diff --git "a/last-checkpoint/trainer_state.json" "b/last-checkpoint/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/last-checkpoint/trainer_state.json"
@@ -0,0 +1,16048 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999744863680664,
+  "global_step": 26723,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001235516251971484,
+      "loss": 9.7101,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016074437039452377,
+      "loss": 7.6038,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001825007316338653,
+      "loss": 7.042,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001979371155918992,
+      "loss": 6.9961,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020991050519692135,
+      "loss": 6.9923,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021969347683124067,
+      "loss": 6.9805,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022796486149183456,
+      "loss": 7.0688,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002351298607892746,
+      "loss": 6.9749,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00024144983807058217,
+      "loss": 6.8547,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002471032503942968,
+      "loss": 6.7304,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002522173839168694,
+      "loss": 6.5608,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002568862220286161,
+      "loss": 6.3845,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00026118113675221766,
+      "loss": 6.2594,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026515760668920996,
+      "loss": 6.103,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026885961163363825,
+      "loss": 5.9774,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027232260598665,
+      "loss": 5.9004,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027557558915551237,
+      "loss": 5.84,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002786425832679576,
+      "loss": 5.8035,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028154371082157617,
+      "loss": 5.7301,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028429599559167215,
+      "loss": 5.6815,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002869139679285514,
+      "loss": 5.6691,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002894101291142448,
+      "loss": 5.6405,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029179531241331795,
+      "loss": 5.6051,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002940789672259915,
+      "loss": 5.5766,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002962693851966943,
+      "loss": 5.5354,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002983738819495931,
+      "loss": 5.5009,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.487,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.4546,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.427,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.4093,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.3689,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.3502,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.3148,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.2891,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.2705,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.237,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.2159,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.1819,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.1519,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 5.1317,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 5.1057,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 5.0917,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 5.0573,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 5.037,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.9967,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.974,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.9363,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.9044,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.8957,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.8705,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.8355,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.7903,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.7654,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.7788,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.7238,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.702,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.677,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.644,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.6304,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.5954,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.5857,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.5478,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.5277,
+      "step": 630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.5062,
+      "step": 640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.489,
+      "step": 650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003,
+      "loss": 4.4573,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.424,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.4252,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.4776,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.4074,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.3595,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.3665,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.3289,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.2975,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.2768,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.2522,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.2516,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.2293,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.1977,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.1959,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.1727,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.1453,
+      "step": 820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.1332,
+      "step": 830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.1316,
+      "step": 840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.0866,
+      "step": 850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.0641,
+      "step": 860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.0743,
+      "step": 870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.0448,
+      "step": 880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.0229,
+      "step": 890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.0076,
+      "step": 900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 4.001,
+      "step": 910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 3.9774,
+      "step": 920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003,
+      "loss": 3.9594,
+      "step": 930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.9476,
+      "step": 940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.9371,
+      "step": 950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.9347,
+      "step": 960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.8973,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.8849,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.8597,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.8536,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.8369,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.8365,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.8008,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.8174,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.7789,
+      "step": 1050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.7691,
+      "step": 1060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.757,
+      "step": 1070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.7431,
+      "step": 1080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.7431,
+      "step": 1090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.7317,
+      "step": 1100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.7134,
+      "step": 1110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.6965,
+      "step": 1120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.6835,
+      "step": 1130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.684,
+      "step": 1140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.6692,
+      "step": 1150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.6632,
+      "step": 1160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.6516,
+      "step": 1170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.6382,
+      "step": 1180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.6217,
+      "step": 1190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003,
+      "loss": 3.6105,
+      "step": 1200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.6118,
+      "step": 1210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.6149,
+      "step": 1220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5861,
+      "step": 1230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5845,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5751,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5727,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5613,
+      "step": 1270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5416,
+      "step": 1280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5428,
+      "step": 1290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5352,
+      "step": 1300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5174,
+      "step": 1310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5158,
+      "step": 1320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.5054,
+      "step": 1330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4948,
+      "step": 1340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4971,
+      "step": 1350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.501,
+      "step": 1360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4902,
+      "step": 1370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4768,
+      "step": 1380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4724,
+      "step": 1390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4574,
+      "step": 1400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4646,
+      "step": 1410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4541,
+      "step": 1420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4468,
+      "step": 1430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4446,
+      "step": 1440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4323,
+      "step": 1450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003,
+      "loss": 3.4257,
+      "step": 1460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.4307,
+      "step": 1470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.4166,
+      "step": 1480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.406,
+      "step": 1490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.4004,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3979,
+      "step": 1510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3915,
+      "step": 1520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3901,
+      "step": 1530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3987,
+      "step": 1540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3813,
+      "step": 1550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.375,
+      "step": 1560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.377,
+      "step": 1570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3655,
+      "step": 1580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.359,
+      "step": 1590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3493,
+      "step": 1600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3476,
+      "step": 1610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3609,
+      "step": 1620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3517,
+      "step": 1630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3347,
+      "step": 1640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3278,
+      "step": 1650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3354,
+      "step": 1660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3221,
+      "step": 1670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3239,
+      "step": 1680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3175,
+      "step": 1690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3084,
+      "step": 1700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3133,
+      "step": 1710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.3025,
+      "step": 1720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003,
+      "loss": 3.2993,
+      "step": 1730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2994,
+      "step": 1740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.3022,
+      "step": 1750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2921,
+      "step": 1760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2971,
+      "step": 1770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2882,
+      "step": 1780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.278,
+      "step": 1790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.28,
+      "step": 1800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2779,
+      "step": 1810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2742,
+      "step": 1820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2669,
+      "step": 1830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2578,
+      "step": 1840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2598,
+      "step": 1850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2536,
+      "step": 1860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2594,
+      "step": 1870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2395,
+      "step": 1880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2412,
+      "step": 1890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2359,
+      "step": 1900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2464,
+      "step": 1910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.247,
+      "step": 1920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2366,
+      "step": 1930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.235,
+      "step": 1940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2274,
+      "step": 1950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2349,
+      "step": 1960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.22,
+      "step": 1970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2229,
+      "step": 1980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2194,
+      "step": 1990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003,
+      "loss": 3.2101,
+      "step": 2000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.2106,
+      "step": 2010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.2018,
+      "step": 2020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1991,
+      "step": 2030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.2054,
+      "step": 2040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.2011,
+      "step": 2050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1911,
+      "step": 2060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1809,
+      "step": 2070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1955,
+      "step": 2080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1863,
+      "step": 2090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1924,
+      "step": 2100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1831,
+      "step": 2110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1798,
+      "step": 2120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1778,
+      "step": 2130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1795,
+      "step": 2140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1753,
+      "step": 2150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.167,
+      "step": 2160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1646,
+      "step": 2170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1656,
+      "step": 2180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1595,
+      "step": 2190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1704,
+      "step": 2200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1622,
+      "step": 2210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1548,
+      "step": 2220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1536,
+      "step": 2230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1501,
+      "step": 2240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1503,
+      "step": 2250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1496,
+      "step": 2260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003,
+      "loss": 3.1491,
+      "step": 2270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1401,
+      "step": 2280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1425,
+      "step": 2290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1426,
+      "step": 2300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1376,
+      "step": 2310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1237,
+      "step": 2320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1415,
+      "step": 2330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1379,
+      "step": 2340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1286,
+      "step": 2350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1329,
+      "step": 2360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1293,
+      "step": 2370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1182,
+      "step": 2380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1235,
+      "step": 2390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1252,
+      "step": 2400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1226,
+      "step": 2410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1227,
+      "step": 2420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1138,
+      "step": 2430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1184,
+      "step": 2440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.104,
+      "step": 2450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1097,
+      "step": 2460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1034,
+      "step": 2470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.1031,
+      "step": 2480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.0951,
+      "step": 2490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.0954,
+      "step": 2500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.0961,
+      "step": 2510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.0986,
+      "step": 2520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003,
+      "loss": 3.0915,
+      "step": 2530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.1012,
+      "step": 2540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.1305,
+      "step": 2550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.1001,
+      "step": 2560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0923,
+      "step": 2570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0938,
+      "step": 2580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0851,
+      "step": 2590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0769,
+      "step": 2600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0782,
+      "step": 2610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0896,
+      "step": 2620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.078,
+      "step": 2630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0731,
+      "step": 2640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0629,
+      "step": 2650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0884,
+      "step": 2660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0863,
+      "step": 2670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0682,
+      "step": 2680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0736,
+      "step": 2690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0601,
+      "step": 2700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.072,
+      "step": 2710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0646,
+      "step": 2720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0616,
+      "step": 2730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0591,
+      "step": 2740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0561,
+      "step": 2750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0694,
+      "step": 2760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0541,
+      "step": 2770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0598,
+      "step": 2780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0644,
+      "step": 2790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003,
+      "loss": 3.0512,
+      "step": 2800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.05,
+      "step": 2810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0437,
+      "step": 2820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0477,
+      "step": 2830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.054,
+      "step": 2840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0417,
+      "step": 2850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0322,
+      "step": 2860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.037,
+      "step": 2870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0487,
+      "step": 2880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0413,
+      "step": 2890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.037,
+      "step": 2900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0297,
+      "step": 2910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0293,
+      "step": 2920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0369,
+      "step": 2930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0356,
+      "step": 2940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0341,
+      "step": 2950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0304,
+      "step": 2960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0358,
+      "step": 2970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0374,
+      "step": 2980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0293,
+      "step": 2990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0252,
+      "step": 3000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0296,
+      "step": 3010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0266,
+      "step": 3020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0284,
+      "step": 3030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0161,
+      "step": 3040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0153,
+      "step": 3050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.0302,
+      "step": 3060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003,
+      "loss": 3.016,
+      "step": 3070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0181,
+      "step": 3080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0141,
+      "step": 3090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0241,
+      "step": 3100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0155,
+      "step": 3110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0083,
+      "step": 3120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0106,
+      "step": 3130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0027,
+      "step": 3140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0116,
+      "step": 3150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0082,
+      "step": 3160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0061,
+      "step": 3170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0005,
+      "step": 3180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9999,
+      "step": 3190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0029,
+      "step": 3200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0003,
+      "step": 3210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9998,
+      "step": 3220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0168,
+      "step": 3230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9882,
+      "step": 3240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9955,
+      "step": 3250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9942,
+      "step": 3260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9953,
+      "step": 3270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0024,
+      "step": 3280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 3.0011,
+      "step": 3290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9952,
+      "step": 3300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9903,
+      "step": 3310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9934,
+      "step": 3320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9849,
+      "step": 3330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003,
+      "loss": 2.9893,
+      "step": 3340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9915,
+      "step": 3350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9946,
+      "step": 3360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9749,
+      "step": 3370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9845,
+      "step": 3380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9836,
+      "step": 3390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9805,
+      "step": 3400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9829,
+      "step": 3410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.982,
+      "step": 3420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9782,
+      "step": 3430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9758,
+      "step": 3440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.97,
+      "step": 3450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9744,
+      "step": 3460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9799,
+      "step": 3470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9707,
+      "step": 3480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9814,
+      "step": 3490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.977,
+      "step": 3500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9708,
+      "step": 3510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9658,
+      "step": 3520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9625,
+      "step": 3530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9666,
+      "step": 3540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9654,
+      "step": 3550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9676,
+      "step": 3560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9701,
+      "step": 3570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9626,
+      "step": 3580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9567,
+      "step": 3590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": 2.9653,
+      "step": 3600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9521,
+      "step": 3610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9594,
+      "step": 3620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9615,
+      "step": 3630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9589,
+      "step": 3640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9598,
+      "step": 3650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9559,
+      "step": 3660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9593,
+      "step": 3670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9549,
+      "step": 3680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9543,
+      "step": 3690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9539,
+      "step": 3700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9524,
+      "step": 3710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9554,
+      "step": 3720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9518,
+      "step": 3730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9544,
+      "step": 3740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9424,
+      "step": 3750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9606,
+      "step": 3760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9437,
+      "step": 3770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9439,
+      "step": 3780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.943,
+      "step": 3790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9502,
+      "step": 3800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9392,
+      "step": 3810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9507,
+      "step": 3820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9315,
+      "step": 3830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9512,
+      "step": 3840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9379,
+      "step": 3850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9555,
+      "step": 3860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003,
+      "loss": 2.9359,
+      "step": 3870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9396,
+      "step": 3880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9409,
+      "step": 3890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9348,
+      "step": 3900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9323,
+      "step": 3910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9345,
+      "step": 3920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9323,
+      "step": 3930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9402,
+      "step": 3940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.93,
+      "step": 3950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9306,
+      "step": 3960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9373,
+      "step": 3970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9371,
+      "step": 3980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9221,
+      "step": 3990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 3.0766,
+      "step": 4000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9973,
+      "step": 4010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9502,
+      "step": 4020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9465,
+      "step": 4030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9326,
+      "step": 4040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9321,
+      "step": 4050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9218,
+      "step": 4060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9208,
+      "step": 4070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9284,
+      "step": 4080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9228,
+      "step": 4090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9272,
+      "step": 4100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9145,
+      "step": 4110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9182,
+      "step": 4120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9334,
+      "step": 4130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003,
+      "loss": 2.9081,
+      "step": 4140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9205,
+      "step": 4150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9101,
+      "step": 4160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9127,
+      "step": 4170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.916,
+      "step": 4180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9245,
+      "step": 4190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9174,
+      "step": 4200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9082,
+      "step": 4210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9126,
+      "step": 4220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9121,
+      "step": 4230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9221,
+      "step": 4240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9057,
+      "step": 4250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.913,
+      "step": 4260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9073,
+      "step": 4270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9098,
+      "step": 4280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9138,
+      "step": 4290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9133,
+      "step": 4300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9051,
+      "step": 4310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9157,
+      "step": 4320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.8976,
+      "step": 4330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9031,
+      "step": 4340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9091,
+      "step": 4350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9073,
+      "step": 4360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9099,
+      "step": 4370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9064,
+      "step": 4380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.9066,
+      "step": 4390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003,
+      "loss": 2.906,
+      "step": 4400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8988,
+      "step": 4410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.9052,
+      "step": 4420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.9027,
+      "step": 4430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.9017,
+      "step": 4440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.902,
+      "step": 4450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.899,
+      "step": 4460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.9038,
+      "step": 4470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8957,
+      "step": 4480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.9066,
+      "step": 4490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.9017,
+      "step": 4500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8884,
+      "step": 4510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8987,
+      "step": 4520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8889,
+      "step": 4530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8952,
+      "step": 4540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.886,
+      "step": 4550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8893,
+      "step": 4560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.893,
+      "step": 4570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8868,
+      "step": 4580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8927,
+      "step": 4590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8892,
+      "step": 4600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8887,
+      "step": 4610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8852,
+      "step": 4620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.88,
+      "step": 4630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8901,
+      "step": 4640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.8889,
+      "step": 4650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.89,
+      "step": 4660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003,
+      "loss": 2.885,
+      "step": 4670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.881,
+      "step": 4680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8837,
+      "step": 4690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8786,
+      "step": 4700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8876,
+      "step": 4710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8933,
+      "step": 4720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8868,
+      "step": 4730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8759,
+      "step": 4740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8758,
+      "step": 4750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8803,
+      "step": 4760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8882,
+      "step": 4770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8776,
+      "step": 4780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8776,
+      "step": 4790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8804,
+      "step": 4800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8807,
+      "step": 4810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8726,
+      "step": 4820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8771,
+      "step": 4830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8895,
+      "step": 4840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8679,
+      "step": 4850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.881,
+      "step": 4860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8825,
+      "step": 4870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8777,
+      "step": 4880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8773,
+      "step": 4890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8623,
+      "step": 4900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8678,
+      "step": 4910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8657,
+      "step": 4920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8657,
+      "step": 4930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003,
+      "loss": 2.8689,
+      "step": 4940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8774,
+      "step": 4950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8722,
+      "step": 4960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8609,
+      "step": 4970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8833,
+      "step": 4980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8802,
+      "step": 4990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8729,
+      "step": 5000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8678,
+      "step": 5010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.869,
+      "step": 5020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8614,
+      "step": 5030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8582,
+      "step": 5040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8704,
+      "step": 5050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.858,
+      "step": 5060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8662,
+      "step": 5070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8682,
+      "step": 5080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 3.1008,
+      "step": 5090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.9056,
+      "step": 5100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8802,
+      "step": 5110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8663,
+      "step": 5120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8533,
+      "step": 5130
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8622,
+      "step": 5140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8638,
+      "step": 5150
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8567,
+      "step": 5160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8594,
+      "step": 5170
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8586,
+      "step": 5180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8688,
+      "step": 5190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8564,
+      "step": 5200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003,
+      "loss": 2.8594,
+      "step": 5210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8611,
+      "step": 5220
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8572,
+      "step": 5230
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8631,
+      "step": 5240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8585,
+      "step": 5250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8537,
+      "step": 5260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8642,
+      "step": 5270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8622,
+      "step": 5280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8568,
+      "step": 5290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8542,
+      "step": 5300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8549,
+      "step": 5310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8538,
+      "step": 5320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8562,
+      "step": 5330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8603,
+      "step": 5340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8556,
+      "step": 5350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8451,
+      "step": 5360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8579,
+      "step": 5370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8534,
+      "step": 5380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8561,
+      "step": 5390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8457,
+      "step": 5400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8577,
+      "step": 5410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8537,
+      "step": 5420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8496,
+      "step": 5430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8581,
+      "step": 5440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8439,
+      "step": 5450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8489,
+      "step": 5460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003,
+      "loss": 2.8395,
+      "step": 5470
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8535,
+      "step": 5480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8528,
+      "step": 5490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8482,
+      "step": 5500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8473,
+      "step": 5510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.845,
+      "step": 5520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8457,
+      "step": 5530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8376,
+      "step": 5540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8533,
+      "step": 5550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8405,
+      "step": 5560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.846,
+      "step": 5570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8447,
+      "step": 5580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8395,
+      "step": 5590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8538,
+      "step": 5600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8505,
+      "step": 5610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8404,
+      "step": 5620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8492,
+      "step": 5630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8452,
+      "step": 5640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8426,
+      "step": 5650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8396,
+      "step": 5660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8305,
+      "step": 5670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8355,
+      "step": 5680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8408,
+      "step": 5690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.839,
+      "step": 5700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8474,
+      "step": 5710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8415,
+      "step": 5720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.836,
+      "step": 5730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003,
+      "loss": 2.8326,
+      "step": 5740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8404,
+      "step": 5750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.824,
+      "step": 5760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8387,
+      "step": 5770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8435,
+      "step": 5780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8457,
+      "step": 5790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8356,
+      "step": 5800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8208,
+      "step": 5810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8338,
+      "step": 5820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8293,
+      "step": 5830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8333,
+      "step": 5840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8353,
+      "step": 5850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8236,
+      "step": 5860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8352,
+      "step": 5870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8353,
+      "step": 5880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8204,
+      "step": 5890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8219,
+      "step": 5900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8279,
+      "step": 5910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8171,
+      "step": 5920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8345,
+      "step": 5930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8275,
+      "step": 5940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8342,
+      "step": 5950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8303,
+      "step": 5960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8295,
+      "step": 5970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8305,
+      "step": 5980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8239,
+      "step": 5990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8249,
+      "step": 6000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003,
+      "loss": 2.8291,
+      "step": 6010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8277,
+      "step": 6020
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8318,
+      "step": 6030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8299,
+      "step": 6040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8201,
+      "step": 6050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8351,
+      "step": 6060
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8284,
+      "step": 6070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8304,
+      "step": 6080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8269,
+      "step": 6090
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8216,
+      "step": 6100
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8187,
+      "step": 6110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8217,
+      "step": 6120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8086,
+      "step": 6130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8248,
+      "step": 6140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8155,
+      "step": 6150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8037,
+      "step": 6160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8274,
+      "step": 6170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8095,
+      "step": 6180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8233,
+      "step": 6190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8217,
+      "step": 6200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8141,
+      "step": 6210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8188,
+      "step": 6220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8161,
+      "step": 6230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.824,
+      "step": 6240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8123,
+      "step": 6250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8264,
+      "step": 6260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8218,
+      "step": 6270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003,
+      "loss": 2.8118,
+      "step": 6280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8036,
+      "step": 6290
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8112,
+      "step": 6300
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.812,
+      "step": 6310
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8108,
+      "step": 6320
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.81,
+      "step": 6330
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8191,
+      "step": 6340
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8114,
+      "step": 6350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8145,
+      "step": 6360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8173,
+      "step": 6370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8038,
+      "step": 6380
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.7997,
+      "step": 6390
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.7941,
+      "step": 6400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8161,
+      "step": 6410
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8106,
+      "step": 6420
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8076,
+      "step": 6430
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8113,
+      "step": 6440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8037,
+      "step": 6450
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8152,
+      "step": 6460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.804,
+      "step": 6470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8134,
+      "step": 6480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8076,
+      "step": 6490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8119,
+      "step": 6500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8112,
+      "step": 6510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8065,
+      "step": 6520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8115,
+      "step": 6530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003,
+      "loss": 2.8101,
+      "step": 6540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8136,
+      "step": 6550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8091,
+      "step": 6560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7996,
+      "step": 6570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8072,
+      "step": 6580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8072,
+      "step": 6590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8145,
+      "step": 6600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8063,
+      "step": 6610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8161,
+      "step": 6620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8095,
+      "step": 6630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7942,
+      "step": 6640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7918,
+      "step": 6650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8107,
+      "step": 6660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7931,
+      "step": 6670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8109,
+      "step": 6680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7906,
+      "step": 6690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7931,
+      "step": 6700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8002,
+      "step": 6710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8054,
+      "step": 6720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7946,
+      "step": 6730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8033,
+      "step": 6740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7996,
+      "step": 6750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.805,
+      "step": 6760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7919,
+      "step": 6770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7963,
+      "step": 6780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8026,
+      "step": 6790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.8073,
+      "step": 6800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003,
+      "loss": 2.7961,
+      "step": 6810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7965,
+      "step": 6820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7938,
+      "step": 6830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7969,
+      "step": 6840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.8025,
+      "step": 6850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7939,
+      "step": 6860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7914,
+      "step": 6870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.8013,
+      "step": 6880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7991,
+      "step": 6890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7966,
+      "step": 6900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7931,
+      "step": 6910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7968,
+      "step": 6920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7969,
+      "step": 6930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7859,
+      "step": 6940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7973,
+      "step": 6950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7942,
+      "step": 6960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7975,
+      "step": 6970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7996,
+      "step": 6980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7912,
+      "step": 6990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7968,
+      "step": 7000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7851,
+      "step": 7010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7897,
+      "step": 7020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7841,
+      "step": 7030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7883,
+      "step": 7040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7855,
+      "step": 7050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7938,
+      "step": 7060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.7873,
+      "step": 7070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003,
+      "loss": 2.79,
+      "step": 7080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7808,
+      "step": 7090
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7894,
+      "step": 7100
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7825,
+      "step": 7110
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7994,
+      "step": 7120
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.793,
+      "step": 7130
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7942,
+      "step": 7140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7805,
+      "step": 7150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7883,
+      "step": 7160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7851,
+      "step": 7170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7769,
+      "step": 7180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.784,
+      "step": 7190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.773,
+      "step": 7200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7872,
+      "step": 7210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7811,
+      "step": 7220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7813,
+      "step": 7230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7847,
+      "step": 7240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.784,
+      "step": 7250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7777,
+      "step": 7260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7822,
+      "step": 7270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7812,
+      "step": 7280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7727,
+      "step": 7290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7895,
+      "step": 7300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7939,
+      "step": 7310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7782,
+      "step": 7320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7879,
+      "step": 7330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003,
+      "loss": 2.7801,
+      "step": 7340
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7723,
+      "step": 7350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7903,
+      "step": 7360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7869,
+      "step": 7370
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7821,
+      "step": 7380
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7781,
+      "step": 7390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7866,
+      "step": 7400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7778,
+      "step": 7410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7818,
+      "step": 7420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7713,
+      "step": 7430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7825,
+      "step": 7440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7723,
+      "step": 7450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7726,
+      "step": 7460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7849,
+      "step": 7470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7837,
+      "step": 7480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7698,
+      "step": 7490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7798,
+      "step": 7500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7805,
+      "step": 7510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7744,
+      "step": 7520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7683,
+      "step": 7530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7784,
+      "step": 7540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7848,
+      "step": 7550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7752,
+      "step": 7560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7708,
+      "step": 7570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7748,
+      "step": 7580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7714,
+      "step": 7590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7772,
+      "step": 7600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003,
+      "loss": 2.7789,
+      "step": 7610
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7686,
+      "step": 7620
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7759,
+      "step": 7630
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7685,
+      "step": 7640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.766,
+      "step": 7650
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7795,
+      "step": 7660
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7725,
+      "step": 7670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7705,
+      "step": 7680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7757,
+      "step": 7690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7661,
+      "step": 7700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7864,
+      "step": 7710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7719,
+      "step": 7720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7751,
+      "step": 7730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7743,
+      "step": 7740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7689,
+      "step": 7750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7657,
+      "step": 7760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7656,
+      "step": 7770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7641,
+      "step": 7780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7733,
+      "step": 7790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7665,
+      "step": 7800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7782,
+      "step": 7810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.765,
+      "step": 7820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7732,
+      "step": 7830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7651,
+      "step": 7840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7839,
+      "step": 7850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7685,
+      "step": 7860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7609,
+      "step": 7870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003,
+      "loss": 2.7648,
+      "step": 7880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7737,
+      "step": 7890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7691,
+      "step": 7900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7613,
+      "step": 7910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7707,
+      "step": 7920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7714,
+      "step": 7930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7582,
+      "step": 7940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7641,
+      "step": 7950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7763,
+      "step": 7960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7603,
+      "step": 7970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7732,
+      "step": 7980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.758,
+      "step": 7990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7647,
+      "step": 8000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7579,
+      "step": 8010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7507,
+      "step": 8020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7667,
+      "step": 8030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7498,
+      "step": 8040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7609,
+      "step": 8050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7606,
+      "step": 8060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.762,
+      "step": 8070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7639,
+      "step": 8080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7599,
+      "step": 8090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7609,
+      "step": 8100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7548,
+      "step": 8110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7666,
+      "step": 8120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7588,
+      "step": 8130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7695,
+      "step": 8140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003,
+      "loss": 2.7698,
+      "step": 8150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.8397,
+      "step": 8160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7883,
+      "step": 8170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7775,
+      "step": 8180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7593,
+      "step": 8190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7718,
+      "step": 8200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7637,
+      "step": 8210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7625,
+      "step": 8220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7599,
+      "step": 8230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7574,
+      "step": 8240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.761,
+      "step": 8250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7613,
+      "step": 8260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7631,
+      "step": 8270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7649,
+      "step": 8280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7565,
+      "step": 8290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7584,
+      "step": 8300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7641,
+      "step": 8310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.757,
+      "step": 8320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7593,
+      "step": 8330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7645,
+      "step": 8340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7491,
+      "step": 8350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7543,
+      "step": 8360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7565,
+      "step": 8370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7518,
+      "step": 8380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7475,
+      "step": 8390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.7543,
+      "step": 8400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003,
+      "loss": 2.746,
+      "step": 8410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.766,
+      "step": 8420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7661,
+      "step": 8430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7568,
+      "step": 8440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7553,
+      "step": 8450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7575,
+      "step": 8460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.761,
+      "step": 8470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7522,
+      "step": 8480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7594,
+      "step": 8490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7506,
+      "step": 8500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7624,
+      "step": 8510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7618,
+      "step": 8520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.757,
+      "step": 8530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7522,
+      "step": 8540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7536,
+      "step": 8550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7461,
+      "step": 8560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7567,
+      "step": 8570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7499,
+      "step": 8580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7619,
+      "step": 8590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7466,
+      "step": 8600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7433,
+      "step": 8610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7454,
+      "step": 8620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7541,
+      "step": 8630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7467,
+      "step": 8640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7551,
+      "step": 8650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7441,
+      "step": 8660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7527,
+      "step": 8670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003,
+      "loss": 2.7472,
+      "step": 8680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7431,
+      "step": 8690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7614,
+      "step": 8700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7503,
+      "step": 8710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7521,
+      "step": 8720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7553,
+      "step": 8730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7561,
+      "step": 8740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7449,
+      "step": 8750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.753,
+      "step": 8760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7454,
+      "step": 8770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7546,
+      "step": 8780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7521,
+      "step": 8790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7468,
+      "step": 8800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7413,
+      "step": 8810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7428,
+      "step": 8820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7475,
+      "step": 8830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7436,
+      "step": 8840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7458,
+      "step": 8850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7499,
+      "step": 8860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7498,
+      "step": 8870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.748,
+      "step": 8880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7409,
+      "step": 8890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7505,
+      "step": 8900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7476,
+      "step": 8910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7412,
+      "step": 8920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7344,
+      "step": 8930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7487,
+      "step": 8940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003,
+      "loss": 2.7424,
+      "step": 8950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7413,
+      "step": 8960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.736,
+      "step": 8970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7438,
+      "step": 8980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7538,
+      "step": 8990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.741,
+      "step": 9000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7432,
+      "step": 9010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7545,
+      "step": 9020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7509,
+      "step": 9030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7354,
+      "step": 9040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7323,
+      "step": 9050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7453,
+      "step": 9060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7364,
+      "step": 9070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7369,
+      "step": 9080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7451,
+      "step": 9090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7372,
+      "step": 9100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7387,
+      "step": 9110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7405,
+      "step": 9120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7341,
+      "step": 9130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7452,
+      "step": 9140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7384,
+      "step": 9150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.745,
+      "step": 9160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.74,
+      "step": 9170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7351,
+      "step": 9180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7309,
+      "step": 9190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7331,
+      "step": 9200
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003,
+      "loss": 2.7488,
+      "step": 9210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7292,
+      "step": 9220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7796,
+      "step": 9230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7926,
+      "step": 9240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.745,
+      "step": 9250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7315,
+      "step": 9260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7382,
+      "step": 9270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7356,
+      "step": 9280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.745,
+      "step": 9290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7338,
+      "step": 9300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7239,
+      "step": 9310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7325,
+      "step": 9320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.742,
+      "step": 9330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7391,
+      "step": 9340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7374,
+      "step": 9350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7237,
+      "step": 9360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7383,
+      "step": 9370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7449,
+      "step": 9380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7307,
+      "step": 9390
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.742,
+      "step": 9400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7382,
+      "step": 9410
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7298,
+      "step": 9420
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7302,
+      "step": 9430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7355,
+      "step": 9440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7471,
+      "step": 9450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7442,
+      "step": 9460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7399,
+      "step": 9470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003,
+      "loss": 2.7333,
+      "step": 9480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7303,
+      "step": 9490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7306,
+      "step": 9500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.73,
+      "step": 9510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7432,
+      "step": 9520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7343,
+      "step": 9530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7308,
+      "step": 9540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7377,
+      "step": 9550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.733,
+      "step": 9560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7345,
+      "step": 9570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.725,
+      "step": 9580
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7268,
+      "step": 9590
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7318,
+      "step": 9600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7309,
+      "step": 9610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7286,
+      "step": 9620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7423,
+      "step": 9630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7301,
+      "step": 9640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7263,
+      "step": 9650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7284,
+      "step": 9660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7269,
+      "step": 9670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7283,
+      "step": 9680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7229,
+      "step": 9690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7309,
+      "step": 9700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7196,
+      "step": 9710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7251,
+      "step": 9720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7244,
+      "step": 9730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7253,
+      "step": 9740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003,
+      "loss": 2.7311,
+      "step": 9750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7247,
+      "step": 9760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7261,
+      "step": 9770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7269,
+      "step": 9780
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7291,
+      "step": 9790
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7327,
+      "step": 9800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7293,
+      "step": 9810
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7229,
+      "step": 9820
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7216,
+      "step": 9830
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7312,
+      "step": 9840
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7272,
+      "step": 9850
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7229,
+      "step": 9860
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7274,
+      "step": 9870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7221,
+      "step": 9880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7244,
+      "step": 9890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7206,
+      "step": 9900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7154,
+      "step": 9910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7287,
+      "step": 9920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7256,
+      "step": 9930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7201,
+      "step": 9940
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7244,
+      "step": 9950
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7257,
+      "step": 9960
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7205,
+      "step": 9970
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7235,
+      "step": 9980
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7312,
+      "step": 9990
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7201,
+      "step": 10000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7222,
+      "step": 10010
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003,
+      "loss": 2.7261,
+      "step": 10020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7258,
+      "step": 10030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.726,
+      "step": 10040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7254,
+      "step": 10050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7256,
+      "step": 10060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7215,
+      "step": 10070
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7256,
+      "step": 10080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7215,
+      "step": 10090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7278,
+      "step": 10100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7306,
+      "step": 10110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7234,
+      "step": 10120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7214,
+      "step": 10130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7206,
+      "step": 10140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7201,
+      "step": 10150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7131,
+      "step": 10160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7265,
+      "step": 10170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7152,
+      "step": 10180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7292,
+      "step": 10190
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7193,
+      "step": 10200
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7216,
+      "step": 10210
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7242,
+      "step": 10220
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7182,
+      "step": 10230
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.721,
+      "step": 10240
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7164,
+      "step": 10250
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7119,
+      "step": 10260
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.715,
+      "step": 10270
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003,
+      "loss": 2.7146,
+      "step": 10280
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7282,
+      "step": 10290
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7091,
+      "step": 10300
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7259,
+      "step": 10310
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7042,
+      "step": 10320
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7172,
+      "step": 10330
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7234,
+      "step": 10340
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7139,
+      "step": 10350
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7118,
+      "step": 10360
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7288,
+      "step": 10370
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7119,
+      "step": 10380
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.72,
+      "step": 10390
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.724,
+      "step": 10400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7131,
+      "step": 10410
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7124,
+      "step": 10420
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7054,
+      "step": 10430
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7072,
+      "step": 10440
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7144,
+      "step": 10450
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7156,
+      "step": 10460
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.702,
+      "step": 10470
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7174,
+      "step": 10480
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7065,
+      "step": 10490
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7131,
+      "step": 10500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7027,
+      "step": 10510
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7099,
+      "step": 10520
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7217,
+      "step": 10530
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7115,
+      "step": 10540
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003,
+      "loss": 2.7139,
+      "step": 10550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7127,
+      "step": 10560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7098,
+      "step": 10570
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7096,
+      "step": 10580
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7128,
+      "step": 10590
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7048,
+      "step": 10600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7162,
+      "step": 10610
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7205,
+      "step": 10620
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7109,
+      "step": 10630
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7054,
+      "step": 10640
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7122,
+      "step": 10650
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7178,
+      "step": 10660
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7131,
+      "step": 10670
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7188,
+      "step": 10680
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7122,
+      "step": 10690
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7013,
+      "step": 10700
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7124,
+      "step": 10710
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7064,
+      "step": 10720
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.6945,
+      "step": 10730
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7051,
+      "step": 10740
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7146,
+      "step": 10750
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7161,
+      "step": 10760
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.714,
+      "step": 10770
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.714,
+      "step": 10780
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7162,
+      "step": 10790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7115,
+      "step": 10800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7052,
+      "step": 10810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003,
+      "loss": 2.7106,
+      "step": 10820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7096,
+      "step": 10830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7004,
+      "step": 10840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7084,
+      "step": 10850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7058,
+      "step": 10860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7091,
+      "step": 10870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.6965,
+      "step": 10880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.714,
+      "step": 10890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7144,
+      "step": 10900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7223,
+      "step": 10910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7139,
+      "step": 10920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7015,
+      "step": 10930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7116,
+      "step": 10940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7081,
+      "step": 10950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7125,
+      "step": 10960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7116,
+      "step": 10970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.713,
+      "step": 10980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7119,
+      "step": 10990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7147,
+      "step": 11000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7061,
+      "step": 11010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.6987,
+      "step": 11020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.6999,
+      "step": 11030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7115,
+      "step": 11040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.706,
+      "step": 11050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7087,
+      "step": 11060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.6983,
+      "step": 11070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7035,
+      "step": 11080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003,
+      "loss": 2.7025,
+      "step": 11090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.6958,
+      "step": 11100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7084,
+      "step": 11110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7061,
+      "step": 11120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7149,
+      "step": 11130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7041,
+      "step": 11140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.6954,
+      "step": 11150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7042,
+      "step": 11160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.704,
+      "step": 11170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7062,
+      "step": 11180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7049,
+      "step": 11190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7028,
+      "step": 11200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7035,
+      "step": 11210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7047,
+      "step": 11220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.6994,
+      "step": 11230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.6958,
+      "step": 11240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7097,
+      "step": 11250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.6991,
+      "step": 11260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7106,
+      "step": 11270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7012,
+      "step": 11280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.6988,
+      "step": 11290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7038,
+      "step": 11300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.6969,
+      "step": 11310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7047,
+      "step": 11320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7009,
+      "step": 11330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.702,
+      "step": 11340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003,
+      "loss": 2.7057,
+      "step": 11350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7023,
+      "step": 11360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.703,
+      "step": 11370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.6988,
+      "step": 11380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7045,
+      "step": 11390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7014,
+      "step": 11400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.6984,
+      "step": 11410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7024,
+      "step": 11420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.709,
+      "step": 11430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7043,
+      "step": 11440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7055,
+      "step": 11450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.6969,
+      "step": 11460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7056,
+      "step": 11470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.6969,
+      "step": 11480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7008,
+      "step": 11490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7094,
+      "step": 11500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7026,
+      "step": 11510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7069,
+      "step": 11520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.6982,
+      "step": 11530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7071,
+      "step": 11540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7032,
+      "step": 11550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.696,
+      "step": 11560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7077,
+      "step": 11570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.6974,
+      "step": 11580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.6977,
+      "step": 11590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.6851,
+      "step": 11600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7003,
+      "step": 11610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003,
+      "loss": 2.7154,
+      "step": 11620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6984,
+      "step": 11630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6969,
+      "step": 11640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6893,
+      "step": 11650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.711,
+      "step": 11660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.7121,
+      "step": 11670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6978,
+      "step": 11680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.76,
+      "step": 11690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.7011,
+      "step": 11700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6949,
+      "step": 11710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6896,
+      "step": 11720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.7004,
+      "step": 11730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6935,
+      "step": 11740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.7053,
+      "step": 11750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6967,
+      "step": 11760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6969,
+      "step": 11770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6959,
+      "step": 11780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.7023,
+      "step": 11790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6968,
+      "step": 11800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6946,
+      "step": 11810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6958,
+      "step": 11820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.7012,
+      "step": 11830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6997,
+      "step": 11840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.7034,
+      "step": 11850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6956,
+      "step": 11860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6919,
+      "step": 11870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.6926,
+      "step": 11880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003,
+      "loss": 2.7018,
+      "step": 11890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6939,
+      "step": 11900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.7016,
+      "step": 11910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6902,
+      "step": 11920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6976,
+      "step": 11930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.7865,
+      "step": 11940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.703,
+      "step": 11950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.693,
+      "step": 11960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6963,
+      "step": 11970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.7003,
+      "step": 11980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6959,
+      "step": 11990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6874,
+      "step": 12000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.7008,
+      "step": 12010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.7026,
+      "step": 12020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6927,
+      "step": 12030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.7007,
+      "step": 12040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6807,
+      "step": 12050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6948,
+      "step": 12060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6959,
+      "step": 12070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6972,
+      "step": 12080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6996,
+      "step": 12090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6923,
+      "step": 12100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6953,
+      "step": 12110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6882,
+      "step": 12120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6902,
+      "step": 12130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.688,
+      "step": 12140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003,
+      "loss": 2.6829,
+      "step": 12150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6905,
+      "step": 12160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.692,
+      "step": 12170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6897,
+      "step": 12180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.69,
+      "step": 12190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6935,
+      "step": 12200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6959,
+      "step": 12210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6861,
+      "step": 12220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6981,
+      "step": 12230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.692,
+      "step": 12240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6918,
+      "step": 12250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6919,
+      "step": 12260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.699,
+      "step": 12270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6877,
+      "step": 12280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6849,
+      "step": 12290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6957,
+      "step": 12300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6843,
+      "step": 12310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6969,
+      "step": 12320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6877,
+      "step": 12330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6931,
+      "step": 12340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6931,
+      "step": 12350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6928,
+      "step": 12360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6918,
+      "step": 12370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6797,
+      "step": 12380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6895,
+      "step": 12390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6919,
+      "step": 12400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.687,
+      "step": 12410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003,
+      "loss": 2.6768,
+      "step": 12420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6945,
+      "step": 12430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6846,
+      "step": 12440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6917,
+      "step": 12450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6819,
+      "step": 12460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6831,
+      "step": 12470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6841,
+      "step": 12480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6885,
+      "step": 12490
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6971,
+      "step": 12500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6865,
+      "step": 12510
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6849,
+      "step": 12520
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6809,
+      "step": 12530
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6859,
+      "step": 12540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.691,
+      "step": 12550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6934,
+      "step": 12560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6896,
+      "step": 12570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.689,
+      "step": 12580
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6856,
+      "step": 12590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6862,
+      "step": 12600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6832,
+      "step": 12610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6888,
+      "step": 12620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6893,
+      "step": 12630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6904,
+      "step": 12640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6811,
+      "step": 12650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6866,
+      "step": 12660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6764,
+      "step": 12670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6921,
+      "step": 12680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003,
+      "loss": 2.6891,
+      "step": 12690
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.694,
+      "step": 12700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6753,
+      "step": 12710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6962,
+      "step": 12720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6781,
+      "step": 12730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6989,
+      "step": 12740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.689,
+      "step": 12750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.69,
+      "step": 12760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6835,
+      "step": 12770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6737,
+      "step": 12780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6831,
+      "step": 12790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6877,
+      "step": 12800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6835,
+      "step": 12810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6887,
+      "step": 12820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6821,
+      "step": 12830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6736,
+      "step": 12840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.684,
+      "step": 12850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6851,
+      "step": 12860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6782,
+      "step": 12870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6812,
+      "step": 12880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6889,
+      "step": 12890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6813,
+      "step": 12900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6755,
+      "step": 12910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6837,
+      "step": 12920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6846,
+      "step": 12930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6814,
+      "step": 12940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6815,
+      "step": 12950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003,
+      "loss": 2.6769,
+      "step": 12960
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6843,
+      "step": 12970
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6811,
+      "step": 12980
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6775,
+      "step": 12990
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6756,
+      "step": 13000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6885,
+      "step": 13010
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6982,
+      "step": 13020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6814,
+      "step": 13030
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.686,
+      "step": 13040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6838,
+      "step": 13050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6862,
+      "step": 13060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6883,
+      "step": 13070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6841,
+      "step": 13080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6782,
+      "step": 13090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6791,
+      "step": 13100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.688,
+      "step": 13110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6828,
+      "step": 13120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.692,
+      "step": 13130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6811,
+      "step": 13140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6799,
+      "step": 13150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6816,
+      "step": 13160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6757,
+      "step": 13170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6756,
+      "step": 13180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6794,
+      "step": 13190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6796,
+      "step": 13200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6894,
+      "step": 13210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003,
+      "loss": 2.6722,
+      "step": 13220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6789,
+      "step": 13230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6749,
+      "step": 13240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6904,
+      "step": 13250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6728,
+      "step": 13260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6792,
+      "step": 13270
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6858,
+      "step": 13280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6664,
+      "step": 13290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.678,
+      "step": 13300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6816,
+      "step": 13310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6839,
+      "step": 13320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.679,
+      "step": 13330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6872,
+      "step": 13340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6761,
+      "step": 13350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6723,
+      "step": 13360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6804,
+      "step": 13370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6813,
+      "step": 13380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6689,
+      "step": 13390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6679,
+      "step": 13400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6674,
+      "step": 13410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6728,
+      "step": 13420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.677,
+      "step": 13430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6795,
+      "step": 13440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6712,
+      "step": 13450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6789,
+      "step": 13460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6794,
+      "step": 13470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6725,
+      "step": 13480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003,
+      "loss": 2.6761,
+      "step": 13490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6709,
+      "step": 13500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6722,
+      "step": 13510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6817,
+      "step": 13520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6765,
+      "step": 13530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6764,
+      "step": 13540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.675,
+      "step": 13550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6731,
+      "step": 13560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6687,
+      "step": 13570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6772,
+      "step": 13580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6715,
+      "step": 13590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6702,
+      "step": 13600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6756,
+      "step": 13610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6772,
+      "step": 13620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6672,
+      "step": 13630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6686,
+      "step": 13640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6667,
+      "step": 13650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6721,
+      "step": 13660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6688,
+      "step": 13670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6703,
+      "step": 13680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6769,
+      "step": 13690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6742,
+      "step": 13700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6728,
+      "step": 13710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6638,
+      "step": 13720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6718,
+      "step": 13730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6741,
+      "step": 13740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6788,
+      "step": 13750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003,
+      "loss": 2.6896,
+      "step": 13760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6656,
+      "step": 13770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6737,
+      "step": 13780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6739,
+      "step": 13790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6758,
+      "step": 13800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6729,
+      "step": 13810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6693,
+      "step": 13820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6659,
+      "step": 13830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6794,
+      "step": 13840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6729,
+      "step": 13850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6769,
+      "step": 13860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6792,
+      "step": 13870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6752,
+      "step": 13880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6712,
+      "step": 13890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6863,
+      "step": 13900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6648,
+      "step": 13910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6799,
+      "step": 13920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6753,
+      "step": 13930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6706,
+      "step": 13940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6698,
+      "step": 13950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6722,
+      "step": 13960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6795,
+      "step": 13970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6819,
+      "step": 13980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6678,
+      "step": 13990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.673,
+      "step": 14000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6701,
+      "step": 14010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003,
+      "loss": 2.6729,
+      "step": 14020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.659,
+      "step": 14030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6732,
+      "step": 14040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6813,
+      "step": 14050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.671,
+      "step": 14060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6609,
+      "step": 14070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6689,
+      "step": 14080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6771,
+      "step": 14090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6664,
+      "step": 14100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6723,
+      "step": 14110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6673,
+      "step": 14120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6767,
+      "step": 14130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6693,
+      "step": 14140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.665,
+      "step": 14150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6662,
+      "step": 14160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6641,
+      "step": 14170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6641,
+      "step": 14180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.666,
+      "step": 14190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6669,
+      "step": 14200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.667,
+      "step": 14210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6756,
+      "step": 14220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6688,
+      "step": 14230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6757,
+      "step": 14240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6753,
+      "step": 14250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6593,
+      "step": 14260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.665,
+      "step": 14270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6758,
+      "step": 14280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003,
+      "loss": 2.6764,
+      "step": 14290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6708,
+      "step": 14300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6685,
+      "step": 14310
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6626,
+      "step": 14320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6668,
+      "step": 14330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6678,
+      "step": 14340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.665,
+      "step": 14350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6557,
+      "step": 14360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6539,
+      "step": 14370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6702,
+      "step": 14380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6676,
+      "step": 14390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6578,
+      "step": 14400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6678,
+      "step": 14410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6632,
+      "step": 14420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6591,
+      "step": 14430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6694,
+      "step": 14440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6621,
+      "step": 14450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6732,
+      "step": 14460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6683,
+      "step": 14470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6568,
+      "step": 14480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.678,
+      "step": 14490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6661,
+      "step": 14500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6674,
+      "step": 14510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.666,
+      "step": 14520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6678,
+      "step": 14530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.659,
+      "step": 14540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6666,
+      "step": 14550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003,
+      "loss": 2.6605,
+      "step": 14560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.663,
+      "step": 14570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6583,
+      "step": 14580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6696,
+      "step": 14590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6645,
+      "step": 14600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6749,
+      "step": 14610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6574,
+      "step": 14620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6621,
+      "step": 14630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6599,
+      "step": 14640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.66,
+      "step": 14650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6643,
+      "step": 14660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6683,
+      "step": 14670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6713,
+      "step": 14680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6644,
+      "step": 14690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6576,
+      "step": 14700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6571,
+      "step": 14710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6671,
+      "step": 14720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6688,
+      "step": 14730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6603,
+      "step": 14740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6564,
+      "step": 14750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6554,
+      "step": 14760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6705,
+      "step": 14770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6642,
+      "step": 14780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6654,
+      "step": 14790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6606,
+      "step": 14800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6595,
+      "step": 14810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6653,
+      "step": 14820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003,
+      "loss": 2.6663,
+      "step": 14830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6666,
+      "step": 14840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.671,
+      "step": 14850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6628,
+      "step": 14860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6664,
+      "step": 14870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.66,
+      "step": 14880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6612,
+      "step": 14890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6712,
+      "step": 14900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6583,
+      "step": 14910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6633,
+      "step": 14920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6639,
+      "step": 14930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6683,
+      "step": 14940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6584,
+      "step": 14950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6589,
+      "step": 14960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6559,
+      "step": 14970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6692,
+      "step": 14980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6692,
+      "step": 14990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6644,
+      "step": 15000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6585,
+      "step": 15010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6589,
+      "step": 15020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6538,
+      "step": 15030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6758,
+      "step": 15040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6548,
+      "step": 15050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.667,
+      "step": 15060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6561,
+      "step": 15070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6601,
+      "step": 15080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003,
+      "loss": 2.6607,
+      "step": 15090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6675,
+      "step": 15100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6609,
+      "step": 15110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6491,
+      "step": 15120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6597,
+      "step": 15130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6658,
+      "step": 15140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6611,
+      "step": 15150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6552,
+      "step": 15160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6579,
+      "step": 15170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6617,
+      "step": 15180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6523,
+      "step": 15190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6582,
+      "step": 15200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6545,
+      "step": 15210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6698,
+      "step": 15220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6594,
+      "step": 15230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6571,
+      "step": 15240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6576,
+      "step": 15250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6539,
+      "step": 15260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6626,
+      "step": 15270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6631,
+      "step": 15280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6481,
+      "step": 15290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6532,
+      "step": 15300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6528,
+      "step": 15310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6519,
+      "step": 15320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6611,
+      "step": 15330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6581,
+      "step": 15340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6597,
+      "step": 15350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003,
+      "loss": 2.6548,
+      "step": 15360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6588,
+      "step": 15370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6637,
+      "step": 15380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6702,
+      "step": 15390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6556,
+      "step": 15400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6507,
+      "step": 15410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6524,
+      "step": 15420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6534,
+      "step": 15430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6543,
+      "step": 15440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6566,
+      "step": 15450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6509,
+      "step": 15460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6507,
+      "step": 15470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6558,
+      "step": 15480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6506,
+      "step": 15490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6567,
+      "step": 15500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6617,
+      "step": 15510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6581,
+      "step": 15520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6624,
+      "step": 15530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6578,
+      "step": 15540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6604,
+      "step": 15550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6482,
+      "step": 15560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.658,
+      "step": 15570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6553,
+      "step": 15580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6482,
+      "step": 15590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6602,
+      "step": 15600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.652,
+      "step": 15610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.657,
+      "step": 15620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003,
+      "loss": 2.6609,
+      "step": 15630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6578,
+      "step": 15640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6554,
+      "step": 15650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6535,
+      "step": 15660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.654,
+      "step": 15670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6573,
+      "step": 15680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.647,
+      "step": 15690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6532,
+      "step": 15700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6605,
+      "step": 15710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6546,
+      "step": 15720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6597,
+      "step": 15730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.654,
+      "step": 15740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6508,
+      "step": 15750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6538,
+      "step": 15760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6613,
+      "step": 15770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.649,
+      "step": 15780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6487,
+      "step": 15790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6623,
+      "step": 15800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6608,
+      "step": 15810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6537,
+      "step": 15820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6549,
+      "step": 15830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6587,
+      "step": 15840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.649,
+      "step": 15850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6539,
+      "step": 15860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6477,
+      "step": 15870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6495,
+      "step": 15880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6589,
+      "step": 15890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003,
+      "loss": 2.6519,
+      "step": 15900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6511,
+      "step": 15910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6487,
+      "step": 15920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6554,
+      "step": 15930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6525,
+      "step": 15940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6579,
+      "step": 15950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6545,
+      "step": 15960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6506,
+      "step": 15970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6434,
+      "step": 15980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6492,
+      "step": 15990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6431,
+      "step": 16000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6427,
+      "step": 16010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6544,
+      "step": 16020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6482,
+      "step": 16030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6489,
+      "step": 16040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6555,
+      "step": 16050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6451,
+      "step": 16060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6558,
+      "step": 16070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6551,
+      "step": 16080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6638,
+      "step": 16090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6552,
+      "step": 16100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6479,
+      "step": 16110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6679,
+      "step": 16120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6483,
+      "step": 16130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6497,
+      "step": 16140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6519,
+      "step": 16150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003,
+      "loss": 2.6403,
+      "step": 16160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6591,
+      "step": 16170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6527,
+      "step": 16180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6453,
+      "step": 16190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6402,
+      "step": 16200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6507,
+      "step": 16210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6486,
+      "step": 16220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.653,
+      "step": 16230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6513,
+      "step": 16240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6521,
+      "step": 16250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6563,
+      "step": 16260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6456,
+      "step": 16270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6529,
+      "step": 16280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6487,
+      "step": 16290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6545,
+      "step": 16300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6434,
+      "step": 16310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6482,
+      "step": 16320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6495,
+      "step": 16330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.649,
+      "step": 16340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6542,
+      "step": 16350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6495,
+      "step": 16360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6496,
+      "step": 16370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6487,
+      "step": 16380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6565,
+      "step": 16390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6467,
+      "step": 16400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6545,
+      "step": 16410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6533,
+      "step": 16420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003,
+      "loss": 2.6391,
+      "step": 16430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6486,
+      "step": 16440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6529,
+      "step": 16450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6495,
+      "step": 16460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.654,
+      "step": 16470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6463,
+      "step": 16480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6459,
+      "step": 16490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6383,
+      "step": 16500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6452,
+      "step": 16510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6608,
+      "step": 16520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6432,
+      "step": 16530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6474,
+      "step": 16540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6458,
+      "step": 16550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6549,
+      "step": 16560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6425,
+      "step": 16570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6441,
+      "step": 16580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6426,
+      "step": 16590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6313,
+      "step": 16600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6459,
+      "step": 16610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6433,
+      "step": 16620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6451,
+      "step": 16630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6457,
+      "step": 16640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6472,
+      "step": 16650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6394,
+      "step": 16660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6417,
+      "step": 16670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6416,
+      "step": 16680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6416,
+      "step": 16690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003,
+      "loss": 2.6462,
+      "step": 16700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6475,
+      "step": 16710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6392,
+      "step": 16720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6367,
+      "step": 16730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6488,
+      "step": 16740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6387,
+      "step": 16750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.644,
+      "step": 16760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6328,
+      "step": 16770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6449,
+      "step": 16780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6465,
+      "step": 16790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6437,
+      "step": 16800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6478,
+      "step": 16810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6503,
+      "step": 16820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6437,
+      "step": 16830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6444,
+      "step": 16840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6467,
+      "step": 16850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6368,
+      "step": 16860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6461,
+      "step": 16870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.65,
+      "step": 16880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6412,
+      "step": 16890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6428,
+      "step": 16900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6443,
+      "step": 16910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6484,
+      "step": 16920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6561,
+      "step": 16930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6578,
+      "step": 16940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.6435,
+      "step": 16950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003,
+      "loss": 2.651,
+      "step": 16960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6461,
+      "step": 16970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6432,
+      "step": 16980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6398,
+      "step": 16990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6505,
+      "step": 17000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6431,
+      "step": 17010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6462,
+      "step": 17020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6365,
+      "step": 17030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6383,
+      "step": 17040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6447,
+      "step": 17050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6386,
+      "step": 17060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.65,
+      "step": 17070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6421,
+      "step": 17080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6441,
+      "step": 17090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6425,
+      "step": 17100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6492,
+      "step": 17110
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6431,
+      "step": 17120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6416,
+      "step": 17130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6493,
+      "step": 17140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.638,
+      "step": 17150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6378,
+      "step": 17160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6389,
+      "step": 17170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6397,
+      "step": 17180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6428,
+      "step": 17190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6499,
+      "step": 17200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6457,
+      "step": 17210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.6441,
+      "step": 17220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003,
+      "loss": 2.638,
+      "step": 17230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6349,
+      "step": 17240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6481,
+      "step": 17250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6362,
+      "step": 17260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6377,
+      "step": 17270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6443,
+      "step": 17280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.628,
+      "step": 17290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6415,
+      "step": 17300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6352,
+      "step": 17310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6332,
+      "step": 17320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6399,
+      "step": 17330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6442,
+      "step": 17340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6382,
+      "step": 17350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6413,
+      "step": 17360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6479,
+      "step": 17370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6458,
+      "step": 17380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6494,
+      "step": 17390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6376,
+      "step": 17400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6352,
+      "step": 17410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6414,
+      "step": 17420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6507,
+      "step": 17430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.654,
+      "step": 17440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6419,
+      "step": 17450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6468,
+      "step": 17460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6429,
+      "step": 17470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6471,
+      "step": 17480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6373,
+      "step": 17490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003,
+      "loss": 2.6373,
+      "step": 17500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6357,
+      "step": 17510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6362,
+      "step": 17520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6401,
+      "step": 17530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6457,
+      "step": 17540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6381,
+      "step": 17550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.648,
+      "step": 17560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6546,
+      "step": 17570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6373,
+      "step": 17580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6358,
+      "step": 17590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6443,
+      "step": 17600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6308,
+      "step": 17610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6331,
+      "step": 17620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6413,
+      "step": 17630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6323,
+      "step": 17640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.634,
+      "step": 17650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6362,
+      "step": 17660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6463,
+      "step": 17670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6288,
+      "step": 17680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6394,
+      "step": 17690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.635,
+      "step": 17700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6301,
+      "step": 17710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6427,
+      "step": 17720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6404,
+      "step": 17730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6366,
+      "step": 17740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6358,
+      "step": 17750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6419,
+      "step": 17760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003,
+      "loss": 2.6522,
+      "step": 17770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6381,
+      "step": 17780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6386,
+      "step": 17790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6357,
+      "step": 17800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6411,
+      "step": 17810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6382,
+      "step": 17820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6414,
+      "step": 17830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6302,
+      "step": 17840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.636,
+      "step": 17850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6347,
+      "step": 17860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6352,
+      "step": 17870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.631,
+      "step": 17880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6449,
+      "step": 17890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6375,
+      "step": 17900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6316,
+      "step": 17910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6338,
+      "step": 17920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6414,
+      "step": 17930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6377,
+      "step": 17940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6312,
+      "step": 17950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6336,
+      "step": 17960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6417,
+      "step": 17970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6343,
+      "step": 17980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6388,
+      "step": 17990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6442,
+      "step": 18000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6393,
+      "step": 18010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6321,
+      "step": 18020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003,
+      "loss": 2.6218,
+      "step": 18030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6355,
+      "step": 18040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6405,
+      "step": 18050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6414,
+      "step": 18060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6325,
+      "step": 18070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6436,
+      "step": 18080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.64,
+      "step": 18090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6299,
+      "step": 18100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6317,
+      "step": 18110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6365,
+      "step": 18120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6383,
+      "step": 18130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.633,
+      "step": 18140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6402,
+      "step": 18150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6331,
+      "step": 18160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6301,
+      "step": 18170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6332,
+      "step": 18180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6389,
+      "step": 18190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6303,
+      "step": 18200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6397,
+      "step": 18210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6345,
+      "step": 18220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6399,
+      "step": 18230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6406,
+      "step": 18240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6355,
+      "step": 18250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6444,
+      "step": 18260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6361,
+      "step": 18270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6353,
+      "step": 18280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.6447,
+      "step": 18290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003,
+      "loss": 2.634,
+      "step": 18300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6403,
+      "step": 18310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6263,
+      "step": 18320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6339,
+      "step": 18330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.631,
+      "step": 18340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6419,
+      "step": 18350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6249,
+      "step": 18360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6404,
+      "step": 18370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6469,
+      "step": 18380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6412,
+      "step": 18390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6307,
+      "step": 18400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6302,
+      "step": 18410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.627,
+      "step": 18420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6344,
+      "step": 18430
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6417,
+      "step": 18440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6271,
+      "step": 18450
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6358,
+      "step": 18460
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6354,
+      "step": 18470
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6375,
+      "step": 18480
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6368,
+      "step": 18490
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6367,
+      "step": 18500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6275,
+      "step": 18510
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6396,
+      "step": 18520
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6373,
+      "step": 18530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6272,
+      "step": 18540
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.628,
+      "step": 18550
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6367,
+      "step": 18560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003,
+      "loss": 2.6346,
+      "step": 18570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6331,
+      "step": 18580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6393,
+      "step": 18590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6325,
+      "step": 18600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6438,
+      "step": 18610
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6388,
+      "step": 18620
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6416,
+      "step": 18630
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6404,
+      "step": 18640
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6424,
+      "step": 18650
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6325,
+      "step": 18660
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6242,
+      "step": 18670
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6319,
+      "step": 18680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6223,
+      "step": 18690
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6434,
+      "step": 18700
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6354,
+      "step": 18710
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6302,
+      "step": 18720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6294,
+      "step": 18730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6377,
+      "step": 18740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6306,
+      "step": 18750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6302,
+      "step": 18760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.632,
+      "step": 18770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6449,
+      "step": 18780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6291,
+      "step": 18790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6294,
+      "step": 18800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6371,
+      "step": 18810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6355,
+      "step": 18820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6314,
+      "step": 18830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003,
+      "loss": 2.6259,
+      "step": 18840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6335,
+      "step": 18850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6406,
+      "step": 18860
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6292,
+      "step": 18870
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6244,
+      "step": 18880
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6333,
+      "step": 18890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6308,
+      "step": 18900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6229,
+      "step": 18910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6294,
+      "step": 18920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6309,
+      "step": 18930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6336,
+      "step": 18940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6284,
+      "step": 18950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6188,
+      "step": 18960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6352,
+      "step": 18970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6255,
+      "step": 18980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6177,
+      "step": 18990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6254,
+      "step": 19000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.633,
+      "step": 19010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6246,
+      "step": 19020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6271,
+      "step": 19030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6391,
+      "step": 19040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6228,
+      "step": 19050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6229,
+      "step": 19060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.627,
+      "step": 19070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6277,
+      "step": 19080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6216,
+      "step": 19090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003,
+      "loss": 2.6226,
+      "step": 19100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6313,
+      "step": 19110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6346,
+      "step": 19120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6353,
+      "step": 19130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6332,
+      "step": 19140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6251,
+      "step": 19150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6255,
+      "step": 19160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6231,
+      "step": 19170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6268,
+      "step": 19180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.619,
+      "step": 19190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6309,
+      "step": 19200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6371,
+      "step": 19210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.62,
+      "step": 19220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6248,
+      "step": 19230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6238,
+      "step": 19240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6288,
+      "step": 19250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6307,
+      "step": 19260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.632,
+      "step": 19270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6443,
+      "step": 19280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6267,
+      "step": 19290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6325,
+      "step": 19300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6297,
+      "step": 19310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6226,
+      "step": 19320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.627,
+      "step": 19330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.624,
+      "step": 19340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6243,
+      "step": 19350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6258,
+      "step": 19360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003,
+      "loss": 2.6183,
+      "step": 19370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6227,
+      "step": 19380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.626,
+      "step": 19390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6357,
+      "step": 19400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6442,
+      "step": 19410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6263,
+      "step": 19420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6295,
+      "step": 19430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6332,
+      "step": 19440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6296,
+      "step": 19450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.627,
+      "step": 19460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6176,
+      "step": 19470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6294,
+      "step": 19480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6226,
+      "step": 19490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6232,
+      "step": 19500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.633,
+      "step": 19510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6212,
+      "step": 19520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6138,
+      "step": 19530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6251,
+      "step": 19540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6192,
+      "step": 19550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6402,
+      "step": 19560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6247,
+      "step": 19570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6292,
+      "step": 19580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6356,
+      "step": 19590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6258,
+      "step": 19600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6268,
+      "step": 19610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6304,
+      "step": 19620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6246,
+      "step": 19630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003,
+      "loss": 2.6261,
+      "step": 19640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6197,
+      "step": 19650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6192,
+      "step": 19660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6304,
+      "step": 19670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6334,
+      "step": 19680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6153,
+      "step": 19690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.622,
+      "step": 19700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6289,
+      "step": 19710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6185,
+      "step": 19720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6222,
+      "step": 19730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6267,
+      "step": 19740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6283,
+      "step": 19750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.615,
+      "step": 19760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.622,
+      "step": 19770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6253,
+      "step": 19780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6233,
+      "step": 19790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6149,
+      "step": 19800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6256,
+      "step": 19810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6249,
+      "step": 19820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6347,
+      "step": 19830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6235,
+      "step": 19840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6214,
+      "step": 19850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6222,
+      "step": 19860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6242,
+      "step": 19870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6268,
+      "step": 19880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.628,
+      "step": 19890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003,
+      "loss": 2.6273,
+      "step": 19900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6309,
+      "step": 19910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6246,
+      "step": 19920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6272,
+      "step": 19930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.628,
+      "step": 19940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.631,
+      "step": 19950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6326,
+      "step": 19960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6194,
+      "step": 19970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6192,
+      "step": 19980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6071,
+      "step": 19990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6235,
+      "step": 20000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6149,
+      "step": 20010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6284,
+      "step": 20020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6207,
+      "step": 20030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6224,
+      "step": 20040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6278,
+      "step": 20050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6189,
+      "step": 20060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6246,
+      "step": 20070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6232,
+      "step": 20080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6109,
+      "step": 20090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6205,
+      "step": 20100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6275,
+      "step": 20110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6204,
+      "step": 20120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6199,
+      "step": 20130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6171,
+      "step": 20140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6189,
+      "step": 20150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6159,
+      "step": 20160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003,
+      "loss": 2.6225,
+      "step": 20170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6249,
+      "step": 20180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6196,
+      "step": 20190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6255,
+      "step": 20200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6306,
+      "step": 20210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6222,
+      "step": 20220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.627,
+      "step": 20230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6199,
+      "step": 20240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6312,
+      "step": 20250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6223,
+      "step": 20260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6127,
+      "step": 20270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6281,
+      "step": 20280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6176,
+      "step": 20290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6298,
+      "step": 20300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6271,
+      "step": 20310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6222,
+      "step": 20320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6285,
+      "step": 20330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6153,
+      "step": 20340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6275,
+      "step": 20350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6274,
+      "step": 20360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6345,
+      "step": 20370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.626,
+      "step": 20380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6235,
+      "step": 20390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6201,
+      "step": 20400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6342,
+      "step": 20410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6235,
+      "step": 20420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6208,
+      "step": 20430
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003,
+      "loss": 2.6182,
+      "step": 20440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6293,
+      "step": 20450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6331,
+      "step": 20460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.623,
+      "step": 20470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6227,
+      "step": 20480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6178,
+      "step": 20490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6271,
+      "step": 20500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6174,
+      "step": 20510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6108,
+      "step": 20520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6207,
+      "step": 20530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6254,
+      "step": 20540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6233,
+      "step": 20550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6142,
+      "step": 20560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.609,
+      "step": 20570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6074,
+      "step": 20580
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6116,
+      "step": 20590
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6288,
+      "step": 20600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6302,
+      "step": 20610
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6085,
+      "step": 20620
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6242,
+      "step": 20630
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6189,
+      "step": 20640
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6211,
+      "step": 20650
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6103,
+      "step": 20660
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6094,
+      "step": 20670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6174,
+      "step": 20680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6114,
+      "step": 20690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.6222,
+      "step": 20700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003,
+      "loss": 2.618,
+      "step": 20710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6129,
+      "step": 20720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6138,
+      "step": 20730
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6238,
+      "step": 20740
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6115,
+      "step": 20750
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6182,
+      "step": 20760
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6237,
+      "step": 20770
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6212,
+      "step": 20780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6239,
+      "step": 20790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6214,
+      "step": 20800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6306,
+      "step": 20810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6153,
+      "step": 20820
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6153,
+      "step": 20830
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6107,
+      "step": 20840
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6239,
+      "step": 20850
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6168,
+      "step": 20860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6199,
+      "step": 20870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.618,
+      "step": 20880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6226,
+      "step": 20890
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6148,
+      "step": 20900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6177,
+      "step": 20910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6129,
+      "step": 20920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6217,
+      "step": 20930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6208,
+      "step": 20940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6229,
+      "step": 20950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6199,
+      "step": 20960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003,
+      "loss": 2.6139,
+      "step": 20970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.606,
+      "step": 20980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6269,
+      "step": 20990
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6266,
+      "step": 21000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6224,
+      "step": 21010
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.621,
+      "step": 21020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.622,
+      "step": 21030
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6165,
+      "step": 21040
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6201,
+      "step": 21050
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6138,
+      "step": 21060
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6282,
+      "step": 21070
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6179,
+      "step": 21080
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6198,
+      "step": 21090
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6052,
+      "step": 21100
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6287,
+      "step": 21110
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6155,
+      "step": 21120
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6208,
+      "step": 21130
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.608,
+      "step": 21140
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.611,
+      "step": 21150
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6136,
+      "step": 21160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6176,
+      "step": 21170
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6258,
+      "step": 21180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6226,
+      "step": 21190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6206,
+      "step": 21200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.614,
+      "step": 21210
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6196,
+      "step": 21220
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6152,
+      "step": 21230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003,
+      "loss": 2.6211,
+      "step": 21240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6249,
+      "step": 21250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.614,
+      "step": 21260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6128,
+      "step": 21270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6285,
+      "step": 21280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6186,
+      "step": 21290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6147,
+      "step": 21300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6106,
+      "step": 21310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6094,
+      "step": 21320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6188,
+      "step": 21330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6152,
+      "step": 21340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6214,
+      "step": 21350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6195,
+      "step": 21360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6175,
+      "step": 21370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6266,
+      "step": 21380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6179,
+      "step": 21390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6111,
+      "step": 21400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6094,
+      "step": 21410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6199,
+      "step": 21420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6198,
+      "step": 21430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6025,
+      "step": 21440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.619,
+      "step": 21450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6059,
+      "step": 21460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6216,
+      "step": 21470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6104,
+      "step": 21480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6116,
+      "step": 21490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6144,
+      "step": 21500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003,
+      "loss": 2.6185,
+      "step": 21510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6132,
+      "step": 21520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.616,
+      "step": 21530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6238,
+      "step": 21540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6208,
+      "step": 21550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6164,
+      "step": 21560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6144,
+      "step": 21570
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6144,
+      "step": 21580
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6168,
+      "step": 21590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6087,
+      "step": 21600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6199,
+      "step": 21610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6179,
+      "step": 21620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6219,
+      "step": 21630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6166,
+      "step": 21640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6144,
+      "step": 21650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6116,
+      "step": 21660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6242,
+      "step": 21670
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6219,
+      "step": 21680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6292,
+      "step": 21690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6069,
+      "step": 21700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6098,
+      "step": 21710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6156,
+      "step": 21720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6092,
+      "step": 21730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6186,
+      "step": 21740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.616,
+      "step": 21750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6114,
+      "step": 21760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003,
+      "loss": 2.6203,
+      "step": 21770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6082,
+      "step": 21780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6077,
+      "step": 21790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6202,
+      "step": 21800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6214,
+      "step": 21810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6076,
+      "step": 21820
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.618,
+      "step": 21830
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6214,
+      "step": 21840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6119,
+      "step": 21850
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.603,
+      "step": 21860
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6137,
+      "step": 21870
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6236,
+      "step": 21880
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6022,
+      "step": 21890
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6124,
+      "step": 21900
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6054,
+      "step": 21910
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6169,
+      "step": 21920
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6067,
+      "step": 21930
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6062,
+      "step": 21940
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6095,
+      "step": 21950
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6094,
+      "step": 21960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6084,
+      "step": 21970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.617,
+      "step": 21980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6163,
+      "step": 21990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6198,
+      "step": 22000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6133,
+      "step": 22010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6098,
+      "step": 22020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6078,
+      "step": 22030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003,
+      "loss": 2.6226,
+      "step": 22040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6105,
+      "step": 22050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6161,
+      "step": 22060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6123,
+      "step": 22070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6246,
+      "step": 22080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6003,
+      "step": 22090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6153,
+      "step": 22100
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6059,
+      "step": 22110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6076,
+      "step": 22120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6061,
+      "step": 22130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6031,
+      "step": 22140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6173,
+      "step": 22150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6106,
+      "step": 22160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6135,
+      "step": 22170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.607,
+      "step": 22180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6127,
+      "step": 22190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.613,
+      "step": 22200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6158,
+      "step": 22210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6043,
+      "step": 22220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6043,
+      "step": 22230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6172,
+      "step": 22240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6079,
+      "step": 22250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6079,
+      "step": 22260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6163,
+      "step": 22270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6139,
+      "step": 22280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6112,
+      "step": 22290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6071,
+      "step": 22300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003,
+      "loss": 2.6112,
+      "step": 22310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6158,
+      "step": 22320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.602,
+      "step": 22330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6102,
+      "step": 22340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6137,
+      "step": 22350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6138,
+      "step": 22360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6183,
+      "step": 22370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.611,
+      "step": 22380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6046,
+      "step": 22390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6099,
+      "step": 22400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6057,
+      "step": 22410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6146,
+      "step": 22420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6097,
+      "step": 22430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6193,
+      "step": 22440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6154,
+      "step": 22450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6076,
+      "step": 22460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6093,
+      "step": 22470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6127,
+      "step": 22480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.608,
+      "step": 22490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6054,
+      "step": 22500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6091,
+      "step": 22510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6085,
+      "step": 22520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6052,
+      "step": 22530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6053,
+      "step": 22540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.603,
+      "step": 22550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6046,
+      "step": 22560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6126,
+      "step": 22570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003,
+      "loss": 2.6136,
+      "step": 22580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6121,
+      "step": 22590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6072,
+      "step": 22600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6016,
+      "step": 22610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6099,
+      "step": 22620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6181,
+      "step": 22630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6024,
+      "step": 22640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6146,
+      "step": 22650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6068,
+      "step": 22660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6103,
+      "step": 22670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6123,
+      "step": 22680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6032,
+      "step": 22690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6061,
+      "step": 22700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6048,
+      "step": 22710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6088,
+      "step": 22720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6005,
+      "step": 22730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6061,
+      "step": 22740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6088,
+      "step": 22750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6166,
+      "step": 22760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.604,
+      "step": 22770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6062,
+      "step": 22780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6093,
+      "step": 22790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6023,
+      "step": 22800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.603,
+      "step": 22810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6005,
+      "step": 22820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6072,
+      "step": 22830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003,
+      "loss": 2.6065,
+      "step": 22840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6081,
+      "step": 22850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6086,
+      "step": 22860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6058,
+      "step": 22870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6057,
+      "step": 22880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6065,
+      "step": 22890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6025,
+      "step": 22900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6072,
+      "step": 22910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6023,
+      "step": 22920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6118,
+      "step": 22930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.5991,
+      "step": 22940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6131,
+      "step": 22950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.5992,
+      "step": 22960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6127,
+      "step": 22970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6107,
+      "step": 22980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6057,
+      "step": 22990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6139,
+      "step": 23000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6048,
+      "step": 23010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.5976,
+      "step": 23020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6083,
+      "step": 23030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6097,
+      "step": 23040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6116,
+      "step": 23050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6121,
+      "step": 23060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.604,
+      "step": 23070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6055,
+      "step": 23080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6039,
+      "step": 23090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6007,
+      "step": 23100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003,
+      "loss": 2.6011,
+      "step": 23110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6079,
+      "step": 23120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6024,
+      "step": 23130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.593,
+      "step": 23140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6081,
+      "step": 23150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6079,
+      "step": 23160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6048,
+      "step": 23170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6003,
+      "step": 23180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6014,
+      "step": 23190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6045,
+      "step": 23200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6059,
+      "step": 23210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.5922,
+      "step": 23220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6029,
+      "step": 23230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6101,
+      "step": 23240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6058,
+      "step": 23250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6023,
+      "step": 23260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6126,
+      "step": 23270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6177,
+      "step": 23280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6093,
+      "step": 23290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.5998,
+      "step": 23300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6116,
+      "step": 23310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.5983,
+      "step": 23320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.5956,
+      "step": 23330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.602,
+      "step": 23340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6097,
+      "step": 23350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.6073,
+      "step": 23360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.607,
+      "step": 23370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003,
+      "loss": 2.5968,
+      "step": 23380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6047,
+      "step": 23390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6005,
+      "step": 23400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6093,
+      "step": 23410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6074,
+      "step": 23420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.612,
+      "step": 23430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6068,
+      "step": 23440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.5984,
+      "step": 23450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6119,
+      "step": 23460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.607,
+      "step": 23470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6119,
+      "step": 23480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6036,
+      "step": 23490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6006,
+      "step": 23500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6072,
+      "step": 23510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6067,
+      "step": 23520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.5885,
+      "step": 23530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6075,
+      "step": 23540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6127,
+      "step": 23550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6082,
+      "step": 23560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6043,
+      "step": 23570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6006,
+      "step": 23580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6065,
+      "step": 23590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6084,
+      "step": 23600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.5989,
+      "step": 23610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.5956,
+      "step": 23620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6092,
+      "step": 23630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.6109,
+      "step": 23640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003,
+      "loss": 2.5963,
+      "step": 23650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6094,
+      "step": 23660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.609,
+      "step": 23670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6152,
+      "step": 23680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6133,
+      "step": 23690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6016,
+      "step": 23700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6075,
+      "step": 23710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6077,
+      "step": 23720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6097,
+      "step": 23730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6106,
+      "step": 23740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6022,
+      "step": 23750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.5854,
+      "step": 23760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6087,
+      "step": 23770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6012,
+      "step": 23780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6042,
+      "step": 23790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6069,
+      "step": 23800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6105,
+      "step": 23810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6052,
+      "step": 23820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6123,
+      "step": 23830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.5984,
+      "step": 23840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.5892,
+      "step": 23850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6047,
+      "step": 23860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6104,
+      "step": 23870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6056,
+      "step": 23880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6072,
+      "step": 23890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.6029,
+      "step": 23900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003,
+      "loss": 2.5981,
+      "step": 23910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6044,
+      "step": 23920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6101,
+      "step": 23930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.5993,
+      "step": 23940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6028,
+      "step": 23950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.611,
+      "step": 23960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6029,
+      "step": 23970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6037,
+      "step": 23980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.598,
+      "step": 23990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6053,
+      "step": 24000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.608,
+      "step": 24010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6142,
+      "step": 24020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.5919,
+      "step": 24030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.5994,
+      "step": 24040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6007,
+      "step": 24050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6015,
+      "step": 24060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6031,
+      "step": 24070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.5986,
+      "step": 24080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.5925,
+      "step": 24090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6015,
+      "step": 24100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6145,
+      "step": 24110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.593,
+      "step": 24120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.5967,
+      "step": 24130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6074,
+      "step": 24140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6067,
+      "step": 24150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.5998,
+      "step": 24160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.6023,
+      "step": 24170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003,
+      "loss": 2.5971,
+      "step": 24180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6013,
+      "step": 24190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6006,
+      "step": 24200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6015,
+      "step": 24210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6059,
+      "step": 24220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6019,
+      "step": 24230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.5939,
+      "step": 24240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.5979,
+      "step": 24250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.5992,
+      "step": 24260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.5976,
+      "step": 24270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6117,
+      "step": 24280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.595,
+      "step": 24290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6146,
+      "step": 24300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6024,
+      "step": 24310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6073,
+      "step": 24320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.5976,
+      "step": 24330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.5983,
+      "step": 24340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6025,
+      "step": 24350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6006,
+      "step": 24360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.5958,
+      "step": 24370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6095,
+      "step": 24380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.5959,
+      "step": 24390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6043,
+      "step": 24400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6045,
+      "step": 24410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6009,
+      "step": 24420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6057,
+      "step": 24430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.6125,
+      "step": 24440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003,
+      "loss": 2.593,
+      "step": 24450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.602,
+      "step": 24460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6007,
+      "step": 24470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.5989,
+      "step": 24480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6038,
+      "step": 24490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6133,
+      "step": 24500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.5948,
+      "step": 24510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6073,
+      "step": 24520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.5969,
+      "step": 24530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6004,
+      "step": 24540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.596,
+      "step": 24550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6074,
+      "step": 24560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6031,
+      "step": 24570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6036,
+      "step": 24580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6072,
+      "step": 24590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.5963,
+      "step": 24600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.5951,
+      "step": 24610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6008,
+      "step": 24620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6175,
+      "step": 24630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6256,
+      "step": 24640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6033,
+      "step": 24650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.5946,
+      "step": 24660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6056,
+      "step": 24670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6127,
+      "step": 24680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.5978,
+      "step": 24690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.6,
+      "step": 24700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003,
+      "loss": 2.5969,
+      "step": 24710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5988,
+      "step": 24720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.608,
+      "step": 24730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6044,
+      "step": 24740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6083,
+      "step": 24750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6023,
+      "step": 24760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6075,
+      "step": 24770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5836,
+      "step": 24780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5929,
+      "step": 24790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6019,
+      "step": 24800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5981,
+      "step": 24810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5993,
+      "step": 24820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5916,
+      "step": 24830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5949,
+      "step": 24840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6059,
+      "step": 24850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.607,
+      "step": 24860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5958,
+      "step": 24870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5889,
+      "step": 24880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5879,
+      "step": 24890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6068,
+      "step": 24900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5906,
+      "step": 24910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6022,
+      "step": 24920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6001,
+      "step": 24930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6067,
+      "step": 24940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6017,
+      "step": 24950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6093,
+      "step": 24960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.5975,
+      "step": 24970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003,
+      "loss": 2.6025,
+      "step": 24980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5873,
+      "step": 24990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5953,
+      "step": 25000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5949,
+      "step": 25010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6029,
+      "step": 25020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6073,
+      "step": 25030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5996,
+      "step": 25040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5947,
+      "step": 25050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5991,
+      "step": 25060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6021,
+      "step": 25070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5958,
+      "step": 25080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6081,
+      "step": 25090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6001,
+      "step": 25100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5946,
+      "step": 25110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6,
+      "step": 25120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6041,
+      "step": 25130
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.586,
+      "step": 25140
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.596,
+      "step": 25150
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5922,
+      "step": 25160
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5888,
+      "step": 25170
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5907,
+      "step": 25180
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6058,
+      "step": 25190
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6025,
+      "step": 25200
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6054,
+      "step": 25210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.6048,
+      "step": 25220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5977,
+      "step": 25230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5945,
+      "step": 25240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003,
+      "loss": 2.5977,
+      "step": 25250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6118,
+      "step": 25260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5845,
+      "step": 25270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5953,
+      "step": 25280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5971,
+      "step": 25290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6036,
+      "step": 25300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5882,
+      "step": 25310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5975,
+      "step": 25320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5955,
+      "step": 25330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6061,
+      "step": 25340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5905,
+      "step": 25350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5916,
+      "step": 25360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5866,
+      "step": 25370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5852,
+      "step": 25380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6024,
+      "step": 25390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5912,
+      "step": 25400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6,
+      "step": 25410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6021,
+      "step": 25420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5919,
+      "step": 25430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6014,
+      "step": 25440
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6028,
+      "step": 25450
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.597,
+      "step": 25460
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6073,
+      "step": 25470
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5945,
+      "step": 25480
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5939,
+      "step": 25490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6042,
+      "step": 25500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.6052,
+      "step": 25510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003,
+      "loss": 2.5966,
+      "step": 25520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.594,
+      "step": 25530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5943,
+      "step": 25540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5955,
+      "step": 25550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5978,
+      "step": 25560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5874,
+      "step": 25570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5986,
+      "step": 25580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5955,
+      "step": 25590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5903,
+      "step": 25600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.605,
+      "step": 25610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5922,
+      "step": 25620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5918,
+      "step": 25630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5901,
+      "step": 25640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5921,
+      "step": 25650
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5967,
+      "step": 25660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.6027,
+      "step": 25670
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5928,
+      "step": 25680
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5915,
+      "step": 25690
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5969,
+      "step": 25700
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5902,
+      "step": 25710
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5969,
+      "step": 25720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.6038,
+      "step": 25730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.6034,
+      "step": 25740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5946,
+      "step": 25750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.6,
+      "step": 25760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.6025,
+      "step": 25770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003,
+      "loss": 2.5872,
+      "step": 25780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.597,
+      "step": 25790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.595,
+      "step": 25800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.6066,
+      "step": 25810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.6003,
+      "step": 25820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5951,
+      "step": 25830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5866,
+      "step": 25840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.6014,
+      "step": 25850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.6002,
+      "step": 25860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.6006,
+      "step": 25870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5914,
+      "step": 25880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5929,
+      "step": 25890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5875,
+      "step": 25900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5966,
+      "step": 25910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5981,
+      "step": 25920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5946,
+      "step": 25930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5941,
+      "step": 25940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5937,
+      "step": 25950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5988,
+      "step": 25960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5978,
+      "step": 25970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5877,
+      "step": 25980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5859,
+      "step": 25990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.594,
+      "step": 26000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5854,
+      "step": 26010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5936,
+      "step": 26020
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5948,
+      "step": 26030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.5842,
+      "step": 26040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 2.592,
+      "step": 26050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5809,
+      "step": 26060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5828,
+      "step": 26070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5901,
+      "step": 26080
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5886,
+      "step": 26090
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5879,
+      "step": 26100
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5933,
+      "step": 26110
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.596,
+      "step": 26120
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.6013,
+      "step": 26130
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5905,
+      "step": 26140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5825,
+      "step": 26150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.6031,
+      "step": 26160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5951,
+      "step": 26170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5904,
+      "step": 26180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5948,
+      "step": 26190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5946,
+      "step": 26200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5945,
+      "step": 26210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5909,
+      "step": 26220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5972,
+      "step": 26230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5898,
+      "step": 26240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.592,
+      "step": 26250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5985,
+      "step": 26260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5867,
+      "step": 26270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5976,
+      "step": 26280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5929,
+      "step": 26290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5954,
+      "step": 26300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5979,
+      "step": 26310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003,
+      "loss": 2.5897,
+      "step": 26320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.6016,
+      "step": 26330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5948,
+      "step": 26340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5981,
+      "step": 26350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.596,
+      "step": 26360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5943,
+      "step": 26370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5906,
+      "step": 26380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5982,
+      "step": 26390
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5962,
+      "step": 26400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5866,
+      "step": 26410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.577,
+      "step": 26420
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5993,
+      "step": 26430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.59,
+      "step": 26440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5905,
+      "step": 26450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.6027,
+      "step": 26460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.596,
+      "step": 26470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5866,
+      "step": 26480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.597,
+      "step": 26490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5958,
+      "step": 26500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5929,
+      "step": 26510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5855,
+      "step": 26520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5924,
+      "step": 26530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.584,
+      "step": 26540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5845,
+      "step": 26550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.6001,
+      "step": 26560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5877,
+      "step": 26570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5796,
+      "step": 26580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003,
+      "loss": 2.5812,
+      "step": 26590
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.592,
+      "step": 26600
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5921,
+      "step": 26610
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5849,
+      "step": 26620
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5996,
+      "step": 26630
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5934,
+      "step": 26640
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5871,
+      "step": 26650
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5926,
+      "step": 26660
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5841,
+      "step": 26670
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.594,
+      "step": 26680
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5913,
+      "step": 26690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5875,
+      "step": 26700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5882,
+      "step": 26710
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003,
+      "loss": 2.5943,
+      "step": 26720
+    }
+  ],
+  "max_steps": 267230,
+  "num_train_epochs": 10,
+  "total_flos": 1.2792009085594003e+20,
+  "trial_name": null,
+  "trial_params": null
+}