": 2.5145,
+      "step": 43565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.290629028126019e-05,
+      "loss": 2.5327,
+      "step": 43570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.290469806348751e-05,
+      "loss": 2.4272,
+      "step": 43575
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.290310568069118e-05,
+      "loss": 2.3414,
+      "step": 43580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.290151313287729e-05,
+      "loss": 2.6064,
+      "step": 43585
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.2899920420052e-05,
+      "loss": 2.5405,
+      "step": 43590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.28983275422214e-05,
+      "loss": 2.4765,
+      "step": 43595
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.289673449939165e-05,
+      "loss": 2.556,
+      "step": 43600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.289514129156884e-05,
+      "loss": 2.3689,
+      "step": 43605
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.289354791875914e-05,
+      "loss": 2.4756,
+      "step": 43610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.289195438096865e-05,
+      "loss": 2.5467,
+      "step": 43615
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.289036067820349e-05,
+      "loss": 2.544,
+      "step": 43620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.288876681046983e-05,
+      "loss": 2.6037,
+      "step": 43625
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.288717277777377e-05,
+      "loss": 2.2865,
+      "step": 43630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.288557858012145e-05,
+      "loss": 2.4158,
+      "step": 43635
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.288398421751899e-05,
+      "loss": 2.4249,
+      "step": 43640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.288238968997255e-05,
+      "loss": 2.4796,
+      "step": 43645
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.288079499748823e-05,
+      "loss": 2.4438,
+      "step": 43650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.287920014007219e-05,
+      "loss": 2.4884,
+      "step": 43655
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.287760511773055e-05,
+      "loss": 2.4314,
+      "step": 43660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.287600993046945e-05,
+      "loss": 2.5218,
+      "step": 43665
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.287441457829503e-05,
+      "loss": 2.4107,
+      "step": 43670
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.287281906121339e-05,
+      "loss": 2.507,
+      "step": 43675
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.287122337923072e-05,
+      "loss": 2.3642,
+      "step": 43680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286962753235313e-05,
+      "loss": 2.3816,
+      "step": 43685
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286803152058677e-05,
+      "loss": 2.6138,
+      "step": 43690
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286643534393777e-05,
+      "loss": 2.4276,
+      "step": 43695
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286483900241225e-05,
+      "loss": 2.4823,
+      "step": 43700
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286324249601639e-05,
+      "loss": 2.4625,
+      "step": 43705
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286164582475632e-05,
+      "loss": 2.5355,
+      "step": 43710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286004898863815e-05,
+      "loss": 2.5517,
+      "step": 43715
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.285845198766805e-05,
+      "loss": 2.5038,
+      "step": 43720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.285685482185217e-05,
+      "loss": 2.3698,
+      "step": 43725
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.285525749119664e-05,
+      "loss": 2.565,
+      "step": 43730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.285365999570758e-05,
+      "loss": 2.5372,
+      "step": 43735
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.285206233539116e-05,
+      "loss": 2.4569,
+      "step": 43740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.285046451025354e-05,
+      "loss": 2.4191,
+      "step": 43745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.284886652030084e-05,
+      "loss": 2.4242,
+      "step": 43750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.284726836553924e-05,
+      "loss": 2.3787,
+      "step": 43755
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.284567004597484e-05,
+      "loss": 2.3816,
+      "step": 43760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.284407156161379e-05,
+      "loss": 2.5093,
+      "step": 43765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.284247291246229e-05,
+      "loss": 2.5237,
+      "step": 43770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.284087409852644e-05,
+      "loss": 2.5128,
+      "step": 43775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.283927511981241e-05,
+      "loss": 2.5478,
+      "step": 43780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.283767597632635e-05,
+      "loss": 2.4723,
+      "step": 43785
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.283607666807441e-05,
+      "loss": 2.4357,
+      "step": 43790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.283447719506273e-05,
+      "loss": 2.4937,
+      "step": 43795
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.283287755729748e-05,
+      "loss": 2.5064,
+      "step": 43800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.28312777547848e-05,
+      "loss": 2.5377,
+      "step": 43805
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.282967778753084e-05,
+      "loss": 2.3376,
+      "step": 43810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.282807765554176e-05,
+      "loss": 2.6007,
+      "step": 43815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.282647735882373e-05,
+      "loss": 2.5255,
+      "step": 43820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.282487689738288e-05,
+      "loss": 2.5052,
+      "step": 43825
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.282327627122537e-05,
+      "loss": 2.4993,
+      "step": 43830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.282167548035736e-05,
+      "loss": 2.413,
+      "step": 43835
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.282007452478503e-05,
+      "loss": 2.5209,
+      "step": 43840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.28184734045145e-05,
+      "loss": 2.3901,
+      "step": 43845
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.281687211955195e-05,
+      "loss": 2.4611,
+      "step": 43850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.281527066990354e-05,
+      "loss": 2.4802,
+      "step": 43855
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.281366905557541e-05,
+      "loss": 2.4449,
+      "step": 43860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.281206727657375e-05,
+      "loss": 2.3837,
+      "step": 43865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.28104653329047e-05,
+      "loss": 2.5183,
+      "step": 43870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.280886322457443e-05,
+      "loss": 2.4476,
+      "step": 43875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.28072609515891e-05,
+      "loss": 2.5509,
+      "step": 43880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.280565851395486e-05,
+      "loss": 2.5031,
+      "step": 43885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.280405591167789e-05,
+      "loss": 2.454,
+      "step": 43890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.280245314476436e-05,
+      "loss": 2.5986,
+      "step": 43895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.280085021322042e-05,
+      "loss": 2.5833,
+      "step": 43900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.279924711705223e-05,
+      "loss": 2.4697,
+      "step": 43905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.279764385626596e-05,
+      "loss": 2.4084,
+      "step": 43910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.27960404308678e-05,
+      "loss": 2.3947,
+      "step": 43915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.279443684086389e-05,
+      "loss": 2.4219,
+      "step": 43920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.279283308626042e-05,
+      "loss": 2.4857,
+      "step": 43925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.279122916706353e-05,
+      "loss": 2.4791,
+      "step": 43930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.278962508327942e-05,
+      "loss": 2.5874,
+      "step": 43935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.278802083491422e-05,
+      "loss": 2.5248,
+      "step": 43940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.278641642197415e-05,
+      "loss": 2.5529,
+      "step": 43945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.278481184446535e-05,
+      "loss": 2.4945,
+      "step": 43950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.278320710239399e-05,
+      "loss": 2.4327,
+      "step": 43955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.278160219576625e-05,
+      "loss": 2.4739,
+      "step": 43960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.277999712458832e-05,
+      "loss": 2.4567,
+      "step": 43965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.277839188886634e-05,
+      "loss": 2.4029,
+      "step": 43970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.277678648860651e-05,
+      "loss": 2.398,
+      "step": 43975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.277518092381498e-05,
+      "loss": 2.5702,
+      "step": 43980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.277357519449794e-05,
+      "loss": 2.609,
+      "step": 43985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.277196930066158e-05,
+      "loss": 2.6462,
+      "step": 43990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.277036324231206e-05,
+      "loss": 2.4863,
+      "step": 43995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.276875701945556e-05,
+      "loss": 2.4193,
+      "step": 44000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.276715063209827e-05,
+      "loss": 2.4481,
+      "step": 44005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.276554408024635e-05,
+      "loss": 2.5248,
+      "step": 44010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.276393736390598e-05,
+      "loss": 2.6254,
+      "step": 44015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.276233048308335e-05,
+      "loss": 2.5417,
+      "step": 44020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.276072343778464e-05,
+      "loss": 2.594,
+      "step": 44025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.275911622801603e-05,
+      "loss": 2.5732,
+      "step": 44030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.275750885378369e-05,
+      "loss": 2.3976,
+      "step": 44035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.275590131509382e-05,
+      "loss": 2.4383,
+      "step": 44040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.275429361195261e-05,
+      "loss": 2.5303,
+      "step": 44045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.275268574436622e-05,
+      "loss": 2.4127,
+      "step": 44050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.275107771234084e-05,
+      "loss": 2.4254,
+      "step": 44055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.274946951588266e-05,
+      "loss": 2.5109,
+      "step": 44060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.274786115499785e-05,
+      "loss": 2.4405,
+      "step": 44065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.274625262969263e-05,
+      "loss": 2.4993,
+      "step": 44070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.274464393997315e-05,
+      "loss": 2.5821,
+      "step": 44075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.274303508584564e-05,
+      "loss": 2.6031,
+      "step": 44080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.274142606731623e-05,
+      "loss": 2.5234,
+      "step": 44085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.273981688439118e-05,
+      "loss": 2.5945,
+      "step": 44090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.273820753707661e-05,
+      "loss": 2.4742,
+      "step": 44095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.273659802537876e-05,
+      "loss": 2.4799,
+      "step": 44100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.273498834930379e-05,
+      "loss": 2.6108,
+      "step": 44105
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.27333785088579e-05,
+      "loss": 2.4493,
+      "step": 44110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.27317685040473e-05,
+      "loss": 2.427,
+      "step": 44115
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.273015833487816e-05,
+      "loss": 2.5483,
+      "step": 44120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.272854800135668e-05,
+      "loss": 2.6478,
+      "step": 44125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.272693750348904e-05,
+      "loss": 2.5266,
+      "step": 44130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.272532684128147e-05,
+      "loss": 2.6228,
+      "step": 44135
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.272371601474015e-05,
+      "loss": 2.3817,
+      "step": 44140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.272210502387124e-05,
+      "loss": 2.507,
+      "step": 44145
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.272049386868098e-05,
+      "loss": 2.4556,
+      "step": 44150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.271888254917556e-05,
+      "loss": 2.4274,
+      "step": 44155
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.271727106536117e-05,
+      "loss": 2.4785,
+      "step": 44160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.2715659417244e-05,
+      "loss": 2.5166,
+      "step": 44165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.271404760483027e-05,
+      "loss": 2.5448,
+      "step": 44170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.271243562812614e-05,
+      "loss": 2.4433,
+      "step": 44175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.271082348713787e-05,
+      "loss": 2.4531,
+      "step": 44180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.270921118187161e-05,
+      "loss": 2.5044,
+      "step": 44185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.270759871233358e-05,
+      "loss": 2.5211,
+      "step": 44190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.270598607852999e-05,
+      "loss": 2.4792,
+      "step": 44195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.270437328046703e-05,
+      "loss": 2.4989,
+      "step": 44200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.270276031815091e-05,
+      "loss": 2.5679,
+      "step": 44205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.270114719158782e-05,
+      "loss": 2.5558,
+      "step": 44210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.269953390078399e-05,
+      "loss": 2.501,
+      "step": 44215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.26979204457456e-05,
+      "loss": 2.4947,
+      "step": 44220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.269630682647885e-05,
+      "loss": 2.4278,
+      "step": 44225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.269469304298998e-05,
+      "loss": 2.4432,
+      "step": 44230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.26930790952852e-05,
+      "loss": 2.4737,
+      "step": 44235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.269146498337067e-05,
+      "loss": 2.4301,
+      "step": 44240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.268985070725264e-05,
+      "loss": 2.3845,
+      "step": 44245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.268823626693731e-05,
+      "loss": 2.485,
+      "step": 44250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.268662166243087e-05,
+      "loss": 2.4568,
+      "step": 44255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.268500689373955e-05,
+      "loss": 2.4361,
+      "step": 44260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.268339196086956e-05,
+      "loss": 2.5891,
+      "step": 44265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.268177686382711e-05,
+      "loss": 2.4885,
+      "step": 44270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.26801616026184e-05,
+      "loss": 2.3317,
+      "step": 44275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.267854617724965e-05,
+      "loss": 2.4752,
+      "step": 44280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.267693058772708e-05,
+      "loss": 2.5367,
+      "step": 44285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.267531483405691e-05,
+      "loss": 2.4901,
+      "step": 44290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.267369891624534e-05,
+      "loss": 2.5092,
+      "step": 44295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.267208283429858e-05,
+      "loss": 2.4681,
+      "step": 44300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.267046658822286e-05,
+      "loss": 2.4177,
+      "step": 44305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.266885017802439e-05,
+      "loss": 2.496,
+      "step": 44310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.266723360370939e-05,
+      "loss": 2.4199,
+      "step": 44315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.26656168652841e-05,
+      "loss": 2.5093,
+      "step": 44320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.266399996275468e-05,
+      "loss": 2.4971,
+      "step": 44325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.26623828961274e-05,
+      "loss": 2.3616,
+      "step": 44330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.266076566540847e-05,
+      "loss": 2.5015,
+      "step": 44335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.26591482706041e-05,
+      "loss": 2.3437,
+      "step": 44340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.265753071172049e-05,
+      "loss": 2.5404,
+      "step": 44345
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.265591298876392e-05,
+      "loss": 2.5121,
+      "step": 44350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.265429510174057e-05,
+      "loss": 2.373,
+      "step": 44355
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.265267705065668e-05,
+      "loss": 2.4812,
+      "step": 44360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.265105883551845e-05,
+      "loss": 2.498,
+      "step": 44365
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.264944045633212e-05,
+      "loss": 2.4623,
+      "step": 44370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.264782191310392e-05,
+      "loss": 2.464,
+      "step": 44375
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.264620320584007e-05,
+      "loss": 2.5331,
+      "step": 44380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.264458433454682e-05,
+      "loss": 2.5258,
+      "step": 44385
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.264296529923033e-05,
+      "loss": 2.463,
+      "step": 44390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.26413460998969e-05,
+      "loss": 2.371,
+      "step": 44395
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.263972673655271e-05,
+      "loss": 2.4624,
+      "step": 44400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.263810720920401e-05,
+      "loss": 2.5358,
+      "step": 44405
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.263648751785703e-05,
+      "loss": 2.3411,
+      "step": 44410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.2634867662518e-05,
+      "loss": 2.5256,
+      "step": 44415
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.263324764319315e-05,
+      "loss": 2.5063,
+      "step": 44420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.263162745988868e-05,
+      "loss": 2.6504,
+      "step": 44425
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.263000711261087e-05,
+      "loss": 2.1986,
+      "step": 44430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.262838660136592e-05,
+      "loss": 2.4252,
+      "step": 44435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.262676592616009e-05,
+      "loss": 2.4581,
+      "step": 44440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.262514508699957e-05,
+      "loss": 2.5091,
+      "step": 44445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.262352408389065e-05,
+      "loss": 2.4581,
+      "step": 44450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.262190291683951e-05,
+      "loss": 2.491,
+      "step": 44455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.262028158585243e-05,
+      "loss": 2.5455,
+      "step": 44460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.261866009093563e-05,
+      "loss": 2.4743,
+      "step": 44465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.261703843209533e-05,
+      "loss": 2.4978,
+      "step": 44470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.261541660933779e-05,
+      "loss": 2.6496,
+      "step": 44475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.261379462266924e-05,
+      "loss": 2.4834,
+      "step": 44480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.261217247209592e-05,
+      "loss": 2.4977,
+      "step": 44485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.261055015762406e-05,
+      "loss": 2.4987,
+      "step": 44490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.260892767925991e-05,
+      "loss": 2.5266,
+      "step": 44495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.260730503700972e-05,
+      "loss": 2.3658,
+      "step": 44500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.260568223087971e-05,
+      "loss": 2.5115,
+      "step": 44505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.260405926087612e-05,
+      "loss": 2.4175,
+      "step": 44510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.260243612700522e-05,
+      "loss": 2.5181,
+      "step": 44515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.260081282927324e-05,
+      "loss": 2.5515,
+      "step": 44520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.259918936768641e-05,
+      "loss": 2.5051,
+      "step": 44525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.259756574225097e-05,
+      "loss": 2.3042,
+      "step": 44530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.259594195297321e-05,
+      "loss": 2.5157,
+      "step": 44535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.259431799985933e-05,
+      "loss": 2.4357,
+      "step": 44540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.259269388291559e-05,
+      "loss": 2.6492,
+      "step": 44545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.259106960214825e-05,
+      "loss": 2.4381,
+      "step": 44550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.258944515756353e-05,
+      "loss": 2.4491,
+      "step": 44555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.258782054916769e-05,
+      "loss": 2.3933,
+      "step": 44560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.258619577696698e-05,
+      "loss": 2.4981,
+      "step": 44565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.258457084096768e-05,
+      "loss": 2.4735,
+      "step": 44570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.258294574117598e-05,
+      "loss": 2.4459,
+      "step": 44575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.258132047759818e-05,
+      "loss": 2.6085,
+      "step": 44580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.257969505024052e-05,
+      "loss": 2.4704,
+      "step": 44585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.257806945910923e-05,
+      "loss": 2.4753,
+      "step": 44590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.257644370421057e-05,
+      "loss": 2.542,
+      "step": 44595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.257481778555081e-05,
+      "loss": 2.5676,
+      "step": 44600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25731917031362e-05,
+      "loss": 2.4886,
+      "step": 44605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.257156545697298e-05,
+      "loss": 2.4235,
+      "step": 44610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.256993904706743e-05,
+      "loss": 2.5173,
+      "step": 44615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.256831247342577e-05,
+      "loss": 2.4499,
+      "step": 44620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.256668573605429e-05,
+      "loss": 2.4645,
+      "step": 44625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.256505883495922e-05,
+      "loss": 2.6402,
+      "step": 44630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.256343177014683e-05,
+      "loss": 2.6496,
+      "step": 44635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25618045416234e-05,
+      "loss": 2.5398,
+      "step": 44640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.256017714939515e-05,
+      "loss": 2.5601,
+      "step": 44645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.255854959346837e-05,
+      "loss": 2.4484,
+      "step": 44650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.255692187384928e-05,
+      "loss": 2.4586,
+      "step": 44655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.255529399054419e-05,
+      "loss": 2.4625,
+      "step": 44660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.255366594355934e-05,
+      "loss": 2.6692,
+      "step": 44665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.255203773290098e-05,
+      "loss": 2.5686,
+      "step": 44670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.255040935857537e-05,
+      "loss": 2.4088,
+      "step": 44675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25487808205888e-05,
+      "loss": 2.4041,
+      "step": 44680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.254715211894753e-05,
+      "loss": 2.4274,
+      "step": 44685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.254552325365781e-05,
+      "loss": 2.4562,
+      "step": 44690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.254389422472591e-05,
+      "loss": 2.4983,
+      "step": 44695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.254226503215808e-05,
+      "loss": 2.436,
+      "step": 44700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25406356759606e-05,
+      "loss": 2.3594,
+      "step": 44705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.253900615613977e-05,
+      "loss": 2.4452,
+      "step": 44710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25373764727018e-05,
+      "loss": 2.6191,
+      "step": 44715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.2535746625653e-05,
+      "loss": 2.4604,
+      "step": 44720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.253411661499962e-05,
+      "loss": 2.5453,
+      "step": 44725
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.253248644074793e-05,
+      "loss": 2.5051,
+      "step": 44730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25308561029042e-05,
+      "loss": 2.2936,
+      "step": 44735
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25292256014747e-05,
+      "loss": 2.3698,
+      "step": 44740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25275949364657e-05,
+      "loss": 2.5948,
+      "step": 44745
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.252596410788349e-05,
+      "loss": 2.4579,
+      "step": 44750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.252433311573432e-05,
+      "loss": 2.3964,
+      "step": 44755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.252270196002448e-05,
+      "loss": 2.5614,
+      "step": 44760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.252107064076023e-05,
+      "loss": 2.5296,
+      "step": 44765
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.251943915794787e-05,
+      "loss": 2.523,
+      "step": 44770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.251780751159365e-05,
+      "loss": 2.4791,
+      "step": 44775
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.251617570170383e-05,
+      "loss": 2.3215,
+      "step": 44780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.251454372828473e-05,
+      "loss": 2.4549,
+      "step": 44785
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25129115913426e-05,
+      "loss": 2.4175,
+      "step": 44790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.251127929088372e-05,
+      "loss": 2.3376,
+      "step": 44795
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.250964682691438e-05,
+      "loss": 2.5517,
+      "step": 44800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.250801419944083e-05,
+      "loss": 2.5108,
+      "step": 44805
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25063814084694e-05,
+      "loss": 2.5328,
+      "step": 44810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.250474845400631e-05,
+      "loss": 2.4407,
+      "step": 44815
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.25031153360579e-05,
+      "loss": 2.4374,
+      "step": 44820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.250148205463041e-05,
+      "loss": 2.5084,
+      "step": 44825
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.249984860973013e-05,
+      "loss": 2.5205,
+      "step": 44830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.249821500136336e-05,
+      "loss": 2.4375,
+      "step": 44835
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.249658122953635e-05,
+      "loss": 2.286,
+      "step": 44840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.249494729425542e-05,
+      "loss": 2.5029,
+      "step": 44845
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.249331319552683e-05,
+      "loss": 2.4391,
+      "step": 44850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.249167893335689e-05,
+      "loss": 2.4687,
+      "step": 44855
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.249004450775184e-05,
+      "loss": 2.4232,
+      "step": 44860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.248840991871803e-05,
+      "loss": 2.4727,
+      "step": 44865
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.24867751662617e-05,
+      "loss": 2.4535,
+      "step": 44870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.248514025038915e-05,
+      "loss": 2.5639,
+      "step": 44875
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.248350517110666e-05,
+      "loss": 2.5583,
+      "step": 44880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.248186992842056e-05,
+      "loss": 2.4671,
+      "step": 44885
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.248023452233708e-05,
+      "loss": 2.4375,
+      "step": 44890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.247859895286255e-05,
+      "loss": 2.4435,
+      "step": 44895
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.247696322000324e-05,
+      "loss": 2.5211,
+      "step": 44900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.247532732376546e-05,
+      "loss": 2.5603,
+      "step": 44905
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.247369126415549e-05,
+      "loss": 2.5311,
+      "step": 44910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.247205504117963e-05,
+      "loss": 2.4744,
+      "step": 44915
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.247041865484416e-05,
+      "loss": 2.496,
+      "step": 44920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.246878210515539e-05,
+      "loss": 2.5086,
+      "step": 44925
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.24671453921196e-05,
+      "loss": 2.5619,
+      "step": 44930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.246550851574312e-05,
+      "loss": 2.5785,
+      "step": 44935
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.246387147603219e-05,
+      "loss": 2.4394,
+      "step": 44940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.246223427299315e-05,
+      "loss": 2.5962,
+      "step": 44945
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.246059690663227e-05,
+      "loss": 2.5145,
+      "step": 44950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.245895937695586e-05,
+      "loss": 2.4981,
+      "step": 44955
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.245732168397023e-05,
+      "loss": 2.5555,
+      "step": 44960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.245568382768166e-05,
+      "loss": 2.5253,
+      "step": 44965
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.245404580809647e-05,
+      "loss": 2.4904,
+      "step": 44970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.245240762522093e-05,
+      "loss": 2.4261,
+      "step": 44975
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.245076927906138e-05,
+      "loss": 2.4242,
+      "step": 44980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.244913076962409e-05,
+      "loss": 2.4248,
+      "step": 44985
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.244749209691537e-05,
+      "loss": 2.5146,
+      "step": 44990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.244585326094153e-05,
+      "loss": 2.5308,
+      "step": 44995
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.244421426170886e-05,
+      "loss": 2.459,
+      "step": 45000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.244257509922368e-05,
+      "loss": 2.3218,
+      "step": 45005
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.244093577349228e-05,
+      "loss": 2.5513,
+      "step": 45010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.243929628452097e-05,
+      "loss": 2.5777,
+      "step": 45015
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.243765663231608e-05,
+      "loss": 2.4246,
+      "step": 45020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.243601681688387e-05,
+      "loss": 2.4812,
+      "step": 45025
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.243437683823068e-05,
+      "loss": 2.6024,
+      "step": 45030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.243273669636283e-05,
+      "loss": 2.477,
+      "step": 45035
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.243109639128659e-05,
+      "loss": 2.4505,
+      "step": 45040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.24294559230083e-05,
+      "loss": 2.5677,
+      "step": 45045
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.242781529153424e-05,
+      "loss": 2.4894,
+      "step": 45050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.242617449687074e-05,
+      "loss": 2.5127,
+      "step": 45055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.242453353902411e-05,
+      "loss": 2.4774,
+      "step": 45060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.242289241800068e-05,
+      "loss": 2.499,
+      "step": 45065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.242125113380673e-05,
+      "loss": 2.5516,
+      "step": 45070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.241960968644858e-05,
+      "loss": 2.4568,
+      "step": 45075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.241796807593255e-05,
+      "loss": 2.5542,
+      "step": 45080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.241632630226497e-05,
+      "loss": 2.3733,
+      "step": 45085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.241468436545211e-05,
+      "loss": 2.4239,
+      "step": 45090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.241304226550034e-05,
+      "loss": 2.5052,
+      "step": 45095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.241140000241592e-05,
+      "loss": 2.403,
+      "step": 45100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.240975757620522e-05,
+      "loss": 2.4342,
+      "step": 45105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.240811498687453e-05,
+      "loss": 2.4903,
+      "step": 45110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.240647223443016e-05,
+      "loss": 2.4735,
+      "step": 45115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.240482931887845e-05,
+      "loss": 2.3812,
+      "step": 45120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.24031862402257e-05,
+      "loss": 2.554,
+      "step": 45125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.240154299847825e-05,
+      "loss": 2.4079,
+      "step": 45130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.23998995936424e-05,
+      "loss": 2.6252,
+      "step": 45135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.239825602572447e-05,
+      "loss": 2.3388,
+      "step": 45140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.239661229473081e-05,
+      "loss": 2.3908,
+      "step": 45145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.239496840066772e-05,
+      "loss": 2.5602,
+      "step": 45150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.239332434354151e-05,
+      "loss": 2.4535,
+      "step": 45155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.239168012335854e-05,
+      "loss": 2.4521,
+      "step": 45160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.239003574012511e-05,
+      "loss": 2.4757,
+      "step": 45165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.238839119384754e-05,
+      "loss": 2.4077,
+      "step": 45170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.238674648453217e-05,
+      "loss": 2.3215,
+      "step": 45175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.23851016121853e-05,
+      "loss": 2.6804,
+      "step": 45180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.23834565768133e-05,
+      "loss": 2.3679,
+      "step": 45185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.238181137842248e-05,
+      "loss": 2.5328,
+      "step": 45190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.238016601701915e-05,
+      "loss": 2.4487,
+      "step": 45195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.237852049260964e-05,
+      "loss": 2.4087,
+      "step": 45200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.237687480520032e-05,
+      "loss": 2.5841,
+      "step": 45205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.237522895479747e-05,
+      "loss": 2.4814,
+      "step": 45210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.237358294140743e-05,
+      "loss": 2.3525,
+      "step": 45215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.237193676503655e-05,
+      "loss": 2.4018,
+      "step": 45220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.237029042569116e-05,
+      "loss": 2.471,
+      "step": 45225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.236864392337758e-05,
+      "loss": 2.3944,
+      "step": 45230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.236699725810214e-05,
+      "loss": 2.403,
+      "step": 45235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.23653504298712e-05,
+      "loss": 2.5127,
+      "step": 45240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.236370343869105e-05,
+      "loss": 2.4318,
+      "step": 45245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.236205628456806e-05,
+      "loss": 2.4959,
+      "step": 45250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.236040896750857e-05,
+      "loss": 2.5283,
+      "step": 45255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.23587614875189e-05,
+      "loss": 2.581,
+      "step": 45260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.235711384460537e-05,
+      "loss": 2.4066,
+      "step": 45265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.235546603877434e-05,
+      "loss": 2.5143,
+      "step": 45270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.235381807003215e-05,
+      "loss": 2.3935,
+      "step": 45275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.235216993838514e-05,
+      "loss": 2.4534,
+      "step": 45280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.235052164383964e-05,
+      "loss": 2.6045,
+      "step": 45285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.234887318640197e-05,
+      "loss": 2.4149,
+      "step": 45290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.234722456607852e-05,
+      "loss": 2.4831,
+      "step": 45295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.234557578287559e-05,
+      "loss": 2.6692,
+      "step": 45300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.234392683679954e-05,
+      "loss": 2.4777,
+      "step": 45305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.234227772785668e-05,
+      "loss": 2.3795,
+      "step": 45310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.23406284560534e-05,
+      "loss": 2.4858,
+      "step": 45315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.233897902139604e-05,
+      "loss": 2.5603,
+      "step": 45320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.233732942389091e-05,
+      "loss": 2.54,
+      "step": 45325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.233567966354436e-05,
+      "loss": 2.34,
+      "step": 45330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.233402974036276e-05,
+      "loss": 2.537,
+      "step": 45335
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.233237965435244e-05,
+      "loss": 2.4027,
+      "step": 45340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.233072940551977e-05,
+      "loss": 2.5258,
+      "step": 45345
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.232907899387106e-05,
+      "loss": 2.54,
+      "step": 45350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.232742841941267e-05,
+      "loss": 2.5593,
+      "step": 45355
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.232577768215096e-05,
+      "loss": 2.5382,
+      "step": 45360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.232412678209228e-05,
+      "loss": 2.5389,
+      "step": 45365
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.232247571924297e-05,
+      "loss": 2.4687,
+      "step": 45370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.232082449360938e-05,
+      "loss": 2.4622,
+      "step": 45375
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.231917310519787e-05,
+      "loss": 2.5683,
+      "step": 45380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.231752155401478e-05,
+      "loss": 2.4898,
+      "step": 45385
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.231586984006648e-05,
+      "loss": 2.4876,
+      "step": 45390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.23142179633593e-05,
+      "loss": 2.3933,
+      "step": 45395
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.231256592389961e-05,
+      "loss": 2.5825,
+      "step": 45400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.231091372169375e-05,
+      "loss": 2.5305,
+      "step": 45405
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.23092613567481e-05,
+      "loss": 2.5092,
+      "step": 45410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.2307608829069e-05,
+      "loss": 2.4745,
+      "step": 45415
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.230595613866281e-05,
+      "loss": 2.5657,
+      "step": 45420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.230430328553588e-05,
+      "loss": 2.4878,
+      "step": 45425
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.230265026969456e-05,
+      "loss": 2.3583,
+      "step": 45430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.230099709114521e-05,
+      "loss": 2.5596,
+      "step": 45435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.229934374989422e-05,
+      "loss": 2.6537,
+      "step": 45440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.229769024594792e-05,
+      "loss": 2.4323,
+      "step": 45445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.229603657931267e-05,
+      "loss": 2.5607,
+      "step": 45450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.229438274999482e-05,
+      "loss": 2.5586,
+      "step": 45455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.229272875800077e-05,
+      "loss": 2.4482,
+      "step": 45460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.229107460333686e-05,
+      "loss": 2.4027,
+      "step": 45465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.228942028600944e-05,
+      "loss": 2.5406,
+      "step": 45470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22877658060249e-05,
+      "loss": 2.4594,
+      "step": 45475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.228611116338956e-05,
+      "loss": 2.539,
+      "step": 45480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.228445635810984e-05,
+      "loss": 2.4938,
+      "step": 45485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.228280139019206e-05,
+      "loss": 2.5017,
+      "step": 45490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.228114625964261e-05,
+      "loss": 2.4745,
+      "step": 45495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.227949096646787e-05,
+      "loss": 2.608,
+      "step": 45500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.227783551067415e-05,
+      "loss": 2.5543,
+      "step": 45505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.227617989226787e-05,
+      "loss": 2.4501,
+      "step": 45510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.227452411125536e-05,
+      "loss": 2.4195,
+      "step": 45515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.227286816764304e-05,
+      "loss": 2.6267,
+      "step": 45520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.227121206143724e-05,
+      "loss": 2.4189,
+      "step": 45525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.226955579264434e-05,
+      "loss": 2.521,
+      "step": 45530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22678993612707e-05,
+      "loss": 2.5987,
+      "step": 45535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22662427673227e-05,
+      "loss": 2.5284,
+      "step": 45540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22645860108067e-05,
+      "loss": 2.4393,
+      "step": 45545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.226292909172911e-05,
+      "loss": 2.4933,
+      "step": 45550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.226127201009627e-05,
+      "loss": 2.5177,
+      "step": 45555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.225961476591457e-05,
+      "loss": 2.6508,
+      "step": 45560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.225795735919034e-05,
+      "loss": 2.4765,
+      "step": 45565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.225629978993001e-05,
+      "loss": 2.395,
+      "step": 45570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.225464205813995e-05,
+      "loss": 2.4266,
+      "step": 45575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22529841638265e-05,
+      "loss": 2.3765,
+      "step": 45580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.225132610699606e-05,
+      "loss": 2.518,
+      "step": 45585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.224966788765502e-05,
+      "loss": 2.579,
+      "step": 45590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.224800950580973e-05,
+      "loss": 2.4737,
+      "step": 45595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.224635096146659e-05,
+      "loss": 2.4854,
+      "step": 45600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.224469225463196e-05,
+      "loss": 2.4459,
+      "step": 45605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.224303338531223e-05,
+      "loss": 2.4233,
+      "step": 45610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22413743535138e-05,
+      "loss": 2.4735,
+      "step": 45615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.223971515924303e-05,
+      "loss": 2.3779,
+      "step": 45620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.223805580250629e-05,
+      "loss": 2.514,
+      "step": 45625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.223639628330999e-05,
+      "loss": 2.5593,
+      "step": 45630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.223473660166048e-05,
+      "loss": 2.4392,
+      "step": 45635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.223307675756418e-05,
+      "loss": 2.4542,
+      "step": 45640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.223141675102746e-05,
+      "loss": 2.5566,
+      "step": 45645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22297565820567e-05,
+      "loss": 2.5858,
+      "step": 45650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22280962506583e-05,
+      "loss": 2.4238,
+      "step": 45655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22264357568386e-05,
+      "loss": 2.3913,
+      "step": 45660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.222477510060405e-05,
+      "loss": 2.5761,
+      "step": 45665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.2223114281961e-05,
+      "loss": 2.5194,
+      "step": 45670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.222145330091584e-05,
+      "loss": 2.4867,
+      "step": 45675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.221979215747498e-05,
+      "loss": 2.549,
+      "step": 45680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.221813085164478e-05,
+      "loss": 2.4372,
+      "step": 45685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.221646938343164e-05,
+      "loss": 2.4878,
+      "step": 45690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.221480775284197e-05,
+      "loss": 2.4681,
+      "step": 45695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.221314595988213e-05,
+      "loss": 2.5677,
+      "step": 45700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.221148400455855e-05,
+      "loss": 2.3688,
+      "step": 45705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.220982188687758e-05,
+      "loss": 2.5081,
+      "step": 45710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.220815960684564e-05,
+      "loss": 2.4644,
+      "step": 45715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22064971644691e-05,
+      "loss": 2.4707,
+      "step": 45720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.220483455975439e-05,
+      "loss": 2.475,
+      "step": 45725
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.220317179270788e-05,
+      "loss": 2.5648,
+      "step": 45730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.220150886333597e-05,
+      "loss": 2.4461,
+      "step": 45735
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.219984577164505e-05,
+      "loss": 2.4278,
+      "step": 45740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.219818251764154e-05,
+      "loss": 2.5225,
+      "step": 45745
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.219651910133181e-05,
+      "loss": 2.5065,
+      "step": 45750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.219485552272226e-05,
+      "loss": 2.3768,
+      "step": 45755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.219319178181931e-05,
+      "loss": 2.6743,
+      "step": 45760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.219152787862934e-05,
+      "loss": 2.3925,
+      "step": 45765
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.218986381315876e-05,
+      "loss": 2.4126,
+      "step": 45770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.218819958541397e-05,
+      "loss": 2.406,
+      "step": 45775
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.218653519540137e-05,
+      "loss": 2.4673,
+      "step": 45780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.218487064312735e-05,
+      "loss": 2.3663,
+      "step": 45785
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.218320592859833e-05,
+      "loss": 2.4213,
+      "step": 45790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.21815410518207e-05,
+      "loss": 2.4946,
+      "step": 45795
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.217987601280087e-05,
+      "loss": 2.5178,
+      "step": 45800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.217821081154525e-05,
+      "loss": 2.4241,
+      "step": 45805
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.217654544806022e-05,
+      "loss": 2.4986,
+      "step": 45810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.217487992235221e-05,
+      "loss": 2.448,
+      "step": 45815
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.217321423442763e-05,
+      "loss": 2.6049,
+      "step": 45820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.217154838429287e-05,
+      "loss": 2.4737,
+      "step": 45825
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.216988237195434e-05,
+      "loss": 2.4974,
+      "step": 45830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.216821619741845e-05,
+      "loss": 2.5114,
+      "step": 45835
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.216654986069162e-05,
+      "loss": 2.4035,
+      "step": 45840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.216488336178024e-05,
+      "loss": 2.4678,
+      "step": 45845
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.216321670069072e-05,
+      "loss": 2.4839,
+      "step": 45850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.216154987742949e-05,
+      "loss": 2.6288,
+      "step": 45855
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.215988289200297e-05,
+      "loss": 2.5808,
+      "step": 45860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.215821574441752e-05,
+      "loss": 2.5492,
+      "step": 45865
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.21565484346796e-05,
+      "loss": 2.5393,
+      "step": 45870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.21548809627956e-05,
+      "loss": 2.4863,
+      "step": 45875
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.215321332877195e-05,
+      "loss": 2.5051,
+      "step": 45880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.215154553261505e-05,
+      "loss": 2.5793,
+      "step": 45885
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.214987757433131e-05,
+      "loss": 2.4065,
+      "step": 45890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.214820945392717e-05,
+      "loss": 2.557,
+      "step": 45895
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.214654117140904e-05,
+      "loss": 2.5144,
+      "step": 45900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.214487272678331e-05,
+      "loss": 2.462,
+      "step": 45905
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.214320412005642e-05,
+      "loss": 2.517,
+      "step": 45910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.214153535123479e-05,
+      "loss": 2.546,
+      "step": 45915
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.213986642032482e-05,
+      "loss": 2.4713,
+      "step": 45920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.213819732733296e-05,
+      "loss": 2.4994,
+      "step": 45925
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.213652807226558e-05,
+      "loss": 2.5628,
+      "step": 45930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.213485865512916e-05,
+      "loss": 2.3983,
+      "step": 45935
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.213318907593008e-05,
+      "loss": 2.3335,
+      "step": 45940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.213151933467478e-05,
+      "loss": 2.5066,
+      "step": 45945
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.212984943136967e-05,
+      "loss": 2.4194,
+      "step": 45950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.212817936602118e-05,
+      "loss": 2.5119,
+      "step": 45955
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.212650913863574e-05,
+      "loss": 2.5211,
+      "step": 45960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.212483874921975e-05,
+      "loss": 2.4513,
+      "step": 45965
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.212316819777966e-05,
+      "loss": 2.537,
+      "step": 45970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.21214974843219e-05,
+      "loss": 2.5073,
+      "step": 45975
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.211982660885288e-05,
+      "loss": 2.5341,
+      "step": 45980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.211815557137902e-05,
+      "loss": 2.5228,
+      "step": 45985
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.211648437190676e-05,
+      "loss": 2.6229,
+      "step": 45990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.211481301044253e-05,
+      "loss": 2.3498,
+      "step": 45995
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.211314148699275e-05,
+      "loss": 2.4981,
+      "step": 46000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.211146980156385e-05,
+      "loss": 2.5531,
+      "step": 46005
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.210979795416225e-05,
+      "loss": 2.5318,
+      "step": 46010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.210812594479442e-05,
+      "loss": 2.4177,
+      "step": 46015
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.210645377346676e-05,
+      "loss": 2.5014,
+      "step": 46020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.210478144018568e-05,
+      "loss": 2.415,
+      "step": 46025
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.210310894495767e-05,
+      "loss": 2.4794,
+      "step": 46030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.210143628778912e-05,
+      "loss": 2.5022,
+      "step": 46035
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.209976346868648e-05,
+      "loss": 2.3963,
+      "step": 46040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.209809048765616e-05,
+      "loss": 2.4412,
+      "step": 46045
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.209641734470462e-05,
+      "loss": 2.4518,
+      "step": 46050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.209474403983829e-05,
+      "loss": 2.4818,
+      "step": 46055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20930705730636e-05,
+      "loss": 2.5965,
+      "step": 46060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.2091396944387e-05,
+      "loss": 2.4994,
+      "step": 46065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.208972315381493e-05,
+      "loss": 2.4229,
+      "step": 46070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.208804920135378e-05,
+      "loss": 2.5609,
+      "step": 46075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.208637508701006e-05,
+      "loss": 2.2497,
+      "step": 46080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.208470081079015e-05,
+      "loss": 2.4637,
+      "step": 46085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.208302637270054e-05,
+      "loss": 2.4764,
+      "step": 46090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.208135177274761e-05,
+      "loss": 2.5069,
+      "step": 46095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.207967701093787e-05,
+      "loss": 2.2347,
+      "step": 46100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20780020872777e-05,
+      "loss": 2.5682,
+      "step": 46105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.207632700177358e-05,
+      "loss": 2.399,
+      "step": 46110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.207465175443194e-05,
+      "loss": 2.4982,
+      "step": 46115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.207297634525923e-05,
+      "loss": 2.4723,
+      "step": 46120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20713007742619e-05,
+      "loss": 2.4529,
+      "step": 46125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.206962504144636e-05,
+      "loss": 2.4435,
+      "step": 46130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20679491468191e-05,
+      "loss": 2.4303,
+      "step": 46135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.206627309038652e-05,
+      "loss": 2.5613,
+      "step": 46140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.206459687215511e-05,
+      "loss": 2.527,
+      "step": 46145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20629204921313e-05,
+      "loss": 2.4768,
+      "step": 46150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.206124395032153e-05,
+      "loss": 2.5597,
+      "step": 46155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.205956724673226e-05,
+      "loss": 2.4504,
+      "step": 46160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.205789038136994e-05,
+      "loss": 2.5016,
+      "step": 46165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.2056213354241e-05,
+      "loss": 2.5678,
+      "step": 46170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.205453616535191e-05,
+      "loss": 2.4416,
+      "step": 46175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20528588147091e-05,
+      "loss": 2.4165,
+      "step": 46180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.205118130231907e-05,
+      "loss": 2.6413,
+      "step": 46185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.204950362818822e-05,
+      "loss": 2.4367,
+      "step": 46190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.204782579232303e-05,
+      "loss": 2.3519,
+      "step": 46195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.204614779472994e-05,
+      "loss": 2.5433,
+      "step": 46200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.204446963541542e-05,
+      "loss": 2.5353,
+      "step": 46205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20427913143859e-05,
+      "loss": 2.3349,
+      "step": 46210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.204111283164784e-05,
+      "loss": 2.5554,
+      "step": 46215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.203943418720773e-05,
+      "loss": 2.5046,
+      "step": 46220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.2037755381072e-05,
+      "loss": 2.6297,
+      "step": 46225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.203607641324709e-05,
+      "loss": 2.4426,
+      "step": 46230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.203439728373951e-05,
+      "loss": 2.3928,
+      "step": 46235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.203271799255564e-05,
+      "loss": 2.3779,
+      "step": 46240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.203103853970202e-05,
+      "loss": 2.4643,
+      "step": 46245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.202935892518507e-05,
+      "loss": 2.4225,
+      "step": 46250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.202767914901124e-05,
+      "loss": 2.4521,
+      "step": 46255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.202599921118702e-05,
+      "loss": 2.5783,
+      "step": 46260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.202431911171885e-05,
+      "loss": 2.5455,
+      "step": 46265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.202263885061319e-05,
+      "loss": 2.6023,
+      "step": 46270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.202095842787652e-05,
+      "loss": 2.5308,
+      "step": 46275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20192778435153e-05,
+      "loss": 2.5047,
+      "step": 46280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.201759709753598e-05,
+      "loss": 2.54,
+      "step": 46285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.201591618994504e-05,
+      "loss": 2.3191,
+      "step": 46290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.201423512074894e-05,
+      "loss": 2.4344,
+      "step": 46295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.201255388995414e-05,
+      "loss": 2.4057,
+      "step": 46300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20108724975671e-05,
+      "loss": 2.4647,
+      "step": 46305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.200919094359433e-05,
+      "loss": 2.5139,
+      "step": 46310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.200750922804224e-05,
+      "loss": 2.6056,
+      "step": 46315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.200582735091733e-05,
+      "loss": 2.4906,
+      "step": 46320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.200414531222606e-05,
+      "loss": 2.4447,
+      "step": 46325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.200246311197491e-05,
+      "loss": 2.5045,
+      "step": 46330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.200078075017034e-05,
+      "loss": 2.5048,
+      "step": 46335
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.199909822681882e-05,
+      "loss": 2.6232,
+      "step": 46340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.199741554192684e-05,
+      "loss": 2.3879,
+      "step": 46345
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.199573269550085e-05,
+      "loss": 2.5088,
+      "step": 46350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.199404968754732e-05,
+      "loss": 2.4966,
+      "step": 46355
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.199236651807275e-05,
+      "loss": 2.4488,
+      "step": 46360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19906831870836e-05,
+      "loss": 2.5685,
+      "step": 46365
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.198899969458632e-05,
+      "loss": 2.5597,
+      "step": 46370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.198731604058743e-05,
+      "loss": 2.4669,
+      "step": 46375
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.198563222509337e-05,
+      "loss": 2.435,
+      "step": 46380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.198394824811062e-05,
+      "loss": 2.3582,
+      "step": 46385
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19822641096457e-05,
+      "loss": 2.533,
+      "step": 46390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.198057980970502e-05,
+      "loss": 2.4354,
+      "step": 46395
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.197889534829512e-05,
+      "loss": 2.5758,
+      "step": 46400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.197721072542244e-05,
+      "loss": 2.4699,
+      "step": 46405
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.197552594109346e-05,
+      "loss": 2.6937,
+      "step": 46410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19738409953147e-05,
+      "loss": 2.457,
+      "step": 46415
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.197215588809257e-05,
+      "loss": 2.5096,
+      "step": 46420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.197047061943363e-05,
+      "loss": 2.4271,
+      "step": 46425
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.196878518934431e-05,
+      "loss": 2.5223,
+      "step": 46430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.196709959783112e-05,
+      "loss": 2.6083,
+      "step": 46435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.196541384490052e-05,
+      "loss": 2.4466,
+      "step": 46440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.196372793055901e-05,
+      "loss": 2.4697,
+      "step": 46445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.196204185481306e-05,
+      "loss": 2.5301,
+      "step": 46450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.196035561766918e-05,
+      "loss": 2.4379,
+      "step": 46455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.195866921913383e-05,
+      "loss": 2.4062,
+      "step": 46460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19569826592135e-05,
+      "loss": 2.3994,
+      "step": 46465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19552959379147e-05,
+      "loss": 2.4602,
+      "step": 46470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.195360905524391e-05,
+      "loss": 2.4347,
+      "step": 46475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.195192201120759e-05,
+      "loss": 2.4212,
+      "step": 46480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.195023480581224e-05,
+      "loss": 2.4546,
+      "step": 46485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.194854743906438e-05,
+      "loss": 2.367,
+      "step": 46490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.194685991097048e-05,
+      "loss": 2.4969,
+      "step": 46495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.194517222153701e-05,
+      "loss": 2.4688,
+      "step": 46500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19434843707705e-05,
+      "loss": 2.4119,
+      "step": 46505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.194179635867741e-05,
+      "loss": 2.3568,
+      "step": 46510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.194010818526425e-05,
+      "loss": 2.5601,
+      "step": 46515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19384198505375e-05,
+      "loss": 2.4034,
+      "step": 46520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.193673135450367e-05,
+      "loss": 2.5024,
+      "step": 46525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.193504269716924e-05,
+      "loss": 2.5498,
+      "step": 46530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.193335387854072e-05,
+      "loss": 2.476,
+      "step": 46535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.193166489862459e-05,
+      "loss": 2.3578,
+      "step": 46540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.192997575742735e-05,
+      "loss": 2.504,
+      "step": 46545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.192828645495552e-05,
+      "loss": 2.4918,
+      "step": 46550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.192659699121555e-05,
+      "loss": 2.5312,
+      "step": 46555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.192490736621398e-05,
+      "loss": 2.4464,
+      "step": 46560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19232175799573e-05,
+      "loss": 2.4583,
+      "step": 46565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.192152763245202e-05,
+      "loss": 2.5273,
+      "step": 46570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.191983752370459e-05,
+      "loss": 2.5432,
+      "step": 46575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.191814725372156e-05,
+      "loss": 2.4685,
+      "step": 46580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19164568225094e-05,
+      "loss": 2.5028,
+      "step": 46585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.191476623007464e-05,
+      "loss": 2.4604,
+      "step": 46590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.191307547642379e-05,
+      "loss": 2.4724,
+      "step": 46595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19113845615633e-05,
+      "loss": 2.583,
+      "step": 46600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.190969348549974e-05,
+      "loss": 2.6412,
+      "step": 46605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.190800224823956e-05,
+      "loss": 2.5496,
+      "step": 46610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.190631084978929e-05,
+      "loss": 2.4841,
+      "step": 46615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.190461929015543e-05,
+      "loss": 2.5061,
+      "step": 46620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19029275693445e-05,
+      "loss": 2.5019,
+      "step": 46625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.190123568736299e-05,
+      "loss": 2.4483,
+      "step": 46630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.189954364421741e-05,
+      "loss": 2.4618,
+      "step": 46635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.189785143991426e-05,
+      "loss": 2.6338,
+      "step": 46640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.189615907446008e-05,
+      "loss": 2.5948,
+      "step": 46645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.189446654786135e-05,
+      "loss": 2.4186,
+      "step": 46650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.189277386012457e-05,
+      "loss": 2.644,
+      "step": 46655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.18910810112563e-05,
+      "loss": 2.4127,
+      "step": 46660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.188938800126301e-05,
+      "loss": 2.4846,
+      "step": 46665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.188769483015123e-05,
+      "loss": 2.5249,
+      "step": 46670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.188600149792745e-05,
+      "loss": 2.366,
+      "step": 46675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.18843080045982e-05,
+      "loss": 2.45,
+      "step": 46680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.188261435017e-05,
+      "loss": 2.3768,
+      "step": 46685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.188092053464934e-05,
+      "loss": 2.4698,
+      "step": 46690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.187922655804277e-05,
+      "loss": 2.4929,
+      "step": 46695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.187753242035677e-05,
+      "loss": 2.5425,
+      "step": 46700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.187583812159788e-05,
+      "loss": 2.542,
+      "step": 46705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.187414366177262e-05,
+      "loss": 2.4999,
+      "step": 46710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.187244904088747e-05,
+      "loss": 2.5233,
+      "step": 46715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.1870754258949e-05,
+      "loss": 2.5394,
+      "step": 46720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.186905931596369e-05,
+      "loss": 2.4536,
+      "step": 46725
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.186736421193807e-05,
+      "loss": 2.4747,
+      "step": 46730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.186566894687867e-05,
+      "loss": 2.516,
+      "step": 46735
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.1863973520792e-05,
+      "loss": 2.4624,
+      "step": 46740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.186227793368458e-05,
+      "loss": 2.436,
+      "step": 46745
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.186058218556295e-05,
+      "loss": 2.3427,
+      "step": 46750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.18588862764336e-05,
+      "loss": 2.5398,
+      "step": 46755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.185719020630309e-05,
+      "loss": 2.3586,
+      "step": 46760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.18554939751779e-05,
+      "loss": 2.5814,
+      "step": 46765
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.185379758306458e-05,
+      "loss": 2.4191,
+      "step": 46770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.185210102996966e-05,
+      "loss": 2.4605,
+      "step": 46775
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.185040431589967e-05,
+      "loss": 2.5977,
+      "step": 46780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.184870744086112e-05,
+      "loss": 2.5194,
+      "step": 46785
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.184701040486053e-05,
+      "loss": 2.4515,
+      "step": 46790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.184531320790445e-05,
+      "loss": 2.3266,
+      "step": 46795
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.184361584999938e-05,
+      "loss": 2.522,
+      "step": 46800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.18419183311519e-05,
+      "loss": 2.5344,
+      "step": 46805
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.184022065136848e-05,
+      "loss": 2.5028,
+      "step": 46810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.183852281065567e-05,
+      "loss": 2.5582,
+      "step": 46815
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.183682480902e-05,
+      "loss": 2.4995,
+      "step": 46820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.183512664646803e-05,
+      "loss": 2.5592,
+      "step": 46825
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.183342832300625e-05,
+      "loss": 2.4598,
+      "step": 46830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.18317298386412e-05,
+      "loss": 2.5236,
+      "step": 46835
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.183003119337944e-05,
+      "loss": 2.5565,
+      "step": 46840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.182833238722747e-05,
+      "loss": 2.5691,
+      "step": 46845
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.182663342019187e-05,
+      "loss": 2.5549,
+      "step": 46850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.182493429227912e-05,
+      "loss": 2.5516,
+      "step": 46855
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.182323500349577e-05,
+      "loss": 2.5545,
+      "step": 46860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.182153555384838e-05,
+      "loss": 2.416,
+      "step": 46865
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.181983594334347e-05,
+      "loss": 2.411,
+      "step": 46870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.181813617198758e-05,
+      "loss": 2.5624,
+      "step": 46875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.181643623978726e-05,
+      "loss": 2.5677,
+      "step": 46880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.1814736146749e-05,
+      "loss": 2.4685,
+      "step": 46885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.18130358928794e-05,
+      "loss": 2.5113,
+      "step": 46890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.181133547818498e-05,
+      "loss": 2.5357,
+      "step": 46895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.180963490267227e-05,
+      "loss": 2.5815,
+      "step": 46900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.18079341663478e-05,
+      "loss": 2.5396,
+      "step": 46905
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.180623326921815e-05,
+      "loss": 2.5344,
+      "step": 46910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.180453221128981e-05,
+      "loss": 2.4926,
+      "step": 46915
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.180283099256938e-05,
+      "loss": 2.4948,
+      "step": 46920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.180112961306336e-05,
+      "loss": 2.3809,
+      "step": 46925
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.179942807277832e-05,
+      "loss": 2.5146,
+      "step": 46930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.179772637172078e-05,
+      "loss": 2.5207,
+      "step": 46935
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.17960245098973e-05,
+      "loss": 2.532,
+      "step": 46940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.179432248731445e-05,
+      "loss": 2.6748,
+      "step": 46945
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.179262030397873e-05,
+      "loss": 2.4403,
+      "step": 46950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.179091795989671e-05,
+      "loss": 2.3024,
+      "step": 46955
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.178921545507492e-05,
+      "loss": 2.4049,
+      "step": 46960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.178751278951996e-05,
+      "loss": 2.4263,
+      "step": 46965
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.178580996323831e-05,
+      "loss": 2.46,
+      "step": 46970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.178410697623657e-05,
+      "loss": 2.4684,
+      "step": 46975
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.178240382852128e-05,
+      "loss": 2.5516,
+      "step": 46980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.178070052009898e-05,
+      "loss": 2.4363,
+      "step": 46985
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.17789970509762e-05,
+      "loss": 2.2991,
+      "step": 46990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.177729342115955e-05,
+      "loss": 2.3849,
+      "step": 46995
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.177558963065553e-05,
+      "loss": 2.3792,
+      "step": 47000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 253379,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "total_flos": 7.040172784877568e+18,
+  "trial_name": null,
+  "trial_params": null
+}