Training in progress, epoch 3
Browse files
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 347665996
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:435103d5ed2c372d3266482cd4324ec5b2fcc679dfe98fa9cb84aece1f68854b
|
| 3 |
size 347665996
|
tmp-checkpoint-5781/model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 347665996
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:435103d5ed2c372d3266482cd4324ec5b2fcc679dfe98fa9cb84aece1f68854b
|
| 3 |
size 347665996
|
tmp-checkpoint-5781/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 695604061
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e545c3e35f74920b90837b4ca778af6936c15d018aa3ab842e2593504cc5de8
|
| 3 |
size 695604061
|
tmp-checkpoint-5781/rng_state.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14244
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2a27ea0b368f2b0a9a6d5994fb69481fa5ac9538808338b024c7eed5d7b9c0c0
|
| 3 |
size 14244
|
tmp-checkpoint-5781/trainer_state.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric":
|
| 3 |
"best_model_checkpoint": "swinv2-base-patch4-window12-192-22k-ConcreteClassifier-PVT\\checkpoint-1927",
|
| 4 |
"epoch": 3.0,
|
| 5 |
"eval_steps": 500,
|
|
@@ -11,3529 +11,3529 @@
|
|
| 11 |
{
|
| 12 |
"epoch": 0.01,
|
| 13 |
"learning_rate": 1.7298045320878742e-06,
|
| 14 |
-
"loss": 2.
|
| 15 |
"step": 10
|
| 16 |
},
|
| 17 |
{
|
| 18 |
"epoch": 0.01,
|
| 19 |
"learning_rate": 3.4596090641757485e-06,
|
| 20 |
-
"loss":
|
| 21 |
"step": 20
|
| 22 |
},
|
| 23 |
{
|
| 24 |
"epoch": 0.02,
|
| 25 |
"learning_rate": 5.1894135962636225e-06,
|
| 26 |
-
"loss":
|
| 27 |
"step": 30
|
| 28 |
},
|
| 29 |
{
|
| 30 |
"epoch": 0.02,
|
| 31 |
"learning_rate": 6.919218128351497e-06,
|
| 32 |
-
"loss": 1.
|
| 33 |
"step": 40
|
| 34 |
},
|
| 35 |
{
|
| 36 |
"epoch": 0.03,
|
| 37 |
"learning_rate": 8.649022660439371e-06,
|
| 38 |
-
"loss": 1.
|
| 39 |
"step": 50
|
| 40 |
},
|
| 41 |
{
|
| 42 |
"epoch": 0.03,
|
| 43 |
"learning_rate": 1.0378827192527245e-05,
|
| 44 |
-
"loss": 1.
|
| 45 |
"step": 60
|
| 46 |
},
|
| 47 |
{
|
| 48 |
"epoch": 0.04,
|
| 49 |
"learning_rate": 1.2108631724615119e-05,
|
| 50 |
-
"loss": 1.
|
| 51 |
"step": 70
|
| 52 |
},
|
| 53 |
{
|
| 54 |
"epoch": 0.04,
|
| 55 |
"learning_rate": 1.3838436256702994e-05,
|
| 56 |
-
"loss": 1.
|
| 57 |
"step": 80
|
| 58 |
},
|
| 59 |
{
|
| 60 |
"epoch": 0.05,
|
| 61 |
"learning_rate": 1.5568240788790867e-05,
|
| 62 |
-
"loss": 1.
|
| 63 |
"step": 90
|
| 64 |
},
|
| 65 |
{
|
| 66 |
"epoch": 0.05,
|
| 67 |
"learning_rate": 1.7298045320878743e-05,
|
| 68 |
-
"loss": 1.
|
| 69 |
"step": 100
|
| 70 |
},
|
| 71 |
{
|
| 72 |
"epoch": 0.06,
|
| 73 |
"learning_rate": 1.9027849852966615e-05,
|
| 74 |
-
"loss": 1.
|
| 75 |
"step": 110
|
| 76 |
},
|
| 77 |
{
|
| 78 |
"epoch": 0.06,
|
| 79 |
"learning_rate": 2.075765438505449e-05,
|
| 80 |
-
"loss":
|
| 81 |
"step": 120
|
| 82 |
},
|
| 83 |
{
|
| 84 |
"epoch": 0.07,
|
| 85 |
"learning_rate": 2.2487458917142365e-05,
|
| 86 |
-
"loss": 1.
|
| 87 |
"step": 130
|
| 88 |
},
|
| 89 |
{
|
| 90 |
"epoch": 0.07,
|
| 91 |
"learning_rate": 2.4217263449230237e-05,
|
| 92 |
-
"loss": 1.
|
| 93 |
"step": 140
|
| 94 |
},
|
| 95 |
{
|
| 96 |
"epoch": 0.08,
|
| 97 |
"learning_rate": 2.5947067981318112e-05,
|
| 98 |
-
"loss": 0.
|
| 99 |
"step": 150
|
| 100 |
},
|
| 101 |
{
|
| 102 |
"epoch": 0.08,
|
| 103 |
"learning_rate": 2.7676872513405988e-05,
|
| 104 |
-
"loss": 1.
|
| 105 |
"step": 160
|
| 106 |
},
|
| 107 |
{
|
| 108 |
"epoch": 0.09,
|
| 109 |
"learning_rate": 2.940667704549386e-05,
|
| 110 |
-
"loss":
|
| 111 |
"step": 170
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"epoch": 0.09,
|
| 115 |
"learning_rate": 3.1136481577581735e-05,
|
| 116 |
-
"loss":
|
| 117 |
"step": 180
|
| 118 |
},
|
| 119 |
{
|
| 120 |
"epoch": 0.1,
|
| 121 |
"learning_rate": 3.286628610966961e-05,
|
| 122 |
-
"loss":
|
| 123 |
"step": 190
|
| 124 |
},
|
| 125 |
{
|
| 126 |
"epoch": 0.1,
|
| 127 |
"learning_rate": 3.4596090641757486e-05,
|
| 128 |
-
"loss": 1.
|
| 129 |
"step": 200
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.11,
|
| 133 |
"learning_rate": 3.632589517384536e-05,
|
| 134 |
-
"loss": 0.
|
| 135 |
"step": 210
|
| 136 |
},
|
| 137 |
{
|
| 138 |
"epoch": 0.11,
|
| 139 |
"learning_rate": 3.805569970593323e-05,
|
| 140 |
-
"loss": 0.
|
| 141 |
"step": 220
|
| 142 |
},
|
| 143 |
{
|
| 144 |
"epoch": 0.12,
|
| 145 |
"learning_rate": 3.97855042380211e-05,
|
| 146 |
-
"loss": 0.
|
| 147 |
"step": 230
|
| 148 |
},
|
| 149 |
{
|
| 150 |
"epoch": 0.12,
|
| 151 |
"learning_rate": 4.151530877010898e-05,
|
| 152 |
-
"loss": 1.
|
| 153 |
"step": 240
|
| 154 |
},
|
| 155 |
{
|
| 156 |
"epoch": 0.13,
|
| 157 |
"learning_rate": 4.324511330219685e-05,
|
| 158 |
-
"loss":
|
| 159 |
"step": 250
|
| 160 |
},
|
| 161 |
{
|
| 162 |
"epoch": 0.13,
|
| 163 |
"learning_rate": 4.497491783428473e-05,
|
| 164 |
-
"loss":
|
| 165 |
"step": 260
|
| 166 |
},
|
| 167 |
{
|
| 168 |
"epoch": 0.14,
|
| 169 |
"learning_rate": 4.67047223663726e-05,
|
| 170 |
-
"loss":
|
| 171 |
"step": 270
|
| 172 |
},
|
| 173 |
{
|
| 174 |
"epoch": 0.15,
|
| 175 |
"learning_rate": 4.8434526898460474e-05,
|
| 176 |
-
"loss":
|
| 177 |
"step": 280
|
| 178 |
},
|
| 179 |
{
|
| 180 |
"epoch": 0.15,
|
| 181 |
"learning_rate": 5.0164331430548346e-05,
|
| 182 |
-
"loss": 0.
|
| 183 |
"step": 290
|
| 184 |
},
|
| 185 |
{
|
| 186 |
"epoch": 0.16,
|
| 187 |
"learning_rate": 5.1894135962636225e-05,
|
| 188 |
-
"loss": 0.
|
| 189 |
"step": 300
|
| 190 |
},
|
| 191 |
{
|
| 192 |
"epoch": 0.16,
|
| 193 |
"learning_rate": 5.36239404947241e-05,
|
| 194 |
-
"loss":
|
| 195 |
"step": 310
|
| 196 |
},
|
| 197 |
{
|
| 198 |
"epoch": 0.17,
|
| 199 |
"learning_rate": 5.5353745026811976e-05,
|
| 200 |
-
"loss": 0.
|
| 201 |
"step": 320
|
| 202 |
},
|
| 203 |
{
|
| 204 |
"epoch": 0.17,
|
| 205 |
"learning_rate": 5.708354955889985e-05,
|
| 206 |
-
"loss": 0.
|
| 207 |
"step": 330
|
| 208 |
},
|
| 209 |
{
|
| 210 |
"epoch": 0.18,
|
| 211 |
"learning_rate": 5.881335409098772e-05,
|
| 212 |
-
"loss": 1.
|
| 213 |
"step": 340
|
| 214 |
},
|
| 215 |
{
|
| 216 |
"epoch": 0.18,
|
| 217 |
"learning_rate": 6.054315862307559e-05,
|
| 218 |
-
"loss": 0.
|
| 219 |
"step": 350
|
| 220 |
},
|
| 221 |
{
|
| 222 |
"epoch": 0.19,
|
| 223 |
"learning_rate": 6.227296315516347e-05,
|
| 224 |
-
"loss": 1.
|
| 225 |
"step": 360
|
| 226 |
},
|
| 227 |
{
|
| 228 |
"epoch": 0.19,
|
| 229 |
"learning_rate": 6.400276768725135e-05,
|
| 230 |
-
"loss": 1.
|
| 231 |
"step": 370
|
| 232 |
},
|
| 233 |
{
|
| 234 |
"epoch": 0.2,
|
| 235 |
"learning_rate": 6.573257221933921e-05,
|
| 236 |
-
"loss":
|
| 237 |
"step": 380
|
| 238 |
},
|
| 239 |
{
|
| 240 |
"epoch": 0.2,
|
| 241 |
"learning_rate": 6.746237675142709e-05,
|
| 242 |
-
"loss":
|
| 243 |
"step": 390
|
| 244 |
},
|
| 245 |
{
|
| 246 |
"epoch": 0.21,
|
| 247 |
"learning_rate": 6.919218128351497e-05,
|
| 248 |
-
"loss": 0.
|
| 249 |
"step": 400
|
| 250 |
},
|
| 251 |
{
|
| 252 |
"epoch": 0.21,
|
| 253 |
"learning_rate": 7.092198581560285e-05,
|
| 254 |
-
"loss": 1.
|
| 255 |
"step": 410
|
| 256 |
},
|
| 257 |
{
|
| 258 |
"epoch": 0.22,
|
| 259 |
"learning_rate": 7.265179034769071e-05,
|
| 260 |
-
"loss":
|
| 261 |
"step": 420
|
| 262 |
},
|
| 263 |
{
|
| 264 |
"epoch": 0.22,
|
| 265 |
"learning_rate": 7.438159487977858e-05,
|
| 266 |
-
"loss":
|
| 267 |
"step": 430
|
| 268 |
},
|
| 269 |
{
|
| 270 |
"epoch": 0.23,
|
| 271 |
"learning_rate": 7.611139941186646e-05,
|
| 272 |
-
"loss": 0.
|
| 273 |
"step": 440
|
| 274 |
},
|
| 275 |
{
|
| 276 |
"epoch": 0.23,
|
| 277 |
"learning_rate": 7.784120394395434e-05,
|
| 278 |
-
"loss": 0.
|
| 279 |
"step": 450
|
| 280 |
},
|
| 281 |
{
|
| 282 |
"epoch": 0.24,
|
| 283 |
"learning_rate": 7.95710084760422e-05,
|
| 284 |
-
"loss": 1.
|
| 285 |
"step": 460
|
| 286 |
},
|
| 287 |
{
|
| 288 |
"epoch": 0.24,
|
| 289 |
"learning_rate": 8.130081300813008e-05,
|
| 290 |
-
"loss": 0.
|
| 291 |
"step": 470
|
| 292 |
},
|
| 293 |
{
|
| 294 |
"epoch": 0.25,
|
| 295 |
"learning_rate": 8.303061754021796e-05,
|
| 296 |
-
"loss":
|
| 297 |
"step": 480
|
| 298 |
},
|
| 299 |
{
|
| 300 |
"epoch": 0.25,
|
| 301 |
"learning_rate": 8.476042207230583e-05,
|
| 302 |
-
"loss": 0.
|
| 303 |
"step": 490
|
| 304 |
},
|
| 305 |
{
|
| 306 |
"epoch": 0.26,
|
| 307 |
"learning_rate": 8.64902266043937e-05,
|
| 308 |
-
"loss": 1.
|
| 309 |
"step": 500
|
| 310 |
},
|
| 311 |
{
|
| 312 |
"epoch": 0.26,
|
| 313 |
"learning_rate": 8.822003113648158e-05,
|
| 314 |
-
"loss":
|
| 315 |
"step": 510
|
| 316 |
},
|
| 317 |
{
|
| 318 |
"epoch": 0.27,
|
| 319 |
"learning_rate": 8.994983566856946e-05,
|
| 320 |
-
"loss":
|
| 321 |
"step": 520
|
| 322 |
},
|
| 323 |
{
|
| 324 |
"epoch": 0.28,
|
| 325 |
"learning_rate": 9.167964020065733e-05,
|
| 326 |
-
"loss": 0.
|
| 327 |
"step": 530
|
| 328 |
},
|
| 329 |
{
|
| 330 |
"epoch": 0.28,
|
| 331 |
"learning_rate": 9.34094447327452e-05,
|
| 332 |
-
"loss":
|
| 333 |
"step": 540
|
| 334 |
},
|
| 335 |
{
|
| 336 |
"epoch": 0.29,
|
| 337 |
"learning_rate": 9.513924926483307e-05,
|
| 338 |
-
"loss": 1.
|
| 339 |
"step": 550
|
| 340 |
},
|
| 341 |
{
|
| 342 |
"epoch": 0.29,
|
| 343 |
"learning_rate": 9.686905379692095e-05,
|
| 344 |
-
"loss":
|
| 345 |
"step": 560
|
| 346 |
},
|
| 347 |
{
|
| 348 |
"epoch": 0.3,
|
| 349 |
"learning_rate": 9.859885832900881e-05,
|
| 350 |
-
"loss":
|
| 351 |
"step": 570
|
| 352 |
},
|
| 353 |
{
|
| 354 |
"epoch": 0.3,
|
| 355 |
"learning_rate": 0.00010032866286109669,
|
| 356 |
-
"loss": 1.
|
| 357 |
"step": 580
|
| 358 |
},
|
| 359 |
{
|
| 360 |
"epoch": 0.31,
|
| 361 |
"learning_rate": 0.00010205846739318457,
|
| 362 |
-
"loss": 0.
|
| 363 |
"step": 590
|
| 364 |
},
|
| 365 |
{
|
| 366 |
"epoch": 0.31,
|
| 367 |
"learning_rate": 0.00010378827192527245,
|
| 368 |
-
"loss": 0.
|
| 369 |
"step": 600
|
| 370 |
},
|
| 371 |
{
|
| 372 |
"epoch": 0.32,
|
| 373 |
"learning_rate": 0.00010551807645736032,
|
| 374 |
-
"loss":
|
| 375 |
"step": 610
|
| 376 |
},
|
| 377 |
{
|
| 378 |
"epoch": 0.32,
|
| 379 |
"learning_rate": 0.0001072478809894482,
|
| 380 |
-
"loss":
|
| 381 |
"step": 620
|
| 382 |
},
|
| 383 |
{
|
| 384 |
"epoch": 0.33,
|
| 385 |
"learning_rate": 0.00010897768552153607,
|
| 386 |
-
"loss": 0.
|
| 387 |
"step": 630
|
| 388 |
},
|
| 389 |
{
|
| 390 |
"epoch": 0.33,
|
| 391 |
"learning_rate": 0.00011070749005362395,
|
| 392 |
-
"loss": 1.
|
| 393 |
"step": 640
|
| 394 |
},
|
| 395 |
{
|
| 396 |
"epoch": 0.34,
|
| 397 |
"learning_rate": 0.00011243729458571182,
|
| 398 |
-
"loss":
|
| 399 |
"step": 650
|
| 400 |
},
|
| 401 |
{
|
| 402 |
"epoch": 0.34,
|
| 403 |
"learning_rate": 0.0001141670991177997,
|
| 404 |
-
"loss":
|
| 405 |
"step": 660
|
| 406 |
},
|
| 407 |
{
|
| 408 |
"epoch": 0.35,
|
| 409 |
"learning_rate": 0.00011589690364988756,
|
| 410 |
-
"loss":
|
| 411 |
"step": 670
|
| 412 |
},
|
| 413 |
{
|
| 414 |
"epoch": 0.35,
|
| 415 |
"learning_rate": 0.00011762670818197544,
|
| 416 |
-
"loss":
|
| 417 |
"step": 680
|
| 418 |
},
|
| 419 |
{
|
| 420 |
"epoch": 0.36,
|
| 421 |
"learning_rate": 0.0001193565127140633,
|
| 422 |
-
"loss":
|
| 423 |
"step": 690
|
| 424 |
},
|
| 425 |
{
|
| 426 |
"epoch": 0.36,
|
| 427 |
"learning_rate": 0.00012108631724615118,
|
| 428 |
-
"loss":
|
| 429 |
"step": 700
|
| 430 |
},
|
| 431 |
{
|
| 432 |
"epoch": 0.37,
|
| 433 |
"learning_rate": 0.00012281612177823906,
|
| 434 |
-
"loss": 1.
|
| 435 |
"step": 710
|
| 436 |
},
|
| 437 |
{
|
| 438 |
"epoch": 0.37,
|
| 439 |
"learning_rate": 0.00012454592631032694,
|
| 440 |
-
"loss": 1.
|
| 441 |
"step": 720
|
| 442 |
},
|
| 443 |
{
|
| 444 |
"epoch": 0.38,
|
| 445 |
"learning_rate": 0.00012627573084241482,
|
| 446 |
-
"loss":
|
| 447 |
"step": 730
|
| 448 |
},
|
| 449 |
{
|
| 450 |
"epoch": 0.38,
|
| 451 |
"learning_rate": 0.0001280055353745027,
|
| 452 |
-
"loss":
|
| 453 |
"step": 740
|
| 454 |
},
|
| 455 |
{
|
| 456 |
"epoch": 0.39,
|
| 457 |
"learning_rate": 0.00012973533990659055,
|
| 458 |
-
"loss":
|
| 459 |
"step": 750
|
| 460 |
},
|
| 461 |
{
|
| 462 |
"epoch": 0.39,
|
| 463 |
"learning_rate": 0.00013146514443867843,
|
| 464 |
-
"loss": 1.
|
| 465 |
"step": 760
|
| 466 |
},
|
| 467 |
{
|
| 468 |
"epoch": 0.4,
|
| 469 |
"learning_rate": 0.0001331949489707663,
|
| 470 |
-
"loss":
|
| 471 |
"step": 770
|
| 472 |
},
|
| 473 |
{
|
| 474 |
"epoch": 0.4,
|
| 475 |
"learning_rate": 0.00013492475350285418,
|
| 476 |
-
"loss": 1.
|
| 477 |
"step": 780
|
| 478 |
},
|
| 479 |
{
|
| 480 |
"epoch": 0.41,
|
| 481 |
"learning_rate": 0.00013665455803494206,
|
| 482 |
-
"loss": 1.
|
| 483 |
"step": 790
|
| 484 |
},
|
| 485 |
{
|
| 486 |
"epoch": 0.42,
|
| 487 |
"learning_rate": 0.00013838436256702994,
|
| 488 |
-
"loss": 0.
|
| 489 |
"step": 800
|
| 490 |
},
|
| 491 |
{
|
| 492 |
"epoch": 0.42,
|
| 493 |
"learning_rate": 0.00014011416709911782,
|
| 494 |
-
"loss":
|
| 495 |
"step": 810
|
| 496 |
},
|
| 497 |
{
|
| 498 |
"epoch": 0.43,
|
| 499 |
"learning_rate": 0.0001418439716312057,
|
| 500 |
-
"loss": 0.
|
| 501 |
"step": 820
|
| 502 |
},
|
| 503 |
{
|
| 504 |
"epoch": 0.43,
|
| 505 |
"learning_rate": 0.00014357377616329355,
|
| 506 |
-
"loss": 1.
|
| 507 |
"step": 830
|
| 508 |
},
|
| 509 |
{
|
| 510 |
"epoch": 0.44,
|
| 511 |
"learning_rate": 0.00014530358069538143,
|
| 512 |
-
"loss":
|
| 513 |
"step": 840
|
| 514 |
},
|
| 515 |
{
|
| 516 |
"epoch": 0.44,
|
| 517 |
"learning_rate": 0.0001470333852274693,
|
| 518 |
-
"loss": 0.
|
| 519 |
"step": 850
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.45,
|
| 523 |
"learning_rate": 0.00014876318975955716,
|
| 524 |
-
"loss":
|
| 525 |
"step": 860
|
| 526 |
},
|
| 527 |
{
|
| 528 |
"epoch": 0.45,
|
| 529 |
"learning_rate": 0.00015049299429164504,
|
| 530 |
-
"loss":
|
| 531 |
"step": 870
|
| 532 |
},
|
| 533 |
{
|
| 534 |
"epoch": 0.46,
|
| 535 |
"learning_rate": 0.00015222279882373292,
|
| 536 |
-
"loss":
|
| 537 |
"step": 880
|
| 538 |
},
|
| 539 |
{
|
| 540 |
"epoch": 0.46,
|
| 541 |
"learning_rate": 0.0001539526033558208,
|
| 542 |
-
"loss": 1.
|
| 543 |
"step": 890
|
| 544 |
},
|
| 545 |
{
|
| 546 |
"epoch": 0.47,
|
| 547 |
"learning_rate": 0.00015568240788790867,
|
| 548 |
-
"loss": 0.
|
| 549 |
"step": 900
|
| 550 |
},
|
| 551 |
{
|
| 552 |
"epoch": 0.47,
|
| 553 |
"learning_rate": 0.00015741221241999653,
|
| 554 |
-
"loss": 1.
|
| 555 |
"step": 910
|
| 556 |
},
|
| 557 |
{
|
| 558 |
"epoch": 0.48,
|
| 559 |
"learning_rate": 0.0001591420169520844,
|
| 560 |
-
"loss":
|
| 561 |
"step": 920
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.48,
|
| 565 |
"learning_rate": 0.00016087182148417228,
|
| 566 |
-
"loss": 0.
|
| 567 |
"step": 930
|
| 568 |
},
|
| 569 |
{
|
| 570 |
"epoch": 0.49,
|
| 571 |
"learning_rate": 0.00016260162601626016,
|
| 572 |
-
"loss": 1.
|
| 573 |
"step": 940
|
| 574 |
},
|
| 575 |
{
|
| 576 |
"epoch": 0.49,
|
| 577 |
"learning_rate": 0.00016433143054834804,
|
| 578 |
-
"loss":
|
| 579 |
"step": 950
|
| 580 |
},
|
| 581 |
{
|
| 582 |
"epoch": 0.5,
|
| 583 |
"learning_rate": 0.00016606123508043592,
|
| 584 |
-
"loss": 1.
|
| 585 |
"step": 960
|
| 586 |
},
|
| 587 |
{
|
| 588 |
"epoch": 0.5,
|
| 589 |
"learning_rate": 0.0001677910396125238,
|
| 590 |
-
"loss":
|
| 591 |
"step": 970
|
| 592 |
},
|
| 593 |
{
|
| 594 |
"epoch": 0.51,
|
| 595 |
"learning_rate": 0.00016952084414461165,
|
| 596 |
-
"loss": 1.
|
| 597 |
"step": 980
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.51,
|
| 601 |
"learning_rate": 0.00017125064867669953,
|
| 602 |
-
"loss":
|
| 603 |
"step": 990
|
| 604 |
},
|
| 605 |
{
|
| 606 |
"epoch": 0.52,
|
| 607 |
"learning_rate": 0.0001729804532087874,
|
| 608 |
-
"loss": 1.
|
| 609 |
"step": 1000
|
| 610 |
},
|
| 611 |
{
|
| 612 |
"epoch": 0.52,
|
| 613 |
"learning_rate": 0.00017471025774087529,
|
| 614 |
-
"loss": 1.
|
| 615 |
"step": 1010
|
| 616 |
},
|
| 617 |
{
|
| 618 |
"epoch": 0.53,
|
| 619 |
"learning_rate": 0.00017644006227296316,
|
| 620 |
-
"loss": 0.
|
| 621 |
"step": 1020
|
| 622 |
},
|
| 623 |
{
|
| 624 |
"epoch": 0.53,
|
| 625 |
"learning_rate": 0.00017816986680505104,
|
| 626 |
-
"loss":
|
| 627 |
"step": 1030
|
| 628 |
},
|
| 629 |
{
|
| 630 |
"epoch": 0.54,
|
| 631 |
"learning_rate": 0.00017989967133713892,
|
| 632 |
-
"loss":
|
| 633 |
"step": 1040
|
| 634 |
},
|
| 635 |
{
|
| 636 |
"epoch": 0.54,
|
| 637 |
"learning_rate": 0.0001816294758692268,
|
| 638 |
-
"loss": 1.
|
| 639 |
"step": 1050
|
| 640 |
},
|
| 641 |
{
|
| 642 |
"epoch": 0.55,
|
| 643 |
"learning_rate": 0.00018335928040131465,
|
| 644 |
-
"loss": 1.
|
| 645 |
"step": 1060
|
| 646 |
},
|
| 647 |
{
|
| 648 |
"epoch": 0.56,
|
| 649 |
"learning_rate": 0.00018508908493340253,
|
| 650 |
-
"loss":
|
| 651 |
"step": 1070
|
| 652 |
},
|
| 653 |
{
|
| 654 |
"epoch": 0.56,
|
| 655 |
"learning_rate": 0.0001868188894654904,
|
| 656 |
-
"loss": 1.
|
| 657 |
"step": 1080
|
| 658 |
},
|
| 659 |
{
|
| 660 |
"epoch": 0.57,
|
| 661 |
"learning_rate": 0.0001885486939975783,
|
| 662 |
-
"loss": 1.
|
| 663 |
"step": 1090
|
| 664 |
},
|
| 665 |
{
|
| 666 |
"epoch": 0.57,
|
| 667 |
"learning_rate": 0.00019027849852966614,
|
| 668 |
-
"loss": 1.
|
| 669 |
"step": 1100
|
| 670 |
},
|
| 671 |
{
|
| 672 |
"epoch": 0.58,
|
| 673 |
"learning_rate": 0.00019200830306175402,
|
| 674 |
-
"loss":
|
| 675 |
"step": 1110
|
| 676 |
},
|
| 677 |
{
|
| 678 |
"epoch": 0.58,
|
| 679 |
"learning_rate": 0.0001937381075938419,
|
| 680 |
-
"loss": 1.
|
| 681 |
"step": 1120
|
| 682 |
},
|
| 683 |
{
|
| 684 |
"epoch": 0.59,
|
| 685 |
"learning_rate": 0.00019546791212592978,
|
| 686 |
-
"loss": 1.
|
| 687 |
"step": 1130
|
| 688 |
},
|
| 689 |
{
|
| 690 |
"epoch": 0.59,
|
| 691 |
"learning_rate": 0.00019719771665801763,
|
| 692 |
-
"loss": 1.
|
| 693 |
"step": 1140
|
| 694 |
},
|
| 695 |
{
|
| 696 |
"epoch": 0.6,
|
| 697 |
"learning_rate": 0.0001989275211901055,
|
| 698 |
-
"loss": 1.
|
| 699 |
"step": 1150
|
| 700 |
},
|
| 701 |
{
|
| 702 |
"epoch": 0.6,
|
| 703 |
"learning_rate": 0.00020065732572219339,
|
| 704 |
-
"loss": 1.
|
| 705 |
"step": 1160
|
| 706 |
},
|
| 707 |
{
|
| 708 |
"epoch": 0.61,
|
| 709 |
"learning_rate": 0.00020238713025428126,
|
| 710 |
-
"loss": 1.
|
| 711 |
"step": 1170
|
| 712 |
},
|
| 713 |
{
|
| 714 |
"epoch": 0.61,
|
| 715 |
"learning_rate": 0.00020411693478636914,
|
| 716 |
-
"loss":
|
| 717 |
"step": 1180
|
| 718 |
},
|
| 719 |
{
|
| 720 |
"epoch": 0.62,
|
| 721 |
"learning_rate": 0.00020584673931845702,
|
| 722 |
-
"loss":
|
| 723 |
"step": 1190
|
| 724 |
},
|
| 725 |
{
|
| 726 |
"epoch": 0.62,
|
| 727 |
"learning_rate": 0.0002075765438505449,
|
| 728 |
-
"loss": 1.
|
| 729 |
"step": 1200
|
| 730 |
},
|
| 731 |
{
|
| 732 |
"epoch": 0.63,
|
| 733 |
"learning_rate": 0.00020930634838263278,
|
| 734 |
-
"loss": 1.
|
| 735 |
"step": 1210
|
| 736 |
},
|
| 737 |
{
|
| 738 |
"epoch": 0.63,
|
| 739 |
"learning_rate": 0.00021103615291472063,
|
| 740 |
-
"loss":
|
| 741 |
"step": 1220
|
| 742 |
},
|
| 743 |
{
|
| 744 |
"epoch": 0.64,
|
| 745 |
"learning_rate": 0.0002127659574468085,
|
| 746 |
-
"loss":
|
| 747 |
"step": 1230
|
| 748 |
},
|
| 749 |
{
|
| 750 |
"epoch": 0.64,
|
| 751 |
"learning_rate": 0.0002144957619788964,
|
| 752 |
-
"loss": 1.
|
| 753 |
"step": 1240
|
| 754 |
},
|
| 755 |
{
|
| 756 |
"epoch": 0.65,
|
| 757 |
"learning_rate": 0.00021622556651098427,
|
| 758 |
-
"loss": 1.
|
| 759 |
"step": 1250
|
| 760 |
},
|
| 761 |
{
|
| 762 |
"epoch": 0.65,
|
| 763 |
"learning_rate": 0.00021795537104307214,
|
| 764 |
-
"loss": 1.
|
| 765 |
"step": 1260
|
| 766 |
},
|
| 767 |
{
|
| 768 |
"epoch": 0.66,
|
| 769 |
"learning_rate": 0.00021968517557516002,
|
| 770 |
-
"loss":
|
| 771 |
"step": 1270
|
| 772 |
},
|
| 773 |
{
|
| 774 |
"epoch": 0.66,
|
| 775 |
"learning_rate": 0.0002214149801072479,
|
| 776 |
-
"loss": 1.
|
| 777 |
"step": 1280
|
| 778 |
},
|
| 779 |
{
|
| 780 |
"epoch": 0.67,
|
| 781 |
"learning_rate": 0.00022314478463933578,
|
| 782 |
-
"loss":
|
| 783 |
"step": 1290
|
| 784 |
},
|
| 785 |
{
|
| 786 |
"epoch": 0.67,
|
| 787 |
"learning_rate": 0.00022487458917142363,
|
| 788 |
-
"loss": 1.
|
| 789 |
"step": 1300
|
| 790 |
},
|
| 791 |
{
|
| 792 |
"epoch": 0.68,
|
| 793 |
"learning_rate": 0.0002266043937035115,
|
| 794 |
-
"loss": 1.
|
| 795 |
"step": 1310
|
| 796 |
},
|
| 797 |
{
|
| 798 |
"epoch": 0.69,
|
| 799 |
"learning_rate": 0.0002283341982355994,
|
| 800 |
-
"loss":
|
| 801 |
"step": 1320
|
| 802 |
},
|
| 803 |
{
|
| 804 |
"epoch": 0.69,
|
| 805 |
"learning_rate": 0.00023006400276768727,
|
| 806 |
-
"loss": 1.
|
| 807 |
"step": 1330
|
| 808 |
},
|
| 809 |
{
|
| 810 |
"epoch": 0.7,
|
| 811 |
"learning_rate": 0.00023179380729977512,
|
| 812 |
-
"loss": 1.
|
| 813 |
"step": 1340
|
| 814 |
},
|
| 815 |
{
|
| 816 |
"epoch": 0.7,
|
| 817 |
"learning_rate": 0.000233523611831863,
|
| 818 |
-
"loss": 1.
|
| 819 |
"step": 1350
|
| 820 |
},
|
| 821 |
{
|
| 822 |
"epoch": 0.71,
|
| 823 |
"learning_rate": 0.00023525341636395088,
|
| 824 |
-
"loss":
|
| 825 |
"step": 1360
|
| 826 |
},
|
| 827 |
{
|
| 828 |
"epoch": 0.71,
|
| 829 |
"learning_rate": 0.00023698322089603876,
|
| 830 |
-
"loss": 1.
|
| 831 |
"step": 1370
|
| 832 |
},
|
| 833 |
{
|
| 834 |
"epoch": 0.72,
|
| 835 |
"learning_rate": 0.0002387130254281266,
|
| 836 |
-
"loss": 1.
|
| 837 |
"step": 1380
|
| 838 |
},
|
| 839 |
{
|
| 840 |
"epoch": 0.72,
|
| 841 |
"learning_rate": 0.00024044282996021449,
|
| 842 |
-
"loss": 1.
|
| 843 |
"step": 1390
|
| 844 |
},
|
| 845 |
{
|
| 846 |
"epoch": 0.73,
|
| 847 |
"learning_rate": 0.00024217263449230237,
|
| 848 |
-
"loss": 1.
|
| 849 |
"step": 1400
|
| 850 |
},
|
| 851 |
{
|
| 852 |
"epoch": 0.73,
|
| 853 |
"learning_rate": 0.00024390243902439024,
|
| 854 |
-
"loss": 0.
|
| 855 |
"step": 1410
|
| 856 |
},
|
| 857 |
{
|
| 858 |
"epoch": 0.74,
|
| 859 |
"learning_rate": 0.0002456322435564781,
|
| 860 |
-
"loss": 1.
|
| 861 |
"step": 1420
|
| 862 |
},
|
| 863 |
{
|
| 864 |
"epoch": 0.74,
|
| 865 |
"learning_rate": 0.000247362048088566,
|
| 866 |
-
"loss": 1.
|
| 867 |
"step": 1430
|
| 868 |
},
|
| 869 |
{
|
| 870 |
"epoch": 0.75,
|
| 871 |
"learning_rate": 0.0002490918526206539,
|
| 872 |
-
"loss":
|
| 873 |
"step": 1440
|
| 874 |
},
|
| 875 |
{
|
| 876 |
"epoch": 0.75,
|
| 877 |
"learning_rate": 0.00025082165715274176,
|
| 878 |
-
"loss": 1.
|
| 879 |
"step": 1450
|
| 880 |
},
|
| 881 |
{
|
| 882 |
"epoch": 0.76,
|
| 883 |
"learning_rate": 0.00025255146168482964,
|
| 884 |
-
"loss": 1.
|
| 885 |
"step": 1460
|
| 886 |
},
|
| 887 |
{
|
| 888 |
"epoch": 0.76,
|
| 889 |
"learning_rate": 0.0002542812662169175,
|
| 890 |
-
"loss": 1.
|
| 891 |
"step": 1470
|
| 892 |
},
|
| 893 |
{
|
| 894 |
"epoch": 0.77,
|
| 895 |
"learning_rate": 0.0002560110707490054,
|
| 896 |
-
"loss": 1.
|
| 897 |
"step": 1480
|
| 898 |
},
|
| 899 |
{
|
| 900 |
"epoch": 0.77,
|
| 901 |
"learning_rate": 0.0002577408752810932,
|
| 902 |
-
"loss": 1.
|
| 903 |
"step": 1490
|
| 904 |
},
|
| 905 |
{
|
| 906 |
"epoch": 0.78,
|
| 907 |
"learning_rate": 0.0002594706798131811,
|
| 908 |
-
"loss": 1.
|
| 909 |
"step": 1500
|
| 910 |
},
|
| 911 |
{
|
| 912 |
"epoch": 0.78,
|
| 913 |
"learning_rate": 0.000261200484345269,
|
| 914 |
-
"loss": 1.
|
| 915 |
"step": 1510
|
| 916 |
},
|
| 917 |
{
|
| 918 |
"epoch": 0.79,
|
| 919 |
"learning_rate": 0.00026293028887735685,
|
| 920 |
-
"loss": 0.
|
| 921 |
"step": 1520
|
| 922 |
},
|
| 923 |
{
|
| 924 |
"epoch": 0.79,
|
| 925 |
"learning_rate": 0.00026466009340944473,
|
| 926 |
-
"loss": 1.
|
| 927 |
"step": 1530
|
| 928 |
},
|
| 929 |
{
|
| 930 |
"epoch": 0.8,
|
| 931 |
"learning_rate": 0.0002663898979415326,
|
| 932 |
-
"loss":
|
| 933 |
"step": 1540
|
| 934 |
},
|
| 935 |
{
|
| 936 |
"epoch": 0.8,
|
| 937 |
"learning_rate": 0.0002681197024736205,
|
| 938 |
-
"loss": 1.
|
| 939 |
"step": 1550
|
| 940 |
},
|
| 941 |
{
|
| 942 |
"epoch": 0.81,
|
| 943 |
"learning_rate": 0.00026984950700570837,
|
| 944 |
-
"loss": 1.
|
| 945 |
"step": 1560
|
| 946 |
},
|
| 947 |
{
|
| 948 |
"epoch": 0.81,
|
| 949 |
"learning_rate": 0.00027157931153779625,
|
| 950 |
-
"loss": 1.
|
| 951 |
"step": 1570
|
| 952 |
},
|
| 953 |
{
|
| 954 |
"epoch": 0.82,
|
| 955 |
"learning_rate": 0.0002733091160698841,
|
| 956 |
-
"loss":
|
| 957 |
"step": 1580
|
| 958 |
},
|
| 959 |
{
|
| 960 |
"epoch": 0.83,
|
| 961 |
"learning_rate": 0.000275038920601972,
|
| 962 |
-
"loss": 1.
|
| 963 |
"step": 1590
|
| 964 |
},
|
| 965 |
{
|
| 966 |
"epoch": 0.83,
|
| 967 |
"learning_rate": 0.0002767687251340599,
|
| 968 |
-
"loss": 1.
|
| 969 |
"step": 1600
|
| 970 |
},
|
| 971 |
{
|
| 972 |
"epoch": 0.84,
|
| 973 |
"learning_rate": 0.00027849852966614776,
|
| 974 |
-
"loss": 1.
|
| 975 |
"step": 1610
|
| 976 |
},
|
| 977 |
{
|
| 978 |
"epoch": 0.84,
|
| 979 |
"learning_rate": 0.00028022833419823564,
|
| 980 |
-
"loss": 1.
|
| 981 |
"step": 1620
|
| 982 |
},
|
| 983 |
{
|
| 984 |
"epoch": 0.85,
|
| 985 |
"learning_rate": 0.0002819581387303235,
|
| 986 |
-
"loss": 1.
|
| 987 |
"step": 1630
|
| 988 |
},
|
| 989 |
{
|
| 990 |
"epoch": 0.85,
|
| 991 |
"learning_rate": 0.0002836879432624114,
|
| 992 |
-
"loss": 1.
|
| 993 |
"step": 1640
|
| 994 |
},
|
| 995 |
{
|
| 996 |
"epoch": 0.86,
|
| 997 |
"learning_rate": 0.0002854177477944992,
|
| 998 |
-
"loss": 1.
|
| 999 |
"step": 1650
|
| 1000 |
},
|
| 1001 |
{
|
| 1002 |
"epoch": 0.86,
|
| 1003 |
"learning_rate": 0.0002871475523265871,
|
| 1004 |
-
"loss":
|
| 1005 |
"step": 1660
|
| 1006 |
},
|
| 1007 |
{
|
| 1008 |
"epoch": 0.87,
|
| 1009 |
"learning_rate": 0.000288877356858675,
|
| 1010 |
-
"loss": 1.
|
| 1011 |
"step": 1670
|
| 1012 |
},
|
| 1013 |
{
|
| 1014 |
"epoch": 0.87,
|
| 1015 |
"learning_rate": 0.00029060716139076286,
|
| 1016 |
-
"loss": 1.
|
| 1017 |
"step": 1680
|
| 1018 |
},
|
| 1019 |
{
|
| 1020 |
"epoch": 0.88,
|
| 1021 |
"learning_rate": 0.00029233696592285074,
|
| 1022 |
-
"loss": 1.
|
| 1023 |
"step": 1690
|
| 1024 |
},
|
| 1025 |
{
|
| 1026 |
"epoch": 0.88,
|
| 1027 |
"learning_rate": 0.0002940667704549386,
|
| 1028 |
-
"loss":
|
| 1029 |
"step": 1700
|
| 1030 |
},
|
| 1031 |
{
|
| 1032 |
"epoch": 0.89,
|
| 1033 |
"learning_rate": 0.00029579657498702644,
|
| 1034 |
-
"loss": 1.
|
| 1035 |
"step": 1710
|
| 1036 |
},
|
| 1037 |
{
|
| 1038 |
"epoch": 0.89,
|
| 1039 |
"learning_rate": 0.0002975263795191143,
|
| 1040 |
-
"loss": 1.
|
| 1041 |
"step": 1720
|
| 1042 |
},
|
| 1043 |
{
|
| 1044 |
"epoch": 0.9,
|
| 1045 |
"learning_rate": 0.0002992561840512022,
|
| 1046 |
-
"loss": 1.
|
| 1047 |
"step": 1730
|
| 1048 |
},
|
| 1049 |
{
|
| 1050 |
"epoch": 0.9,
|
| 1051 |
"learning_rate": 0.0003009859885832901,
|
| 1052 |
-
"loss": 1.
|
| 1053 |
"step": 1740
|
| 1054 |
},
|
| 1055 |
{
|
| 1056 |
"epoch": 0.91,
|
| 1057 |
"learning_rate": 0.00030271579311537796,
|
| 1058 |
-
"loss": 1.
|
| 1059 |
"step": 1750
|
| 1060 |
},
|
| 1061 |
{
|
| 1062 |
"epoch": 0.91,
|
| 1063 |
"learning_rate": 0.00030444559764746583,
|
| 1064 |
-
"loss": 1.
|
| 1065 |
"step": 1760
|
| 1066 |
},
|
| 1067 |
{
|
| 1068 |
"epoch": 0.92,
|
| 1069 |
"learning_rate": 0.0003061754021795537,
|
| 1070 |
-
"loss": 1.
|
| 1071 |
"step": 1770
|
| 1072 |
},
|
| 1073 |
{
|
| 1074 |
"epoch": 0.92,
|
| 1075 |
"learning_rate": 0.0003079052067116416,
|
| 1076 |
-
"loss": 1.
|
| 1077 |
"step": 1780
|
| 1078 |
},
|
| 1079 |
{
|
| 1080 |
"epoch": 0.93,
|
| 1081 |
"learning_rate": 0.00030963501124372947,
|
| 1082 |
-
"loss": 1.
|
| 1083 |
"step": 1790
|
| 1084 |
},
|
| 1085 |
{
|
| 1086 |
"epoch": 0.93,
|
| 1087 |
"learning_rate": 0.00031136481577581735,
|
| 1088 |
-
"loss": 1.
|
| 1089 |
"step": 1800
|
| 1090 |
},
|
| 1091 |
{
|
| 1092 |
"epoch": 0.94,
|
| 1093 |
"learning_rate": 0.0003130946203079052,
|
| 1094 |
-
"loss": 1.
|
| 1095 |
"step": 1810
|
| 1096 |
},
|
| 1097 |
{
|
| 1098 |
"epoch": 0.94,
|
| 1099 |
"learning_rate": 0.00031482442483999305,
|
| 1100 |
-
"loss": 1.
|
| 1101 |
"step": 1820
|
| 1102 |
},
|
| 1103 |
{
|
| 1104 |
"epoch": 0.95,
|
| 1105 |
"learning_rate": 0.00031655422937208093,
|
| 1106 |
-
"loss": 1.
|
| 1107 |
"step": 1830
|
| 1108 |
},
|
| 1109 |
{
|
| 1110 |
"epoch": 0.95,
|
| 1111 |
"learning_rate": 0.0003182840339041688,
|
| 1112 |
-
"loss": 1.
|
| 1113 |
"step": 1840
|
| 1114 |
},
|
| 1115 |
{
|
| 1116 |
"epoch": 0.96,
|
| 1117 |
"learning_rate": 0.0003200138384362567,
|
| 1118 |
-
"loss":
|
| 1119 |
"step": 1850
|
| 1120 |
},
|
| 1121 |
{
|
| 1122 |
"epoch": 0.97,
|
| 1123 |
"learning_rate": 0.00032174364296834457,
|
| 1124 |
-
"loss":
|
| 1125 |
"step": 1860
|
| 1126 |
},
|
| 1127 |
{
|
| 1128 |
"epoch": 0.97,
|
| 1129 |
"learning_rate": 0.00032347344750043245,
|
| 1130 |
-
"loss": 1.
|
| 1131 |
"step": 1870
|
| 1132 |
},
|
| 1133 |
{
|
| 1134 |
"epoch": 0.98,
|
| 1135 |
"learning_rate": 0.0003252032520325203,
|
| 1136 |
-
"loss": 1.
|
| 1137 |
"step": 1880
|
| 1138 |
},
|
| 1139 |
{
|
| 1140 |
"epoch": 0.98,
|
| 1141 |
"learning_rate": 0.0003269330565646082,
|
| 1142 |
-
"loss": 1.
|
| 1143 |
"step": 1890
|
| 1144 |
},
|
| 1145 |
{
|
| 1146 |
"epoch": 0.99,
|
| 1147 |
"learning_rate": 0.0003286628610966961,
|
| 1148 |
-
"loss": 1.
|
| 1149 |
"step": 1900
|
| 1150 |
},
|
| 1151 |
{
|
| 1152 |
"epoch": 0.99,
|
| 1153 |
"learning_rate": 0.00033039266562878396,
|
| 1154 |
-
"loss": 1.
|
| 1155 |
"step": 1910
|
| 1156 |
},
|
| 1157 |
{
|
| 1158 |
"epoch": 1.0,
|
| 1159 |
"learning_rate": 0.00033212247016087184,
|
| 1160 |
-
"loss": 1.
|
| 1161 |
"step": 1920
|
| 1162 |
},
|
| 1163 |
{
|
| 1164 |
"epoch": 1.0,
|
| 1165 |
"eval_accuracy": {
|
| 1166 |
-
"accuracy": 0.
|
| 1167 |
},
|
| 1168 |
"eval_f1": {
|
| 1169 |
-
"f1": 0.
|
| 1170 |
},
|
| 1171 |
-
"eval_loss":
|
| 1172 |
"eval_precision": {
|
| 1173 |
-
"precision": 0.
|
| 1174 |
},
|
| 1175 |
"eval_recall": {
|
| 1176 |
-
"recall": 0.
|
| 1177 |
},
|
| 1178 |
-
"eval_runtime":
|
| 1179 |
-
"eval_samples_per_second":
|
| 1180 |
-
"eval_steps_per_second":
|
| 1181 |
"step": 1927
|
| 1182 |
},
|
| 1183 |
{
|
| 1184 |
"epoch": 1.0,
|
| 1185 |
"learning_rate": 0.0003338522746929597,
|
| 1186 |
-
"loss": 1.
|
| 1187 |
"step": 1930
|
| 1188 |
},
|
| 1189 |
{
|
| 1190 |
"epoch": 1.01,
|
| 1191 |
"learning_rate": 0.0003355820792250476,
|
| 1192 |
-
"loss": 1.
|
| 1193 |
"step": 1940
|
| 1194 |
},
|
| 1195 |
{
|
| 1196 |
"epoch": 1.01,
|
| 1197 |
"learning_rate": 0.0003373118837571355,
|
| 1198 |
-
"loss": 1.
|
| 1199 |
"step": 1950
|
| 1200 |
},
|
| 1201 |
{
|
| 1202 |
"epoch": 1.02,
|
| 1203 |
"learning_rate": 0.0003390416882892233,
|
| 1204 |
-
"loss": 1.
|
| 1205 |
"step": 1960
|
| 1206 |
},
|
| 1207 |
{
|
| 1208 |
"epoch": 1.02,
|
| 1209 |
"learning_rate": 0.0003407714928213112,
|
| 1210 |
-
"loss": 1.
|
| 1211 |
"step": 1970
|
| 1212 |
},
|
| 1213 |
{
|
| 1214 |
"epoch": 1.03,
|
| 1215 |
"learning_rate": 0.00034250129735339906,
|
| 1216 |
-
"loss": 1.
|
| 1217 |
"step": 1980
|
| 1218 |
},
|
| 1219 |
{
|
| 1220 |
"epoch": 1.03,
|
| 1221 |
"learning_rate": 0.00034423110188548694,
|
| 1222 |
-
"loss": 1.
|
| 1223 |
"step": 1990
|
| 1224 |
},
|
| 1225 |
{
|
| 1226 |
"epoch": 1.04,
|
| 1227 |
"learning_rate": 0.0003459609064175748,
|
| 1228 |
-
"loss": 1.
|
| 1229 |
"step": 2000
|
| 1230 |
},
|
| 1231 |
{
|
| 1232 |
"epoch": 1.04,
|
| 1233 |
"learning_rate": 0.0003476907109496627,
|
| 1234 |
-
"loss": 1.
|
| 1235 |
"step": 2010
|
| 1236 |
},
|
| 1237 |
{
|
| 1238 |
"epoch": 1.05,
|
| 1239 |
"learning_rate": 0.00034942051548175057,
|
| 1240 |
-
"loss":
|
| 1241 |
"step": 2020
|
| 1242 |
},
|
| 1243 |
{
|
| 1244 |
"epoch": 1.05,
|
| 1245 |
"learning_rate": 0.00035115032001383845,
|
| 1246 |
-
"loss": 1.
|
| 1247 |
"step": 2030
|
| 1248 |
},
|
| 1249 |
{
|
| 1250 |
"epoch": 1.06,
|
| 1251 |
"learning_rate": 0.00035288012454592633,
|
| 1252 |
-
"loss": 1.
|
| 1253 |
"step": 2040
|
| 1254 |
},
|
| 1255 |
{
|
| 1256 |
"epoch": 1.06,
|
| 1257 |
"learning_rate": 0.0003546099290780142,
|
| 1258 |
-
"loss": 1.
|
| 1259 |
"step": 2050
|
| 1260 |
},
|
| 1261 |
{
|
| 1262 |
"epoch": 1.07,
|
| 1263 |
"learning_rate": 0.0003563397336101021,
|
| 1264 |
-
"loss": 1.
|
| 1265 |
"step": 2060
|
| 1266 |
},
|
| 1267 |
{
|
| 1268 |
"epoch": 1.07,
|
| 1269 |
"learning_rate": 0.00035806953814218997,
|
| 1270 |
-
"loss":
|
| 1271 |
"step": 2070
|
| 1272 |
},
|
| 1273 |
{
|
| 1274 |
"epoch": 1.08,
|
| 1275 |
"learning_rate": 0.00035979934267427784,
|
| 1276 |
-
"loss": 1.
|
| 1277 |
"step": 2080
|
| 1278 |
},
|
| 1279 |
{
|
| 1280 |
"epoch": 1.08,
|
| 1281 |
"learning_rate": 0.0003615291472063657,
|
| 1282 |
-
"loss": 1.
|
| 1283 |
"step": 2090
|
| 1284 |
},
|
| 1285 |
{
|
| 1286 |
"epoch": 1.09,
|
| 1287 |
"learning_rate": 0.0003632589517384536,
|
| 1288 |
-
"loss": 1.
|
| 1289 |
"step": 2100
|
| 1290 |
},
|
| 1291 |
{
|
| 1292 |
"epoch": 1.09,
|
| 1293 |
"learning_rate": 0.0003649887562705415,
|
| 1294 |
-
"loss": 1.
|
| 1295 |
"step": 2110
|
| 1296 |
},
|
| 1297 |
{
|
| 1298 |
"epoch": 1.1,
|
| 1299 |
"learning_rate": 0.0003667185608026293,
|
| 1300 |
-
"loss": 1.
|
| 1301 |
"step": 2120
|
| 1302 |
},
|
| 1303 |
{
|
| 1304 |
"epoch": 1.11,
|
| 1305 |
"learning_rate": 0.0003684483653347172,
|
| 1306 |
-
"loss":
|
| 1307 |
"step": 2130
|
| 1308 |
},
|
| 1309 |
{
|
| 1310 |
"epoch": 1.11,
|
| 1311 |
"learning_rate": 0.00037017816986680506,
|
| 1312 |
-
"loss": 1.
|
| 1313 |
"step": 2140
|
| 1314 |
},
|
| 1315 |
{
|
| 1316 |
"epoch": 1.12,
|
| 1317 |
"learning_rate": 0.00037190797439889294,
|
| 1318 |
-
"loss": 1.
|
| 1319 |
"step": 2150
|
| 1320 |
},
|
| 1321 |
{
|
| 1322 |
"epoch": 1.12,
|
| 1323 |
"learning_rate": 0.0003736377789309808,
|
| 1324 |
-
"loss": 1.
|
| 1325 |
"step": 2160
|
| 1326 |
},
|
| 1327 |
{
|
| 1328 |
"epoch": 1.13,
|
| 1329 |
"learning_rate": 0.0003753675834630687,
|
| 1330 |
-
"loss": 1.
|
| 1331 |
"step": 2170
|
| 1332 |
},
|
| 1333 |
{
|
| 1334 |
"epoch": 1.13,
|
| 1335 |
"learning_rate": 0.0003770973879951566,
|
| 1336 |
-
"loss": 1.
|
| 1337 |
"step": 2180
|
| 1338 |
},
|
| 1339 |
{
|
| 1340 |
"epoch": 1.14,
|
| 1341 |
"learning_rate": 0.0003788271925272444,
|
| 1342 |
-
"loss": 1.
|
| 1343 |
"step": 2190
|
| 1344 |
},
|
| 1345 |
{
|
| 1346 |
"epoch": 1.14,
|
| 1347 |
"learning_rate": 0.0003805569970593323,
|
| 1348 |
-
"loss": 1.
|
| 1349 |
"step": 2200
|
| 1350 |
},
|
| 1351 |
{
|
| 1352 |
"epoch": 1.15,
|
| 1353 |
"learning_rate": 0.00038228680159142016,
|
| 1354 |
-
"loss": 1.
|
| 1355 |
"step": 2210
|
| 1356 |
},
|
| 1357 |
{
|
| 1358 |
"epoch": 1.15,
|
| 1359 |
"learning_rate": 0.00038401660612350804,
|
| 1360 |
-
"loss": 1.
|
| 1361 |
"step": 2220
|
| 1362 |
},
|
| 1363 |
{
|
| 1364 |
"epoch": 1.16,
|
| 1365 |
"learning_rate": 0.0003857464106555959,
|
| 1366 |
-
"loss": 1.
|
| 1367 |
"step": 2230
|
| 1368 |
},
|
| 1369 |
{
|
| 1370 |
"epoch": 1.16,
|
| 1371 |
"learning_rate": 0.0003874762151876838,
|
| 1372 |
-
"loss": 1.
|
| 1373 |
"step": 2240
|
| 1374 |
},
|
| 1375 |
{
|
| 1376 |
"epoch": 1.17,
|
| 1377 |
"learning_rate": 0.0003892060197197717,
|
| 1378 |
-
"loss": 1.
|
| 1379 |
"step": 2250
|
| 1380 |
},
|
| 1381 |
{
|
| 1382 |
"epoch": 1.17,
|
| 1383 |
"learning_rate": 0.00039093582425185955,
|
| 1384 |
-
"loss": 1.
|
| 1385 |
"step": 2260
|
| 1386 |
},
|
| 1387 |
{
|
| 1388 |
"epoch": 1.18,
|
| 1389 |
"learning_rate": 0.00039266562878394743,
|
| 1390 |
-
"loss": 1.
|
| 1391 |
"step": 2270
|
| 1392 |
},
|
| 1393 |
{
|
| 1394 |
"epoch": 1.18,
|
| 1395 |
"learning_rate": 0.00039439543331603526,
|
| 1396 |
-
"loss": 1.
|
| 1397 |
"step": 2280
|
| 1398 |
},
|
| 1399 |
{
|
| 1400 |
"epoch": 1.19,
|
| 1401 |
"learning_rate": 0.00039612523784812313,
|
| 1402 |
-
"loss": 1.
|
| 1403 |
"step": 2290
|
| 1404 |
},
|
| 1405 |
{
|
| 1406 |
"epoch": 1.19,
|
| 1407 |
"learning_rate": 0.000397855042380211,
|
| 1408 |
-
"loss": 1.
|
| 1409 |
"step": 2300
|
| 1410 |
},
|
| 1411 |
{
|
| 1412 |
"epoch": 1.2,
|
| 1413 |
"learning_rate": 0.0003995848469122989,
|
| 1414 |
-
"loss": 1.
|
| 1415 |
"step": 2310
|
| 1416 |
},
|
| 1417 |
{
|
| 1418 |
"epoch": 1.2,
|
| 1419 |
"learning_rate": 0.00040131465144438677,
|
| 1420 |
-
"loss": 1.
|
| 1421 |
"step": 2320
|
| 1422 |
},
|
| 1423 |
{
|
| 1424 |
"epoch": 1.21,
|
| 1425 |
"learning_rate": 0.00040304445597647465,
|
| 1426 |
-
"loss": 1.
|
| 1427 |
"step": 2330
|
| 1428 |
},
|
| 1429 |
{
|
| 1430 |
"epoch": 1.21,
|
| 1431 |
"learning_rate": 0.00040477426050856253,
|
| 1432 |
-
"loss": 1.
|
| 1433 |
"step": 2340
|
| 1434 |
},
|
| 1435 |
{
|
| 1436 |
"epoch": 1.22,
|
| 1437 |
"learning_rate": 0.0004065040650406504,
|
| 1438 |
-
"loss": 1.
|
| 1439 |
"step": 2350
|
| 1440 |
},
|
| 1441 |
{
|
| 1442 |
"epoch": 1.22,
|
| 1443 |
"learning_rate": 0.0004082338695727383,
|
| 1444 |
-
"loss": 1.
|
| 1445 |
"step": 2360
|
| 1446 |
},
|
| 1447 |
{
|
| 1448 |
"epoch": 1.23,
|
| 1449 |
"learning_rate": 0.00040996367410482616,
|
| 1450 |
-
"loss": 1.
|
| 1451 |
"step": 2370
|
| 1452 |
},
|
| 1453 |
{
|
| 1454 |
"epoch": 1.24,
|
| 1455 |
"learning_rate": 0.00041169347863691404,
|
| 1456 |
-
"loss": 1.
|
| 1457 |
"step": 2380
|
| 1458 |
},
|
| 1459 |
{
|
| 1460 |
"epoch": 1.24,
|
| 1461 |
"learning_rate": 0.0004134232831690019,
|
| 1462 |
-
"loss": 1.
|
| 1463 |
"step": 2390
|
| 1464 |
},
|
| 1465 |
{
|
| 1466 |
"epoch": 1.25,
|
| 1467 |
"learning_rate": 0.0004151530877010898,
|
| 1468 |
-
"loss": 1.
|
| 1469 |
"step": 2400
|
| 1470 |
},
|
| 1471 |
{
|
| 1472 |
"epoch": 1.25,
|
| 1473 |
"learning_rate": 0.0004168828922331777,
|
| 1474 |
-
"loss":
|
| 1475 |
"step": 2410
|
| 1476 |
},
|
| 1477 |
{
|
| 1478 |
"epoch": 1.26,
|
| 1479 |
"learning_rate": 0.00041861269676526556,
|
| 1480 |
-
"loss": 1.
|
| 1481 |
"step": 2420
|
| 1482 |
},
|
| 1483 |
{
|
| 1484 |
"epoch": 1.26,
|
| 1485 |
"learning_rate": 0.0004203425012973534,
|
| 1486 |
-
"loss":
|
| 1487 |
"step": 2430
|
| 1488 |
},
|
| 1489 |
{
|
| 1490 |
"epoch": 1.27,
|
| 1491 |
"learning_rate": 0.00042207230582944126,
|
| 1492 |
-
"loss": 1.
|
| 1493 |
"step": 2440
|
| 1494 |
},
|
| 1495 |
{
|
| 1496 |
"epoch": 1.27,
|
| 1497 |
"learning_rate": 0.00042380211036152914,
|
| 1498 |
-
"loss": 1.
|
| 1499 |
"step": 2450
|
| 1500 |
},
|
| 1501 |
{
|
| 1502 |
"epoch": 1.28,
|
| 1503 |
"learning_rate": 0.000425531914893617,
|
| 1504 |
-
"loss": 1.
|
| 1505 |
"step": 2460
|
| 1506 |
},
|
| 1507 |
{
|
| 1508 |
"epoch": 1.28,
|
| 1509 |
"learning_rate": 0.0004272617194257049,
|
| 1510 |
-
"loss": 1.
|
| 1511 |
"step": 2470
|
| 1512 |
},
|
| 1513 |
{
|
| 1514 |
"epoch": 1.29,
|
| 1515 |
"learning_rate": 0.0004289915239577928,
|
| 1516 |
-
"loss": 1.
|
| 1517 |
"step": 2480
|
| 1518 |
},
|
| 1519 |
{
|
| 1520 |
"epoch": 1.29,
|
| 1521 |
"learning_rate": 0.00043072132848988065,
|
| 1522 |
-
"loss": 1.
|
| 1523 |
"step": 2490
|
| 1524 |
},
|
| 1525 |
{
|
| 1526 |
"epoch": 1.3,
|
| 1527 |
"learning_rate": 0.00043245113302196853,
|
| 1528 |
-
"loss":
|
| 1529 |
"step": 2500
|
| 1530 |
},
|
| 1531 |
{
|
| 1532 |
"epoch": 1.3,
|
| 1533 |
"learning_rate": 0.0004341809375540564,
|
| 1534 |
-
"loss": 1.
|
| 1535 |
"step": 2510
|
| 1536 |
},
|
| 1537 |
{
|
| 1538 |
"epoch": 1.31,
|
| 1539 |
"learning_rate": 0.0004359107420861443,
|
| 1540 |
-
"loss": 1.
|
| 1541 |
"step": 2520
|
| 1542 |
},
|
| 1543 |
{
|
| 1544 |
"epoch": 1.31,
|
| 1545 |
"learning_rate": 0.00043764054661823217,
|
| 1546 |
-
"loss": 1.
|
| 1547 |
"step": 2530
|
| 1548 |
},
|
| 1549 |
{
|
| 1550 |
"epoch": 1.32,
|
| 1551 |
"learning_rate": 0.00043937035115032005,
|
| 1552 |
-
"loss": 1.
|
| 1553 |
"step": 2540
|
| 1554 |
},
|
| 1555 |
{
|
| 1556 |
"epoch": 1.32,
|
| 1557 |
"learning_rate": 0.0004411001556824079,
|
| 1558 |
-
"loss": 1.
|
| 1559 |
"step": 2550
|
| 1560 |
},
|
| 1561 |
{
|
| 1562 |
"epoch": 1.33,
|
| 1563 |
"learning_rate": 0.0004428299602144958,
|
| 1564 |
-
"loss": 1.
|
| 1565 |
"step": 2560
|
| 1566 |
},
|
| 1567 |
{
|
| 1568 |
"epoch": 1.33,
|
| 1569 |
"learning_rate": 0.0004445597647465837,
|
| 1570 |
-
"loss":
|
| 1571 |
"step": 2570
|
| 1572 |
},
|
| 1573 |
{
|
| 1574 |
"epoch": 1.34,
|
| 1575 |
"learning_rate": 0.00044628956927867156,
|
| 1576 |
-
"loss": 1.
|
| 1577 |
"step": 2580
|
| 1578 |
},
|
| 1579 |
{
|
| 1580 |
"epoch": 1.34,
|
| 1581 |
"learning_rate": 0.0004480193738107594,
|
| 1582 |
-
"loss": 1.
|
| 1583 |
"step": 2590
|
| 1584 |
},
|
| 1585 |
{
|
| 1586 |
"epoch": 1.35,
|
| 1587 |
"learning_rate": 0.00044974917834284726,
|
| 1588 |
-
"loss": 1.
|
| 1589 |
"step": 2600
|
| 1590 |
},
|
| 1591 |
{
|
| 1592 |
"epoch": 1.35,
|
| 1593 |
"learning_rate": 0.00045147898287493514,
|
| 1594 |
-
"loss": 1.
|
| 1595 |
"step": 2610
|
| 1596 |
},
|
| 1597 |
{
|
| 1598 |
"epoch": 1.36,
|
| 1599 |
"learning_rate": 0.000453208787407023,
|
| 1600 |
-
"loss": 1.
|
| 1601 |
"step": 2620
|
| 1602 |
},
|
| 1603 |
{
|
| 1604 |
"epoch": 1.36,
|
| 1605 |
"learning_rate": 0.0004549385919391109,
|
| 1606 |
-
"loss": 1.
|
| 1607 |
"step": 2630
|
| 1608 |
},
|
| 1609 |
{
|
| 1610 |
"epoch": 1.37,
|
| 1611 |
"learning_rate": 0.0004566683964711988,
|
| 1612 |
-
"loss": 1.
|
| 1613 |
"step": 2640
|
| 1614 |
},
|
| 1615 |
{
|
| 1616 |
"epoch": 1.38,
|
| 1617 |
"learning_rate": 0.00045839820100328666,
|
| 1618 |
-
"loss": 1.
|
| 1619 |
"step": 2650
|
| 1620 |
},
|
| 1621 |
{
|
| 1622 |
"epoch": 1.38,
|
| 1623 |
"learning_rate": 0.00046012800553537454,
|
| 1624 |
-
"loss": 1.
|
| 1625 |
"step": 2660
|
| 1626 |
},
|
| 1627 |
{
|
| 1628 |
"epoch": 1.39,
|
| 1629 |
"learning_rate": 0.0004618578100674624,
|
| 1630 |
-
"loss": 1.
|
| 1631 |
"step": 2670
|
| 1632 |
},
|
| 1633 |
{
|
| 1634 |
"epoch": 1.39,
|
| 1635 |
"learning_rate": 0.00046358761459955024,
|
| 1636 |
-
"loss": 1.
|
| 1637 |
"step": 2680
|
| 1638 |
},
|
| 1639 |
{
|
| 1640 |
"epoch": 1.4,
|
| 1641 |
"learning_rate": 0.0004653174191316381,
|
| 1642 |
-
"loss": 1.
|
| 1643 |
"step": 2690
|
| 1644 |
},
|
| 1645 |
{
|
| 1646 |
"epoch": 1.4,
|
| 1647 |
"learning_rate": 0.000467047223663726,
|
| 1648 |
-
"loss": 1.
|
| 1649 |
"step": 2700
|
| 1650 |
},
|
| 1651 |
{
|
| 1652 |
"epoch": 1.41,
|
| 1653 |
"learning_rate": 0.0004687770281958139,
|
| 1654 |
-
"loss": 1.
|
| 1655 |
"step": 2710
|
| 1656 |
},
|
| 1657 |
{
|
| 1658 |
"epoch": 1.41,
|
| 1659 |
"learning_rate": 0.00047050683272790175,
|
| 1660 |
-
"loss": 1.
|
| 1661 |
"step": 2720
|
| 1662 |
},
|
| 1663 |
{
|
| 1664 |
"epoch": 1.42,
|
| 1665 |
"learning_rate": 0.00047223663725998963,
|
| 1666 |
-
"loss": 1.
|
| 1667 |
"step": 2730
|
| 1668 |
},
|
| 1669 |
{
|
| 1670 |
"epoch": 1.42,
|
| 1671 |
"learning_rate": 0.0004739664417920775,
|
| 1672 |
-
"loss":
|
| 1673 |
"step": 2740
|
| 1674 |
},
|
| 1675 |
{
|
| 1676 |
"epoch": 1.43,
|
| 1677 |
"learning_rate": 0.00047569624632416534,
|
| 1678 |
-
"loss": 1.
|
| 1679 |
"step": 2750
|
| 1680 |
},
|
| 1681 |
{
|
| 1682 |
"epoch": 1.43,
|
| 1683 |
"learning_rate": 0.0004774260508562532,
|
| 1684 |
-
"loss": 1.
|
| 1685 |
"step": 2760
|
| 1686 |
},
|
| 1687 |
{
|
| 1688 |
"epoch": 1.44,
|
| 1689 |
"learning_rate": 0.0004791558553883411,
|
| 1690 |
-
"loss": 1.
|
| 1691 |
"step": 2770
|
| 1692 |
},
|
| 1693 |
{
|
| 1694 |
"epoch": 1.44,
|
| 1695 |
"learning_rate": 0.00048088565992042897,
|
| 1696 |
-
"loss": 1.
|
| 1697 |
"step": 2780
|
| 1698 |
},
|
| 1699 |
{
|
| 1700 |
"epoch": 1.45,
|
| 1701 |
"learning_rate": 0.00048261546445251685,
|
| 1702 |
-
"loss": 1.
|
| 1703 |
"step": 2790
|
| 1704 |
},
|
| 1705 |
{
|
| 1706 |
"epoch": 1.45,
|
| 1707 |
"learning_rate": 0.00048434526898460473,
|
| 1708 |
-
"loss": 1.
|
| 1709 |
"step": 2800
|
| 1710 |
},
|
| 1711 |
{
|
| 1712 |
"epoch": 1.46,
|
| 1713 |
"learning_rate": 0.0004860750735166926,
|
| 1714 |
-
"loss": 1.
|
| 1715 |
"step": 2810
|
| 1716 |
},
|
| 1717 |
{
|
| 1718 |
"epoch": 1.46,
|
| 1719 |
"learning_rate": 0.0004878048780487805,
|
| 1720 |
-
"loss":
|
| 1721 |
"step": 2820
|
| 1722 |
},
|
| 1723 |
{
|
| 1724 |
"epoch": 1.47,
|
| 1725 |
"learning_rate": 0.0004895346825808684,
|
| 1726 |
-
"loss": 1.
|
| 1727 |
"step": 2830
|
| 1728 |
},
|
| 1729 |
{
|
| 1730 |
"epoch": 1.47,
|
| 1731 |
"learning_rate": 0.0004912644871129562,
|
| 1732 |
-
"loss": 1.
|
| 1733 |
"step": 2840
|
| 1734 |
},
|
| 1735 |
{
|
| 1736 |
"epoch": 1.48,
|
| 1737 |
"learning_rate": 0.0004929942916450441,
|
| 1738 |
-
"loss": 1.
|
| 1739 |
"step": 2850
|
| 1740 |
},
|
| 1741 |
{
|
| 1742 |
"epoch": 1.48,
|
| 1743 |
"learning_rate": 0.000494724096177132,
|
| 1744 |
-
"loss": 1.
|
| 1745 |
"step": 2860
|
| 1746 |
},
|
| 1747 |
{
|
| 1748 |
"epoch": 1.49,
|
| 1749 |
"learning_rate": 0.0004964539007092199,
|
| 1750 |
-
"loss": 1.
|
| 1751 |
"step": 2870
|
| 1752 |
},
|
| 1753 |
{
|
| 1754 |
"epoch": 1.49,
|
| 1755 |
"learning_rate": 0.0004981837052413078,
|
| 1756 |
-
"loss": 1.
|
| 1757 |
"step": 2880
|
| 1758 |
},
|
| 1759 |
{
|
| 1760 |
"epoch": 1.5,
|
| 1761 |
"learning_rate": 0.0004999135097733956,
|
| 1762 |
-
"loss": 1.
|
| 1763 |
"step": 2890
|
| 1764 |
},
|
| 1765 |
{
|
| 1766 |
"epoch": 1.5,
|
| 1767 |
"learning_rate": 0.0005016433143054835,
|
| 1768 |
-
"loss": 1.
|
| 1769 |
"step": 2900
|
| 1770 |
},
|
| 1771 |
{
|
| 1772 |
"epoch": 1.51,
|
| 1773 |
"learning_rate": 0.0005033731188375714,
|
| 1774 |
-
"loss":
|
| 1775 |
"step": 2910
|
| 1776 |
},
|
| 1777 |
{
|
| 1778 |
"epoch": 1.52,
|
| 1779 |
"learning_rate": 0.0005051029233696593,
|
| 1780 |
-
"loss": 1.
|
| 1781 |
"step": 2920
|
| 1782 |
},
|
| 1783 |
{
|
| 1784 |
"epoch": 1.52,
|
| 1785 |
"learning_rate": 0.000506832727901747,
|
| 1786 |
-
"loss": 1.
|
| 1787 |
"step": 2930
|
| 1788 |
},
|
| 1789 |
{
|
| 1790 |
"epoch": 1.53,
|
| 1791 |
"learning_rate": 0.000508562532433835,
|
| 1792 |
-
"loss": 1.
|
| 1793 |
"step": 2940
|
| 1794 |
},
|
| 1795 |
{
|
| 1796 |
"epoch": 1.53,
|
| 1797 |
"learning_rate": 0.0005102923369659228,
|
| 1798 |
-
"loss": 1.
|
| 1799 |
"step": 2950
|
| 1800 |
},
|
| 1801 |
{
|
| 1802 |
"epoch": 1.54,
|
| 1803 |
"learning_rate": 0.0005120221414980108,
|
| 1804 |
-
"loss": 1.
|
| 1805 |
"step": 2960
|
| 1806 |
},
|
| 1807 |
{
|
| 1808 |
"epoch": 1.54,
|
| 1809 |
"learning_rate": 0.0005137519460300986,
|
| 1810 |
-
"loss":
|
| 1811 |
"step": 2970
|
| 1812 |
},
|
| 1813 |
{
|
| 1814 |
"epoch": 1.55,
|
| 1815 |
"learning_rate": 0.0005154817505621864,
|
| 1816 |
-
"loss":
|
| 1817 |
"step": 2980
|
| 1818 |
},
|
| 1819 |
{
|
| 1820 |
"epoch": 1.55,
|
| 1821 |
"learning_rate": 0.0005172115550942743,
|
| 1822 |
-
"loss": 1.
|
| 1823 |
"step": 2990
|
| 1824 |
},
|
| 1825 |
{
|
| 1826 |
"epoch": 1.56,
|
| 1827 |
"learning_rate": 0.0005189413596263622,
|
| 1828 |
-
"loss": 1.
|
| 1829 |
"step": 3000
|
| 1830 |
},
|
| 1831 |
{
|
| 1832 |
"epoch": 1.56,
|
| 1833 |
"learning_rate": 0.0005206711641584501,
|
| 1834 |
-
"loss":
|
| 1835 |
"step": 3010
|
| 1836 |
},
|
| 1837 |
{
|
| 1838 |
"epoch": 1.57,
|
| 1839 |
"learning_rate": 0.000522400968690538,
|
| 1840 |
-
"loss": 1.
|
| 1841 |
"step": 3020
|
| 1842 |
},
|
| 1843 |
{
|
| 1844 |
"epoch": 1.57,
|
| 1845 |
"learning_rate": 0.0005241307732226258,
|
| 1846 |
-
"loss": 1.
|
| 1847 |
"step": 3030
|
| 1848 |
},
|
| 1849 |
{
|
| 1850 |
"epoch": 1.58,
|
| 1851 |
"learning_rate": 0.0005258605777547137,
|
| 1852 |
-
"loss":
|
| 1853 |
"step": 3040
|
| 1854 |
},
|
| 1855 |
{
|
| 1856 |
"epoch": 1.58,
|
| 1857 |
"learning_rate": 0.0005275903822868016,
|
| 1858 |
-
"loss": 1.
|
| 1859 |
"step": 3050
|
| 1860 |
},
|
| 1861 |
{
|
| 1862 |
"epoch": 1.59,
|
| 1863 |
"learning_rate": 0.0005293201868188895,
|
| 1864 |
-
"loss":
|
| 1865 |
"step": 3060
|
| 1866 |
},
|
| 1867 |
{
|
| 1868 |
"epoch": 1.59,
|
| 1869 |
"learning_rate": 0.0005310499913509773,
|
| 1870 |
-
"loss":
|
| 1871 |
"step": 3070
|
| 1872 |
},
|
| 1873 |
{
|
| 1874 |
"epoch": 1.6,
|
| 1875 |
"learning_rate": 0.0005327797958830652,
|
| 1876 |
-
"loss":
|
| 1877 |
"step": 3080
|
| 1878 |
},
|
| 1879 |
{
|
| 1880 |
"epoch": 1.6,
|
| 1881 |
"learning_rate": 0.0005345096004151531,
|
| 1882 |
-
"loss":
|
| 1883 |
"step": 3090
|
| 1884 |
},
|
| 1885 |
{
|
| 1886 |
"epoch": 1.61,
|
| 1887 |
"learning_rate": 0.000536239404947241,
|
| 1888 |
-
"loss": 1.
|
| 1889 |
"step": 3100
|
| 1890 |
},
|
| 1891 |
{
|
| 1892 |
"epoch": 1.61,
|
| 1893 |
"learning_rate": 0.0005379692094793289,
|
| 1894 |
-
"loss":
|
| 1895 |
"step": 3110
|
| 1896 |
},
|
| 1897 |
{
|
| 1898 |
"epoch": 1.62,
|
| 1899 |
"learning_rate": 0.0005396990140114167,
|
| 1900 |
-
"loss":
|
| 1901 |
"step": 3120
|
| 1902 |
},
|
| 1903 |
{
|
| 1904 |
"epoch": 1.62,
|
| 1905 |
"learning_rate": 0.0005414288185435046,
|
| 1906 |
-
"loss": 1.
|
| 1907 |
"step": 3130
|
| 1908 |
},
|
| 1909 |
{
|
| 1910 |
"epoch": 1.63,
|
| 1911 |
"learning_rate": 0.0005431586230755925,
|
| 1912 |
-
"loss": 1.
|
| 1913 |
"step": 3140
|
| 1914 |
},
|
| 1915 |
{
|
| 1916 |
"epoch": 1.63,
|
| 1917 |
"learning_rate": 0.0005448884276076803,
|
| 1918 |
-
"loss": 1.
|
| 1919 |
"step": 3150
|
| 1920 |
},
|
| 1921 |
{
|
| 1922 |
"epoch": 1.64,
|
| 1923 |
"learning_rate": 0.0005466182321397683,
|
| 1924 |
-
"loss":
|
| 1925 |
"step": 3160
|
| 1926 |
},
|
| 1927 |
{
|
| 1928 |
"epoch": 1.65,
|
| 1929 |
"learning_rate": 0.000548348036671856,
|
| 1930 |
-
"loss": 1.
|
| 1931 |
"step": 3170
|
| 1932 |
},
|
| 1933 |
{
|
| 1934 |
"epoch": 1.65,
|
| 1935 |
"learning_rate": 0.000550077841203944,
|
| 1936 |
-
"loss": 1.
|
| 1937 |
"step": 3180
|
| 1938 |
},
|
| 1939 |
{
|
| 1940 |
"epoch": 1.66,
|
| 1941 |
"learning_rate": 0.0005518076457360318,
|
| 1942 |
-
"loss": 1.
|
| 1943 |
"step": 3190
|
| 1944 |
},
|
| 1945 |
{
|
| 1946 |
"epoch": 1.66,
|
| 1947 |
"learning_rate": 0.0005535374502681198,
|
| 1948 |
-
"loss": 1.
|
| 1949 |
"step": 3200
|
| 1950 |
},
|
| 1951 |
{
|
| 1952 |
"epoch": 1.67,
|
| 1953 |
"learning_rate": 0.0005552672548002075,
|
| 1954 |
-
"loss": 1.
|
| 1955 |
"step": 3210
|
| 1956 |
},
|
| 1957 |
{
|
| 1958 |
"epoch": 1.67,
|
| 1959 |
"learning_rate": 0.0005569970593322955,
|
| 1960 |
-
"loss": 1.
|
| 1961 |
"step": 3220
|
| 1962 |
},
|
| 1963 |
{
|
| 1964 |
"epoch": 1.68,
|
| 1965 |
"learning_rate": 0.0005587268638643833,
|
| 1966 |
-
"loss": 1.
|
| 1967 |
"step": 3230
|
| 1968 |
},
|
| 1969 |
{
|
| 1970 |
"epoch": 1.68,
|
| 1971 |
"learning_rate": 0.0005604566683964713,
|
| 1972 |
-
"loss": 1.
|
| 1973 |
"step": 3240
|
| 1974 |
},
|
| 1975 |
{
|
| 1976 |
"epoch": 1.69,
|
| 1977 |
"learning_rate": 0.000562186472928559,
|
| 1978 |
-
"loss": 1.
|
| 1979 |
"step": 3250
|
| 1980 |
},
|
| 1981 |
{
|
| 1982 |
"epoch": 1.69,
|
| 1983 |
"learning_rate": 0.000563916277460647,
|
| 1984 |
-
"loss": 1.
|
| 1985 |
"step": 3260
|
| 1986 |
},
|
| 1987 |
{
|
| 1988 |
"epoch": 1.7,
|
| 1989 |
"learning_rate": 0.0005656460819927348,
|
| 1990 |
-
"loss": 1.
|
| 1991 |
"step": 3270
|
| 1992 |
},
|
| 1993 |
{
|
| 1994 |
"epoch": 1.7,
|
| 1995 |
"learning_rate": 0.0005673758865248228,
|
| 1996 |
-
"loss": 1.
|
| 1997 |
"step": 3280
|
| 1998 |
},
|
| 1999 |
{
|
| 2000 |
"epoch": 1.71,
|
| 2001 |
"learning_rate": 0.0005691056910569106,
|
| 2002 |
-
"loss": 1.
|
| 2003 |
"step": 3290
|
| 2004 |
},
|
| 2005 |
{
|
| 2006 |
"epoch": 1.71,
|
| 2007 |
"learning_rate": 0.0005708354955889984,
|
| 2008 |
-
"loss": 1.
|
| 2009 |
"step": 3300
|
| 2010 |
},
|
| 2011 |
{
|
| 2012 |
"epoch": 1.72,
|
| 2013 |
"learning_rate": 0.0005725653001210863,
|
| 2014 |
-
"loss": 1.
|
| 2015 |
"step": 3310
|
| 2016 |
},
|
| 2017 |
{
|
| 2018 |
"epoch": 1.72,
|
| 2019 |
"learning_rate": 0.0005742951046531742,
|
| 2020 |
-
"loss": 1.
|
| 2021 |
"step": 3320
|
| 2022 |
},
|
| 2023 |
{
|
| 2024 |
"epoch": 1.73,
|
| 2025 |
"learning_rate": 0.0005760249091852621,
|
| 2026 |
-
"loss": 1.
|
| 2027 |
"step": 3330
|
| 2028 |
},
|
| 2029 |
{
|
| 2030 |
"epoch": 1.73,
|
| 2031 |
"learning_rate": 0.00057775471371735,
|
| 2032 |
-
"loss": 1.
|
| 2033 |
"step": 3340
|
| 2034 |
},
|
| 2035 |
{
|
| 2036 |
"epoch": 1.74,
|
| 2037 |
"learning_rate": 0.0005794845182494378,
|
| 2038 |
-
"loss": 1.
|
| 2039 |
"step": 3350
|
| 2040 |
},
|
| 2041 |
{
|
| 2042 |
"epoch": 1.74,
|
| 2043 |
"learning_rate": 0.0005812143227815257,
|
| 2044 |
-
"loss":
|
| 2045 |
"step": 3360
|
| 2046 |
},
|
| 2047 |
{
|
| 2048 |
"epoch": 1.75,
|
| 2049 |
"learning_rate": 0.0005829441273136136,
|
| 2050 |
-
"loss": 1.
|
| 2051 |
"step": 3370
|
| 2052 |
},
|
| 2053 |
{
|
| 2054 |
"epoch": 1.75,
|
| 2055 |
"learning_rate": 0.0005846739318457015,
|
| 2056 |
-
"loss": 1.
|
| 2057 |
"step": 3380
|
| 2058 |
},
|
| 2059 |
{
|
| 2060 |
"epoch": 1.76,
|
| 2061 |
"learning_rate": 0.0005864037363777894,
|
| 2062 |
-
"loss": 1.
|
| 2063 |
"step": 3390
|
| 2064 |
},
|
| 2065 |
{
|
| 2066 |
"epoch": 1.76,
|
| 2067 |
"learning_rate": 0.0005881335409098772,
|
| 2068 |
-
"loss": 1.
|
| 2069 |
"step": 3400
|
| 2070 |
},
|
| 2071 |
{
|
| 2072 |
"epoch": 1.77,
|
| 2073 |
"learning_rate": 0.0005898633454419651,
|
| 2074 |
-
"loss": 1.
|
| 2075 |
"step": 3410
|
| 2076 |
},
|
| 2077 |
{
|
| 2078 |
"epoch": 1.77,
|
| 2079 |
"learning_rate": 0.0005915931499740529,
|
| 2080 |
-
"loss": 1.
|
| 2081 |
"step": 3420
|
| 2082 |
},
|
| 2083 |
{
|
| 2084 |
"epoch": 1.78,
|
| 2085 |
"learning_rate": 0.0005933229545061409,
|
| 2086 |
-
"loss": 1.
|
| 2087 |
"step": 3430
|
| 2088 |
},
|
| 2089 |
{
|
| 2090 |
"epoch": 1.79,
|
| 2091 |
"learning_rate": 0.0005950527590382286,
|
| 2092 |
-
"loss": 1.
|
| 2093 |
"step": 3440
|
| 2094 |
},
|
| 2095 |
{
|
| 2096 |
"epoch": 1.79,
|
| 2097 |
"learning_rate": 0.0005967825635703165,
|
| 2098 |
-
"loss": 1.
|
| 2099 |
"step": 3450
|
| 2100 |
},
|
| 2101 |
{
|
| 2102 |
"epoch": 1.8,
|
| 2103 |
"learning_rate": 0.0005985123681024044,
|
| 2104 |
-
"loss": 1.
|
| 2105 |
"step": 3460
|
| 2106 |
},
|
| 2107 |
{
|
| 2108 |
"epoch": 1.8,
|
| 2109 |
"learning_rate": 0.0006002421726344923,
|
| 2110 |
-
"loss":
|
| 2111 |
"step": 3470
|
| 2112 |
},
|
| 2113 |
{
|
| 2114 |
"epoch": 1.81,
|
| 2115 |
"learning_rate": 0.0006019719771665802,
|
| 2116 |
-
"loss": 1.
|
| 2117 |
"step": 3480
|
| 2118 |
},
|
| 2119 |
{
|
| 2120 |
"epoch": 1.81,
|
| 2121 |
"learning_rate": 0.000603701781698668,
|
| 2122 |
-
"loss": 1.
|
| 2123 |
"step": 3490
|
| 2124 |
},
|
| 2125 |
{
|
| 2126 |
"epoch": 1.82,
|
| 2127 |
"learning_rate": 0.0006054315862307559,
|
| 2128 |
-
"loss": 1.
|
| 2129 |
"step": 3500
|
| 2130 |
},
|
| 2131 |
{
|
| 2132 |
"epoch": 1.82,
|
| 2133 |
"learning_rate": 0.0006071613907628438,
|
| 2134 |
-
"loss": 1.
|
| 2135 |
"step": 3510
|
| 2136 |
},
|
| 2137 |
{
|
| 2138 |
"epoch": 1.83,
|
| 2139 |
"learning_rate": 0.0006088911952949317,
|
| 2140 |
-
"loss":
|
| 2141 |
"step": 3520
|
| 2142 |
},
|
| 2143 |
{
|
| 2144 |
"epoch": 1.83,
|
| 2145 |
"learning_rate": 0.0006106209998270195,
|
| 2146 |
-
"loss": 1.
|
| 2147 |
"step": 3530
|
| 2148 |
},
|
| 2149 |
{
|
| 2150 |
"epoch": 1.84,
|
| 2151 |
"learning_rate": 0.0006123508043591074,
|
| 2152 |
-
"loss": 1.
|
| 2153 |
"step": 3540
|
| 2154 |
},
|
| 2155 |
{
|
| 2156 |
"epoch": 1.84,
|
| 2157 |
"learning_rate": 0.0006140806088911953,
|
| 2158 |
-
"loss": 1.
|
| 2159 |
"step": 3550
|
| 2160 |
},
|
| 2161 |
{
|
| 2162 |
"epoch": 1.85,
|
| 2163 |
"learning_rate": 0.0006158104134232832,
|
| 2164 |
-
"loss": 1.
|
| 2165 |
"step": 3560
|
| 2166 |
},
|
| 2167 |
{
|
| 2168 |
"epoch": 1.85,
|
| 2169 |
"learning_rate": 0.0006175402179553711,
|
| 2170 |
-
"loss": 1.
|
| 2171 |
"step": 3570
|
| 2172 |
},
|
| 2173 |
{
|
| 2174 |
"epoch": 1.86,
|
| 2175 |
"learning_rate": 0.0006192700224874589,
|
| 2176 |
-
"loss": 1.
|
| 2177 |
"step": 3580
|
| 2178 |
},
|
| 2179 |
{
|
| 2180 |
"epoch": 1.86,
|
| 2181 |
"learning_rate": 0.0006209998270195468,
|
| 2182 |
-
"loss": 1.
|
| 2183 |
"step": 3590
|
| 2184 |
},
|
| 2185 |
{
|
| 2186 |
"epoch": 1.87,
|
| 2187 |
"learning_rate": 0.0006227296315516347,
|
| 2188 |
-
"loss":
|
| 2189 |
"step": 3600
|
| 2190 |
},
|
| 2191 |
{
|
| 2192 |
"epoch": 1.87,
|
| 2193 |
"learning_rate": 0.0006244594360837226,
|
| 2194 |
-
"loss": 1.
|
| 2195 |
"step": 3610
|
| 2196 |
},
|
| 2197 |
{
|
| 2198 |
"epoch": 1.88,
|
| 2199 |
"learning_rate": 0.0006261892406158103,
|
| 2200 |
-
"loss": 1.
|
| 2201 |
"step": 3620
|
| 2202 |
},
|
| 2203 |
{
|
| 2204 |
"epoch": 1.88,
|
| 2205 |
"learning_rate": 0.0006279190451478983,
|
| 2206 |
-
"loss":
|
| 2207 |
"step": 3630
|
| 2208 |
},
|
| 2209 |
{
|
| 2210 |
"epoch": 1.89,
|
| 2211 |
"learning_rate": 0.0006296488496799861,
|
| 2212 |
-
"loss":
|
| 2213 |
"step": 3640
|
| 2214 |
},
|
| 2215 |
{
|
| 2216 |
"epoch": 1.89,
|
| 2217 |
"learning_rate": 0.0006313786542120741,
|
| 2218 |
-
"loss":
|
| 2219 |
"step": 3650
|
| 2220 |
},
|
| 2221 |
{
|
| 2222 |
"epoch": 1.9,
|
| 2223 |
"learning_rate": 0.0006331084587441619,
|
| 2224 |
-
"loss":
|
| 2225 |
"step": 3660
|
| 2226 |
},
|
| 2227 |
{
|
| 2228 |
"epoch": 1.9,
|
| 2229 |
"learning_rate": 0.0006348382632762499,
|
| 2230 |
-
"loss": 1.
|
| 2231 |
"step": 3670
|
| 2232 |
},
|
| 2233 |
{
|
| 2234 |
"epoch": 1.91,
|
| 2235 |
"learning_rate": 0.0006365680678083376,
|
| 2236 |
-
"loss":
|
| 2237 |
"step": 3680
|
| 2238 |
},
|
| 2239 |
{
|
| 2240 |
"epoch": 1.91,
|
| 2241 |
"learning_rate": 0.0006382978723404256,
|
| 2242 |
-
"loss": 2.
|
| 2243 |
"step": 3690
|
| 2244 |
},
|
| 2245 |
{
|
| 2246 |
"epoch": 1.92,
|
| 2247 |
"learning_rate": 0.0006400276768725134,
|
| 2248 |
-
"loss":
|
| 2249 |
"step": 3700
|
| 2250 |
},
|
| 2251 |
{
|
| 2252 |
"epoch": 1.93,
|
| 2253 |
"learning_rate": 0.0006417574814046014,
|
| 2254 |
-
"loss":
|
| 2255 |
"step": 3710
|
| 2256 |
},
|
| 2257 |
{
|
| 2258 |
"epoch": 1.93,
|
| 2259 |
"learning_rate": 0.0006434872859366891,
|
| 2260 |
-
"loss":
|
| 2261 |
"step": 3720
|
| 2262 |
},
|
| 2263 |
{
|
| 2264 |
"epoch": 1.94,
|
| 2265 |
"learning_rate": 0.0006452170904687771,
|
| 2266 |
-
"loss": 1.
|
| 2267 |
"step": 3730
|
| 2268 |
},
|
| 2269 |
{
|
| 2270 |
"epoch": 1.94,
|
| 2271 |
"learning_rate": 0.0006469468950008649,
|
| 2272 |
-
"loss": 1.
|
| 2273 |
"step": 3740
|
| 2274 |
},
|
| 2275 |
{
|
| 2276 |
"epoch": 1.95,
|
| 2277 |
"learning_rate": 0.0006486766995329529,
|
| 2278 |
-
"loss":
|
| 2279 |
"step": 3750
|
| 2280 |
},
|
| 2281 |
{
|
| 2282 |
"epoch": 1.95,
|
| 2283 |
"learning_rate": 0.0006504065040650406,
|
| 2284 |
-
"loss": 1.
|
| 2285 |
"step": 3760
|
| 2286 |
},
|
| 2287 |
{
|
| 2288 |
"epoch": 1.96,
|
| 2289 |
"learning_rate": 0.0006521363085971285,
|
| 2290 |
-
"loss": 1.
|
| 2291 |
"step": 3770
|
| 2292 |
},
|
| 2293 |
{
|
| 2294 |
"epoch": 1.96,
|
| 2295 |
"learning_rate": 0.0006538661131292164,
|
| 2296 |
-
"loss":
|
| 2297 |
"step": 3780
|
| 2298 |
},
|
| 2299 |
{
|
| 2300 |
"epoch": 1.97,
|
| 2301 |
"learning_rate": 0.0006555959176613043,
|
| 2302 |
-
"loss":
|
| 2303 |
"step": 3790
|
| 2304 |
},
|
| 2305 |
{
|
| 2306 |
"epoch": 1.97,
|
| 2307 |
"learning_rate": 0.0006573257221933922,
|
| 2308 |
-
"loss": 1.
|
| 2309 |
"step": 3800
|
| 2310 |
},
|
| 2311 |
{
|
| 2312 |
"epoch": 1.98,
|
| 2313 |
"learning_rate": 0.00065905552672548,
|
| 2314 |
-
"loss": 1.
|
| 2315 |
"step": 3810
|
| 2316 |
},
|
| 2317 |
{
|
| 2318 |
"epoch": 1.98,
|
| 2319 |
"learning_rate": 0.0006607853312575679,
|
| 2320 |
-
"loss": 1.
|
| 2321 |
"step": 3820
|
| 2322 |
},
|
| 2323 |
{
|
| 2324 |
"epoch": 1.99,
|
| 2325 |
"learning_rate": 0.0006625151357896558,
|
| 2326 |
-
"loss":
|
| 2327 |
"step": 3830
|
| 2328 |
},
|
| 2329 |
{
|
| 2330 |
"epoch": 1.99,
|
| 2331 |
"learning_rate": 0.0006642449403217437,
|
| 2332 |
-
"loss": 1.
|
| 2333 |
"step": 3840
|
| 2334 |
},
|
| 2335 |
{
|
| 2336 |
"epoch": 2.0,
|
| 2337 |
"learning_rate": 0.0006659747448538316,
|
| 2338 |
-
"loss":
|
| 2339 |
"step": 3850
|
| 2340 |
},
|
| 2341 |
{
|
| 2342 |
"epoch": 2.0,
|
| 2343 |
"eval_accuracy": {
|
| 2344 |
-
"accuracy": 0.
|
| 2345 |
},
|
| 2346 |
"eval_f1": {
|
| 2347 |
-
"f1": 0.
|
| 2348 |
},
|
| 2349 |
-
"eval_loss": 1.
|
| 2350 |
"eval_precision": {
|
| 2351 |
-
"precision": 0.
|
| 2352 |
},
|
| 2353 |
"eval_recall": {
|
| 2354 |
-
"recall": 0.
|
| 2355 |
},
|
| 2356 |
-
"eval_runtime":
|
| 2357 |
-
"eval_samples_per_second":
|
| 2358 |
-
"eval_steps_per_second":
|
| 2359 |
"step": 3854
|
| 2360 |
},
|
| 2361 |
{
|
| 2362 |
"epoch": 2.0,
|
| 2363 |
"learning_rate": 0.0006677045493859194,
|
| 2364 |
-
"loss": 1.
|
| 2365 |
"step": 3860
|
| 2366 |
},
|
| 2367 |
{
|
| 2368 |
"epoch": 2.01,
|
| 2369 |
"learning_rate": 0.0006694343539180073,
|
| 2370 |
-
"loss": 1.
|
| 2371 |
"step": 3870
|
| 2372 |
},
|
| 2373 |
{
|
| 2374 |
"epoch": 2.01,
|
| 2375 |
"learning_rate": 0.0006711641584500952,
|
| 2376 |
-
"loss": 1.
|
| 2377 |
"step": 3880
|
| 2378 |
},
|
| 2379 |
{
|
| 2380 |
"epoch": 2.02,
|
| 2381 |
"learning_rate": 0.000672893962982183,
|
| 2382 |
-
"loss": 1.
|
| 2383 |
"step": 3890
|
| 2384 |
},
|
| 2385 |
{
|
| 2386 |
"epoch": 2.02,
|
| 2387 |
"learning_rate": 0.000674623767514271,
|
| 2388 |
-
"loss": 1.
|
| 2389 |
"step": 3900
|
| 2390 |
},
|
| 2391 |
{
|
| 2392 |
"epoch": 2.03,
|
| 2393 |
"learning_rate": 0.0006763535720463587,
|
| 2394 |
-
"loss":
|
| 2395 |
"step": 3910
|
| 2396 |
},
|
| 2397 |
{
|
| 2398 |
"epoch": 2.03,
|
| 2399 |
"learning_rate": 0.0006780833765784466,
|
| 2400 |
-
"loss":
|
| 2401 |
"step": 3920
|
| 2402 |
},
|
| 2403 |
{
|
| 2404 |
"epoch": 2.04,
|
| 2405 |
"learning_rate": 0.0006798131811105345,
|
| 2406 |
-
"loss": 1.
|
| 2407 |
"step": 3930
|
| 2408 |
},
|
| 2409 |
{
|
| 2410 |
"epoch": 2.04,
|
| 2411 |
"learning_rate": 0.0006815429856426224,
|
| 2412 |
-
"loss": 1.
|
| 2413 |
"step": 3940
|
| 2414 |
},
|
| 2415 |
{
|
| 2416 |
"epoch": 2.05,
|
| 2417 |
"learning_rate": 0.0006832727901747102,
|
| 2418 |
-
"loss": 1.
|
| 2419 |
"step": 3950
|
| 2420 |
},
|
| 2421 |
{
|
| 2422 |
"epoch": 2.06,
|
| 2423 |
"learning_rate": 0.0006850025947067981,
|
| 2424 |
-
"loss": 1.
|
| 2425 |
"step": 3960
|
| 2426 |
},
|
| 2427 |
{
|
| 2428 |
"epoch": 2.06,
|
| 2429 |
"learning_rate": 0.000686732399238886,
|
| 2430 |
-
"loss": 1.
|
| 2431 |
"step": 3970
|
| 2432 |
},
|
| 2433 |
{
|
| 2434 |
"epoch": 2.07,
|
| 2435 |
"learning_rate": 0.0006884622037709739,
|
| 2436 |
-
"loss": 1.
|
| 2437 |
"step": 3980
|
| 2438 |
},
|
| 2439 |
{
|
| 2440 |
"epoch": 2.07,
|
| 2441 |
"learning_rate": 0.0006901920083030618,
|
| 2442 |
-
"loss": 1.
|
| 2443 |
"step": 3990
|
| 2444 |
},
|
| 2445 |
{
|
| 2446 |
"epoch": 2.08,
|
| 2447 |
"learning_rate": 0.0006919218128351496,
|
| 2448 |
-
"loss":
|
| 2449 |
"step": 4000
|
| 2450 |
},
|
| 2451 |
{
|
| 2452 |
"epoch": 2.08,
|
| 2453 |
"learning_rate": 0.0006936516173672375,
|
| 2454 |
-
"loss": 1.
|
| 2455 |
"step": 4010
|
| 2456 |
},
|
| 2457 |
{
|
| 2458 |
"epoch": 2.09,
|
| 2459 |
"learning_rate": 0.0006953814218993254,
|
| 2460 |
-
"loss": 1.
|
| 2461 |
"step": 4020
|
| 2462 |
},
|
| 2463 |
{
|
| 2464 |
"epoch": 2.09,
|
| 2465 |
"learning_rate": 0.0006971112264314133,
|
| 2466 |
-
"loss":
|
| 2467 |
"step": 4030
|
| 2468 |
},
|
| 2469 |
{
|
| 2470 |
"epoch": 2.1,
|
| 2471 |
"learning_rate": 0.0006988410309635011,
|
| 2472 |
-
"loss": 1.
|
| 2473 |
"step": 4040
|
| 2474 |
},
|
| 2475 |
{
|
| 2476 |
"epoch": 2.1,
|
| 2477 |
"learning_rate": 0.000700570835495589,
|
| 2478 |
-
"loss": 1.
|
| 2479 |
"step": 4050
|
| 2480 |
},
|
| 2481 |
{
|
| 2482 |
"epoch": 2.11,
|
| 2483 |
"learning_rate": 0.0007023006400276769,
|
| 2484 |
-
"loss": 1.
|
| 2485 |
"step": 4060
|
| 2486 |
},
|
| 2487 |
{
|
| 2488 |
"epoch": 2.11,
|
| 2489 |
"learning_rate": 0.0007040304445597648,
|
| 2490 |
-
"loss": 1.
|
| 2491 |
"step": 4070
|
| 2492 |
},
|
| 2493 |
{
|
| 2494 |
"epoch": 2.12,
|
| 2495 |
"learning_rate": 0.0007057602490918527,
|
| 2496 |
-
"loss": 1.
|
| 2497 |
"step": 4080
|
| 2498 |
},
|
| 2499 |
{
|
| 2500 |
"epoch": 2.12,
|
| 2501 |
"learning_rate": 0.0007074900536239404,
|
| 2502 |
-
"loss": 1.
|
| 2503 |
"step": 4090
|
| 2504 |
},
|
| 2505 |
{
|
| 2506 |
"epoch": 2.13,
|
| 2507 |
"learning_rate": 0.0007092198581560284,
|
| 2508 |
-
"loss": 1.
|
| 2509 |
"step": 4100
|
| 2510 |
},
|
| 2511 |
{
|
| 2512 |
"epoch": 2.13,
|
| 2513 |
"learning_rate": 0.0007109496626881162,
|
| 2514 |
-
"loss": 1.
|
| 2515 |
"step": 4110
|
| 2516 |
},
|
| 2517 |
{
|
| 2518 |
"epoch": 2.14,
|
| 2519 |
"learning_rate": 0.0007126794672202042,
|
| 2520 |
-
"loss": 1.
|
| 2521 |
"step": 4120
|
| 2522 |
},
|
| 2523 |
{
|
| 2524 |
"epoch": 2.14,
|
| 2525 |
"learning_rate": 0.0007144092717522919,
|
| 2526 |
-
"loss": 1.
|
| 2527 |
"step": 4130
|
| 2528 |
},
|
| 2529 |
{
|
| 2530 |
"epoch": 2.15,
|
| 2531 |
"learning_rate": 0.0007161390762843799,
|
| 2532 |
-
"loss": 1.
|
| 2533 |
"step": 4140
|
| 2534 |
},
|
| 2535 |
{
|
| 2536 |
"epoch": 2.15,
|
| 2537 |
"learning_rate": 0.0007178688808164677,
|
| 2538 |
-
"loss": 1.
|
| 2539 |
"step": 4150
|
| 2540 |
},
|
| 2541 |
{
|
| 2542 |
"epoch": 2.16,
|
| 2543 |
"learning_rate": 0.0007195986853485557,
|
| 2544 |
-
"loss": 1.
|
| 2545 |
"step": 4160
|
| 2546 |
},
|
| 2547 |
{
|
| 2548 |
"epoch": 2.16,
|
| 2549 |
"learning_rate": 0.0007213284898806435,
|
| 2550 |
-
"loss": 1.
|
| 2551 |
"step": 4170
|
| 2552 |
},
|
| 2553 |
{
|
| 2554 |
"epoch": 2.17,
|
| 2555 |
"learning_rate": 0.0007230582944127314,
|
| 2556 |
-
"loss": 1.
|
| 2557 |
"step": 4180
|
| 2558 |
},
|
| 2559 |
{
|
| 2560 |
"epoch": 2.17,
|
| 2561 |
"learning_rate": 0.0007247880989448192,
|
| 2562 |
-
"loss": 1.
|
| 2563 |
"step": 4190
|
| 2564 |
},
|
| 2565 |
{
|
| 2566 |
"epoch": 2.18,
|
| 2567 |
"learning_rate": 0.0007265179034769072,
|
| 2568 |
-
"loss": 1.
|
| 2569 |
"step": 4200
|
| 2570 |
},
|
| 2571 |
{
|
| 2572 |
"epoch": 2.18,
|
| 2573 |
"learning_rate": 0.000728247708008995,
|
| 2574 |
-
"loss": 1.
|
| 2575 |
"step": 4210
|
| 2576 |
},
|
| 2577 |
{
|
| 2578 |
"epoch": 2.19,
|
| 2579 |
"learning_rate": 0.000729977512541083,
|
| 2580 |
-
"loss": 1.
|
| 2581 |
"step": 4220
|
| 2582 |
},
|
| 2583 |
{
|
| 2584 |
"epoch": 2.2,
|
| 2585 |
"learning_rate": 0.0007317073170731707,
|
| 2586 |
-
"loss": 1.
|
| 2587 |
"step": 4230
|
| 2588 |
},
|
| 2589 |
{
|
| 2590 |
"epoch": 2.2,
|
| 2591 |
"learning_rate": 0.0007334371216052586,
|
| 2592 |
-
"loss": 1.
|
| 2593 |
"step": 4240
|
| 2594 |
},
|
| 2595 |
{
|
| 2596 |
"epoch": 2.21,
|
| 2597 |
"learning_rate": 0.0007351669261373465,
|
| 2598 |
-
"loss": 1.
|
| 2599 |
"step": 4250
|
| 2600 |
},
|
| 2601 |
{
|
| 2602 |
"epoch": 2.21,
|
| 2603 |
"learning_rate": 0.0007368967306694344,
|
| 2604 |
-
"loss": 1.
|
| 2605 |
"step": 4260
|
| 2606 |
},
|
| 2607 |
{
|
| 2608 |
"epoch": 2.22,
|
| 2609 |
"learning_rate": 0.0007386265352015222,
|
| 2610 |
-
"loss": 1.
|
| 2611 |
"step": 4270
|
| 2612 |
},
|
| 2613 |
{
|
| 2614 |
"epoch": 2.22,
|
| 2615 |
"learning_rate": 0.0007403563397336101,
|
| 2616 |
-
"loss": 1.
|
| 2617 |
"step": 4280
|
| 2618 |
},
|
| 2619 |
{
|
| 2620 |
"epoch": 2.23,
|
| 2621 |
"learning_rate": 0.000742086144265698,
|
| 2622 |
-
"loss": 1.
|
| 2623 |
"step": 4290
|
| 2624 |
},
|
| 2625 |
{
|
| 2626 |
"epoch": 2.23,
|
| 2627 |
"learning_rate": 0.0007438159487977859,
|
| 2628 |
-
"loss": 1.
|
| 2629 |
"step": 4300
|
| 2630 |
},
|
| 2631 |
{
|
| 2632 |
"epoch": 2.24,
|
| 2633 |
"learning_rate": 0.0007455457533298738,
|
| 2634 |
-
"loss": 1.
|
| 2635 |
"step": 4310
|
| 2636 |
},
|
| 2637 |
{
|
| 2638 |
"epoch": 2.24,
|
| 2639 |
"learning_rate": 0.0007472755578619616,
|
| 2640 |
-
"loss": 1.
|
| 2641 |
"step": 4320
|
| 2642 |
},
|
| 2643 |
{
|
| 2644 |
"epoch": 2.25,
|
| 2645 |
"learning_rate": 0.0007490053623940495,
|
| 2646 |
-
"loss": 1.
|
| 2647 |
"step": 4330
|
| 2648 |
},
|
| 2649 |
{
|
| 2650 |
"epoch": 2.25,
|
| 2651 |
"learning_rate": 0.0007507351669261374,
|
| 2652 |
-
"loss": 1.
|
| 2653 |
"step": 4340
|
| 2654 |
},
|
| 2655 |
{
|
| 2656 |
"epoch": 2.26,
|
| 2657 |
"learning_rate": 0.0007524649714582253,
|
| 2658 |
-
"loss": 1.
|
| 2659 |
"step": 4350
|
| 2660 |
},
|
| 2661 |
{
|
| 2662 |
"epoch": 2.26,
|
| 2663 |
"learning_rate": 0.0007541947759903132,
|
| 2664 |
-
"loss": 1.
|
| 2665 |
"step": 4360
|
| 2666 |
},
|
| 2667 |
{
|
| 2668 |
"epoch": 2.27,
|
| 2669 |
"learning_rate": 0.000755924580522401,
|
| 2670 |
-
"loss": 1.
|
| 2671 |
"step": 4370
|
| 2672 |
},
|
| 2673 |
{
|
| 2674 |
"epoch": 2.27,
|
| 2675 |
"learning_rate": 0.0007576543850544888,
|
| 2676 |
-
"loss": 1.
|
| 2677 |
"step": 4380
|
| 2678 |
},
|
| 2679 |
{
|
| 2680 |
"epoch": 2.28,
|
| 2681 |
"learning_rate": 0.0007593841895865767,
|
| 2682 |
-
"loss": 1.
|
| 2683 |
"step": 4390
|
| 2684 |
},
|
| 2685 |
{
|
| 2686 |
"epoch": 2.28,
|
| 2687 |
"learning_rate": 0.0007611139941186646,
|
| 2688 |
-
"loss": 1.
|
| 2689 |
"step": 4400
|
| 2690 |
},
|
| 2691 |
{
|
| 2692 |
"epoch": 2.29,
|
| 2693 |
"learning_rate": 0.0007628437986507524,
|
| 2694 |
-
"loss": 1.
|
| 2695 |
"step": 4410
|
| 2696 |
},
|
| 2697 |
{
|
| 2698 |
"epoch": 2.29,
|
| 2699 |
"learning_rate": 0.0007645736031828403,
|
| 2700 |
-
"loss":
|
| 2701 |
"step": 4420
|
| 2702 |
},
|
| 2703 |
{
|
| 2704 |
"epoch": 2.3,
|
| 2705 |
"learning_rate": 0.0007663034077149282,
|
| 2706 |
-
"loss": 1.
|
| 2707 |
"step": 4430
|
| 2708 |
},
|
| 2709 |
{
|
| 2710 |
"epoch": 2.3,
|
| 2711 |
"learning_rate": 0.0007680332122470161,
|
| 2712 |
-
"loss":
|
| 2713 |
"step": 4440
|
| 2714 |
},
|
| 2715 |
{
|
| 2716 |
"epoch": 2.31,
|
| 2717 |
"learning_rate": 0.000769763016779104,
|
| 2718 |
-
"loss":
|
| 2719 |
"step": 4450
|
| 2720 |
},
|
| 2721 |
{
|
| 2722 |
"epoch": 2.31,
|
| 2723 |
"learning_rate": 0.0007714928213111918,
|
| 2724 |
-
"loss": 1.
|
| 2725 |
"step": 4460
|
| 2726 |
},
|
| 2727 |
{
|
| 2728 |
"epoch": 2.32,
|
| 2729 |
"learning_rate": 0.0007732226258432797,
|
| 2730 |
-
"loss": 1.
|
| 2731 |
"step": 4470
|
| 2732 |
},
|
| 2733 |
{
|
| 2734 |
"epoch": 2.32,
|
| 2735 |
"learning_rate": 0.0007749524303753676,
|
| 2736 |
-
"loss": 1.
|
| 2737 |
"step": 4480
|
| 2738 |
},
|
| 2739 |
{
|
| 2740 |
"epoch": 2.33,
|
| 2741 |
"learning_rate": 0.0007766822349074555,
|
| 2742 |
-
"loss": 1.
|
| 2743 |
"step": 4490
|
| 2744 |
},
|
| 2745 |
{
|
| 2746 |
"epoch": 2.34,
|
| 2747 |
"learning_rate": 0.0007784120394395433,
|
| 2748 |
-
"loss": 1.
|
| 2749 |
"step": 4500
|
| 2750 |
},
|
| 2751 |
{
|
| 2752 |
"epoch": 2.34,
|
| 2753 |
"learning_rate": 0.0007801418439716312,
|
| 2754 |
-
"loss": 1.
|
| 2755 |
"step": 4510
|
| 2756 |
},
|
| 2757 |
{
|
| 2758 |
"epoch": 2.35,
|
| 2759 |
"learning_rate": 0.0007818716485037191,
|
| 2760 |
-
"loss":
|
| 2761 |
"step": 4520
|
| 2762 |
},
|
| 2763 |
{
|
| 2764 |
"epoch": 2.35,
|
| 2765 |
"learning_rate": 0.000783601453035807,
|
| 2766 |
-
"loss": 1.
|
| 2767 |
"step": 4530
|
| 2768 |
},
|
| 2769 |
{
|
| 2770 |
"epoch": 2.36,
|
| 2771 |
"learning_rate": 0.0007853312575678949,
|
| 2772 |
-
"loss": 1.
|
| 2773 |
"step": 4540
|
| 2774 |
},
|
| 2775 |
{
|
| 2776 |
"epoch": 2.36,
|
| 2777 |
"learning_rate": 0.0007870610620999827,
|
| 2778 |
-
"loss": 1.
|
| 2779 |
"step": 4550
|
| 2780 |
},
|
| 2781 |
{
|
| 2782 |
"epoch": 2.37,
|
| 2783 |
"learning_rate": 0.0007887908666320705,
|
| 2784 |
-
"loss": 1.
|
| 2785 |
"step": 4560
|
| 2786 |
},
|
| 2787 |
{
|
| 2788 |
"epoch": 2.37,
|
| 2789 |
"learning_rate": 0.0007905206711641585,
|
| 2790 |
-
"loss": 1.
|
| 2791 |
"step": 4570
|
| 2792 |
},
|
| 2793 |
{
|
| 2794 |
"epoch": 2.38,
|
| 2795 |
"learning_rate": 0.0007922504756962463,
|
| 2796 |
-
"loss": 1.
|
| 2797 |
"step": 4580
|
| 2798 |
},
|
| 2799 |
{
|
| 2800 |
"epoch": 2.38,
|
| 2801 |
"learning_rate": 0.0007939802802283343,
|
| 2802 |
-
"loss": 1.
|
| 2803 |
"step": 4590
|
| 2804 |
},
|
| 2805 |
{
|
| 2806 |
"epoch": 2.39,
|
| 2807 |
"learning_rate": 0.000795710084760422,
|
| 2808 |
-
"loss": 1.
|
| 2809 |
"step": 4600
|
| 2810 |
},
|
| 2811 |
{
|
| 2812 |
"epoch": 2.39,
|
| 2813 |
"learning_rate": 0.00079743988929251,
|
| 2814 |
-
"loss": 1.
|
| 2815 |
"step": 4610
|
| 2816 |
},
|
| 2817 |
{
|
| 2818 |
"epoch": 2.4,
|
| 2819 |
"learning_rate": 0.0007991696938245978,
|
| 2820 |
-
"loss": 1.
|
| 2821 |
"step": 4620
|
| 2822 |
},
|
| 2823 |
{
|
| 2824 |
"epoch": 2.4,
|
| 2825 |
"learning_rate": 0.0008008994983566858,
|
| 2826 |
-
"loss": 1.
|
| 2827 |
"step": 4630
|
| 2828 |
},
|
| 2829 |
{
|
| 2830 |
"epoch": 2.41,
|
| 2831 |
"learning_rate": 0.0008026293028887735,
|
| 2832 |
-
"loss": 1.
|
| 2833 |
"step": 4640
|
| 2834 |
},
|
| 2835 |
{
|
| 2836 |
"epoch": 2.41,
|
| 2837 |
"learning_rate": 0.0008043591074208615,
|
| 2838 |
-
"loss": 1.
|
| 2839 |
"step": 4650
|
| 2840 |
},
|
| 2841 |
{
|
| 2842 |
"epoch": 2.42,
|
| 2843 |
"learning_rate": 0.0008060889119529493,
|
| 2844 |
-
"loss": 1.
|
| 2845 |
"step": 4660
|
| 2846 |
},
|
| 2847 |
{
|
| 2848 |
"epoch": 2.42,
|
| 2849 |
"learning_rate": 0.0008078187164850373,
|
| 2850 |
-
"loss": 1.
|
| 2851 |
"step": 4670
|
| 2852 |
},
|
| 2853 |
{
|
| 2854 |
"epoch": 2.43,
|
| 2855 |
"learning_rate": 0.0008095485210171251,
|
| 2856 |
-
"loss": 1.
|
| 2857 |
"step": 4680
|
| 2858 |
},
|
| 2859 |
{
|
| 2860 |
"epoch": 2.43,
|
| 2861 |
"learning_rate": 0.000811278325549213,
|
| 2862 |
-
"loss": 1.
|
| 2863 |
"step": 4690
|
| 2864 |
},
|
| 2865 |
{
|
| 2866 |
"epoch": 2.44,
|
| 2867 |
"learning_rate": 0.0008130081300813008,
|
| 2868 |
-
"loss": 1.
|
| 2869 |
"step": 4700
|
| 2870 |
},
|
| 2871 |
{
|
| 2872 |
"epoch": 2.44,
|
| 2873 |
"learning_rate": 0.0008147379346133887,
|
| 2874 |
-
"loss": 1.
|
| 2875 |
"step": 4710
|
| 2876 |
},
|
| 2877 |
{
|
| 2878 |
"epoch": 2.45,
|
| 2879 |
"learning_rate": 0.0008164677391454766,
|
| 2880 |
-
"loss": 1.
|
| 2881 |
"step": 4720
|
| 2882 |
},
|
| 2883 |
{
|
| 2884 |
"epoch": 2.45,
|
| 2885 |
"learning_rate": 0.0008181975436775644,
|
| 2886 |
-
"loss": 1.
|
| 2887 |
"step": 4730
|
| 2888 |
},
|
| 2889 |
{
|
| 2890 |
"epoch": 2.46,
|
| 2891 |
"learning_rate": 0.0008199273482096523,
|
| 2892 |
-
"loss": 1.
|
| 2893 |
"step": 4740
|
| 2894 |
},
|
| 2895 |
{
|
| 2896 |
"epoch": 2.46,
|
| 2897 |
"learning_rate": 0.0008216571527417402,
|
| 2898 |
-
"loss": 1.
|
| 2899 |
"step": 4750
|
| 2900 |
},
|
| 2901 |
{
|
| 2902 |
"epoch": 2.47,
|
| 2903 |
"learning_rate": 0.0008233869572738281,
|
| 2904 |
-
"loss": 1.
|
| 2905 |
"step": 4760
|
| 2906 |
},
|
| 2907 |
{
|
| 2908 |
"epoch": 2.48,
|
| 2909 |
"learning_rate": 0.000825116761805916,
|
| 2910 |
-
"loss": 1.
|
| 2911 |
"step": 4770
|
| 2912 |
},
|
| 2913 |
{
|
| 2914 |
"epoch": 2.48,
|
| 2915 |
"learning_rate": 0.0008268465663380038,
|
| 2916 |
-
"loss": 1.
|
| 2917 |
"step": 4780
|
| 2918 |
},
|
| 2919 |
{
|
| 2920 |
"epoch": 2.49,
|
| 2921 |
"learning_rate": 0.0008285763708700917,
|
| 2922 |
-
"loss": 1.
|
| 2923 |
"step": 4790
|
| 2924 |
},
|
| 2925 |
{
|
| 2926 |
"epoch": 2.49,
|
| 2927 |
"learning_rate": 0.0008303061754021796,
|
| 2928 |
-
"loss": 1.
|
| 2929 |
"step": 4800
|
| 2930 |
},
|
| 2931 |
{
|
| 2932 |
"epoch": 2.5,
|
| 2933 |
"learning_rate": 0.0008320359799342675,
|
| 2934 |
-
"loss": 1.
|
| 2935 |
"step": 4810
|
| 2936 |
},
|
| 2937 |
{
|
| 2938 |
"epoch": 2.5,
|
| 2939 |
"learning_rate": 0.0008337657844663554,
|
| 2940 |
-
"loss": 1.
|
| 2941 |
"step": 4820
|
| 2942 |
},
|
| 2943 |
{
|
| 2944 |
"epoch": 2.51,
|
| 2945 |
"learning_rate": 0.0008354955889984432,
|
| 2946 |
-
"loss":
|
| 2947 |
"step": 4830
|
| 2948 |
},
|
| 2949 |
{
|
| 2950 |
"epoch": 2.51,
|
| 2951 |
"learning_rate": 0.0008372253935305311,
|
| 2952 |
-
"loss": 1.
|
| 2953 |
"step": 4840
|
| 2954 |
},
|
| 2955 |
{
|
| 2956 |
"epoch": 2.52,
|
| 2957 |
"learning_rate": 0.000838955198062619,
|
| 2958 |
-
"loss": 1.
|
| 2959 |
"step": 4850
|
| 2960 |
},
|
| 2961 |
{
|
| 2962 |
"epoch": 2.52,
|
| 2963 |
"learning_rate": 0.0008406850025947068,
|
| 2964 |
-
"loss": 1.
|
| 2965 |
"step": 4860
|
| 2966 |
},
|
| 2967 |
{
|
| 2968 |
"epoch": 2.53,
|
| 2969 |
"learning_rate": 0.0008424148071267946,
|
| 2970 |
-
"loss": 1.
|
| 2971 |
"step": 4870
|
| 2972 |
},
|
| 2973 |
{
|
| 2974 |
"epoch": 2.53,
|
| 2975 |
"learning_rate": 0.0008441446116588825,
|
| 2976 |
-
"loss": 1.
|
| 2977 |
"step": 4880
|
| 2978 |
},
|
| 2979 |
{
|
| 2980 |
"epoch": 2.54,
|
| 2981 |
"learning_rate": 0.0008458744161909704,
|
| 2982 |
-
"loss": 1.
|
| 2983 |
"step": 4890
|
| 2984 |
},
|
| 2985 |
{
|
| 2986 |
"epoch": 2.54,
|
| 2987 |
"learning_rate": 0.0008476042207230583,
|
| 2988 |
-
"loss": 1.
|
| 2989 |
"step": 4900
|
| 2990 |
},
|
| 2991 |
{
|
| 2992 |
"epoch": 2.55,
|
| 2993 |
"learning_rate": 0.0008493340252551462,
|
| 2994 |
-
"loss": 1.
|
| 2995 |
"step": 4910
|
| 2996 |
},
|
| 2997 |
{
|
| 2998 |
"epoch": 2.55,
|
| 2999 |
"learning_rate": 0.000851063829787234,
|
| 3000 |
-
"loss": 1.
|
| 3001 |
"step": 4920
|
| 3002 |
},
|
| 3003 |
{
|
| 3004 |
"epoch": 2.56,
|
| 3005 |
"learning_rate": 0.0008527936343193219,
|
| 3006 |
-
"loss": 1.
|
| 3007 |
"step": 4930
|
| 3008 |
},
|
| 3009 |
{
|
| 3010 |
"epoch": 2.56,
|
| 3011 |
"learning_rate": 0.0008545234388514098,
|
| 3012 |
-
"loss": 1.
|
| 3013 |
"step": 4940
|
| 3014 |
},
|
| 3015 |
{
|
| 3016 |
"epoch": 2.57,
|
| 3017 |
"learning_rate": 0.0008562532433834977,
|
| 3018 |
-
"loss": 1.
|
| 3019 |
"step": 4950
|
| 3020 |
},
|
| 3021 |
{
|
| 3022 |
"epoch": 2.57,
|
| 3023 |
"learning_rate": 0.0008579830479155855,
|
| 3024 |
-
"loss": 1.
|
| 3025 |
"step": 4960
|
| 3026 |
},
|
| 3027 |
{
|
| 3028 |
"epoch": 2.58,
|
| 3029 |
"learning_rate": 0.0008597128524476734,
|
| 3030 |
-
"loss": 1.
|
| 3031 |
"step": 4970
|
| 3032 |
},
|
| 3033 |
{
|
| 3034 |
"epoch": 2.58,
|
| 3035 |
"learning_rate": 0.0008614426569797613,
|
| 3036 |
-
"loss": 1.
|
| 3037 |
"step": 4980
|
| 3038 |
},
|
| 3039 |
{
|
| 3040 |
"epoch": 2.59,
|
| 3041 |
"learning_rate": 0.0008631724615118492,
|
| 3042 |
-
"loss": 1.
|
| 3043 |
"step": 4990
|
| 3044 |
},
|
| 3045 |
{
|
| 3046 |
"epoch": 2.59,
|
| 3047 |
"learning_rate": 0.0008649022660439371,
|
| 3048 |
-
"loss": 1.
|
| 3049 |
"step": 5000
|
| 3050 |
},
|
| 3051 |
{
|
| 3052 |
"epoch": 2.6,
|
| 3053 |
"learning_rate": 0.0008666320705760249,
|
| 3054 |
-
"loss": 1.
|
| 3055 |
"step": 5010
|
| 3056 |
},
|
| 3057 |
{
|
| 3058 |
"epoch": 2.61,
|
| 3059 |
"learning_rate": 0.0008683618751081128,
|
| 3060 |
-
"loss": 1.
|
| 3061 |
"step": 5020
|
| 3062 |
},
|
| 3063 |
{
|
| 3064 |
"epoch": 2.61,
|
| 3065 |
"learning_rate": 0.0008700916796402006,
|
| 3066 |
-
"loss": 1.
|
| 3067 |
"step": 5030
|
| 3068 |
},
|
| 3069 |
{
|
| 3070 |
"epoch": 2.62,
|
| 3071 |
"learning_rate": 0.0008718214841722886,
|
| 3072 |
-
"loss": 1.
|
| 3073 |
"step": 5040
|
| 3074 |
},
|
| 3075 |
{
|
| 3076 |
"epoch": 2.62,
|
| 3077 |
"learning_rate": 0.0008735512887043763,
|
| 3078 |
-
"loss":
|
| 3079 |
"step": 5050
|
| 3080 |
},
|
| 3081 |
{
|
| 3082 |
"epoch": 2.63,
|
| 3083 |
"learning_rate": 0.0008752810932364643,
|
| 3084 |
-
"loss": 1.
|
| 3085 |
"step": 5060
|
| 3086 |
},
|
| 3087 |
{
|
| 3088 |
"epoch": 2.63,
|
| 3089 |
"learning_rate": 0.0008770108977685521,
|
| 3090 |
-
"loss": 1.
|
| 3091 |
"step": 5070
|
| 3092 |
},
|
| 3093 |
{
|
| 3094 |
"epoch": 2.64,
|
| 3095 |
"learning_rate": 0.0008787407023006401,
|
| 3096 |
-
"loss":
|
| 3097 |
"step": 5080
|
| 3098 |
},
|
| 3099 |
{
|
| 3100 |
"epoch": 2.64,
|
| 3101 |
"learning_rate": 0.0008804705068327279,
|
| 3102 |
-
"loss": 1.
|
| 3103 |
"step": 5090
|
| 3104 |
},
|
| 3105 |
{
|
| 3106 |
"epoch": 2.65,
|
| 3107 |
"learning_rate": 0.0008822003113648159,
|
| 3108 |
-
"loss": 1.
|
| 3109 |
"step": 5100
|
| 3110 |
},
|
| 3111 |
{
|
| 3112 |
"epoch": 2.65,
|
| 3113 |
"learning_rate": 0.0008839301158969036,
|
| 3114 |
-
"loss":
|
| 3115 |
"step": 5110
|
| 3116 |
},
|
| 3117 |
{
|
| 3118 |
"epoch": 2.66,
|
| 3119 |
"learning_rate": 0.0008856599204289916,
|
| 3120 |
-
"loss": 1.
|
| 3121 |
"step": 5120
|
| 3122 |
},
|
| 3123 |
{
|
| 3124 |
"epoch": 2.66,
|
| 3125 |
"learning_rate": 0.0008873897249610794,
|
| 3126 |
-
"loss":
|
| 3127 |
"step": 5130
|
| 3128 |
},
|
| 3129 |
{
|
| 3130 |
"epoch": 2.67,
|
| 3131 |
"learning_rate": 0.0008891195294931674,
|
| 3132 |
-
"loss": 1.
|
| 3133 |
"step": 5140
|
| 3134 |
},
|
| 3135 |
{
|
| 3136 |
"epoch": 2.67,
|
| 3137 |
"learning_rate": 0.0008908493340252551,
|
| 3138 |
-
"loss": 1.
|
| 3139 |
"step": 5150
|
| 3140 |
},
|
| 3141 |
{
|
| 3142 |
"epoch": 2.68,
|
| 3143 |
"learning_rate": 0.0008925791385573431,
|
| 3144 |
-
"loss":
|
| 3145 |
"step": 5160
|
| 3146 |
},
|
| 3147 |
{
|
| 3148 |
"epoch": 2.68,
|
| 3149 |
"learning_rate": 0.0008943089430894309,
|
| 3150 |
-
"loss":
|
| 3151 |
"step": 5170
|
| 3152 |
},
|
| 3153 |
{
|
| 3154 |
"epoch": 2.69,
|
| 3155 |
"learning_rate": 0.0008960387476215188,
|
| 3156 |
-
"loss":
|
| 3157 |
"step": 5180
|
| 3158 |
},
|
| 3159 |
{
|
| 3160 |
"epoch": 2.69,
|
| 3161 |
"learning_rate": 0.0008977685521536067,
|
| 3162 |
-
"loss":
|
| 3163 |
"step": 5190
|
| 3164 |
},
|
| 3165 |
{
|
| 3166 |
"epoch": 2.7,
|
| 3167 |
"learning_rate": 0.0008994983566856945,
|
| 3168 |
-
"loss": 1.
|
| 3169 |
"step": 5200
|
| 3170 |
},
|
| 3171 |
{
|
| 3172 |
"epoch": 2.7,
|
| 3173 |
"learning_rate": 0.0009012281612177824,
|
| 3174 |
-
"loss":
|
| 3175 |
"step": 5210
|
| 3176 |
},
|
| 3177 |
{
|
| 3178 |
"epoch": 2.71,
|
| 3179 |
"learning_rate": 0.0009029579657498703,
|
| 3180 |
-
"loss": 1.
|
| 3181 |
"step": 5220
|
| 3182 |
},
|
| 3183 |
{
|
| 3184 |
"epoch": 2.71,
|
| 3185 |
"learning_rate": 0.0009046877702819582,
|
| 3186 |
-
"loss": 1.
|
| 3187 |
"step": 5230
|
| 3188 |
},
|
| 3189 |
{
|
| 3190 |
"epoch": 2.72,
|
| 3191 |
"learning_rate": 0.000906417574814046,
|
| 3192 |
-
"loss": 1.
|
| 3193 |
"step": 5240
|
| 3194 |
},
|
| 3195 |
{
|
| 3196 |
"epoch": 2.72,
|
| 3197 |
"learning_rate": 0.0009081473793461339,
|
| 3198 |
-
"loss":
|
| 3199 |
"step": 5250
|
| 3200 |
},
|
| 3201 |
{
|
| 3202 |
"epoch": 2.73,
|
| 3203 |
"learning_rate": 0.0009098771838782218,
|
| 3204 |
-
"loss": 1.
|
| 3205 |
"step": 5260
|
| 3206 |
},
|
| 3207 |
{
|
| 3208 |
"epoch": 2.73,
|
| 3209 |
"learning_rate": 0.0009116069884103097,
|
| 3210 |
-
"loss": 1.
|
| 3211 |
"step": 5270
|
| 3212 |
},
|
| 3213 |
{
|
| 3214 |
"epoch": 2.74,
|
| 3215 |
"learning_rate": 0.0009133367929423976,
|
| 3216 |
-
"loss": 1.
|
| 3217 |
"step": 5280
|
| 3218 |
},
|
| 3219 |
{
|
| 3220 |
"epoch": 2.75,
|
| 3221 |
"learning_rate": 0.0009150665974744854,
|
| 3222 |
-
"loss": 1.
|
| 3223 |
"step": 5290
|
| 3224 |
},
|
| 3225 |
{
|
| 3226 |
"epoch": 2.75,
|
| 3227 |
"learning_rate": 0.0009167964020065733,
|
| 3228 |
-
"loss": 1.
|
| 3229 |
"step": 5300
|
| 3230 |
},
|
| 3231 |
{
|
| 3232 |
"epoch": 2.76,
|
| 3233 |
"learning_rate": 0.0009185262065386612,
|
| 3234 |
-
"loss": 1.
|
| 3235 |
"step": 5310
|
| 3236 |
},
|
| 3237 |
{
|
| 3238 |
"epoch": 2.76,
|
| 3239 |
"learning_rate": 0.0009202560110707491,
|
| 3240 |
-
"loss":
|
| 3241 |
"step": 5320
|
| 3242 |
},
|
| 3243 |
{
|
| 3244 |
"epoch": 2.77,
|
| 3245 |
"learning_rate": 0.0009219858156028368,
|
| 3246 |
-
"loss": 1.
|
| 3247 |
"step": 5330
|
| 3248 |
},
|
| 3249 |
{
|
| 3250 |
"epoch": 2.77,
|
| 3251 |
"learning_rate": 0.0009237156201349248,
|
| 3252 |
-
"loss": 1.
|
| 3253 |
"step": 5340
|
| 3254 |
},
|
| 3255 |
{
|
| 3256 |
"epoch": 2.78,
|
| 3257 |
"learning_rate": 0.0009254454246670126,
|
| 3258 |
-
"loss": 1.
|
| 3259 |
"step": 5350
|
| 3260 |
},
|
| 3261 |
{
|
| 3262 |
"epoch": 2.78,
|
| 3263 |
"learning_rate": 0.0009271752291991005,
|
| 3264 |
-
"loss": 1.
|
| 3265 |
"step": 5360
|
| 3266 |
},
|
| 3267 |
{
|
| 3268 |
"epoch": 2.79,
|
| 3269 |
"learning_rate": 0.0009289050337311884,
|
| 3270 |
-
"loss": 1.
|
| 3271 |
"step": 5370
|
| 3272 |
},
|
| 3273 |
{
|
| 3274 |
"epoch": 2.79,
|
| 3275 |
"learning_rate": 0.0009306348382632762,
|
| 3276 |
-
"loss": 1.
|
| 3277 |
"step": 5380
|
| 3278 |
},
|
| 3279 |
{
|
| 3280 |
"epoch": 2.8,
|
| 3281 |
"learning_rate": 0.0009323646427953641,
|
| 3282 |
-
"loss": 1.
|
| 3283 |
"step": 5390
|
| 3284 |
},
|
| 3285 |
{
|
| 3286 |
"epoch": 2.8,
|
| 3287 |
"learning_rate": 0.000934094447327452,
|
| 3288 |
-
"loss": 1.
|
| 3289 |
"step": 5400
|
| 3290 |
},
|
| 3291 |
{
|
| 3292 |
"epoch": 2.81,
|
| 3293 |
"learning_rate": 0.0009358242518595399,
|
| 3294 |
-
"loss": 1.
|
| 3295 |
"step": 5410
|
| 3296 |
},
|
| 3297 |
{
|
| 3298 |
"epoch": 2.81,
|
| 3299 |
"learning_rate": 0.0009375540563916278,
|
| 3300 |
-
"loss": 1.
|
| 3301 |
"step": 5420
|
| 3302 |
},
|
| 3303 |
{
|
| 3304 |
"epoch": 2.82,
|
| 3305 |
"learning_rate": 0.0009392838609237156,
|
| 3306 |
-
"loss": 1.
|
| 3307 |
"step": 5430
|
| 3308 |
},
|
| 3309 |
{
|
| 3310 |
"epoch": 2.82,
|
| 3311 |
"learning_rate": 0.0009410136654558035,
|
| 3312 |
-
"loss": 1.
|
| 3313 |
"step": 5440
|
| 3314 |
},
|
| 3315 |
{
|
| 3316 |
"epoch": 2.83,
|
| 3317 |
"learning_rate": 0.0009427434699878914,
|
| 3318 |
-
"loss":
|
| 3319 |
"step": 5450
|
| 3320 |
},
|
| 3321 |
{
|
| 3322 |
"epoch": 2.83,
|
| 3323 |
"learning_rate": 0.0009444732745199793,
|
| 3324 |
-
"loss": 1.
|
| 3325 |
"step": 5460
|
| 3326 |
},
|
| 3327 |
{
|
| 3328 |
"epoch": 2.84,
|
| 3329 |
"learning_rate": 0.0009462030790520671,
|
| 3330 |
-
"loss": 1.
|
| 3331 |
"step": 5470
|
| 3332 |
},
|
| 3333 |
{
|
| 3334 |
"epoch": 2.84,
|
| 3335 |
"learning_rate": 0.000947932883584155,
|
| 3336 |
-
"loss": 1.
|
| 3337 |
"step": 5480
|
| 3338 |
},
|
| 3339 |
{
|
| 3340 |
"epoch": 2.85,
|
| 3341 |
"learning_rate": 0.0009496626881162429,
|
| 3342 |
-
"loss": 1.
|
| 3343 |
"step": 5490
|
| 3344 |
},
|
| 3345 |
{
|
| 3346 |
"epoch": 2.85,
|
| 3347 |
"learning_rate": 0.0009513924926483307,
|
| 3348 |
-
"loss":
|
| 3349 |
"step": 5500
|
| 3350 |
},
|
| 3351 |
{
|
| 3352 |
"epoch": 2.86,
|
| 3353 |
"learning_rate": 0.0009531222971804187,
|
| 3354 |
-
"loss": 1.
|
| 3355 |
"step": 5510
|
| 3356 |
},
|
| 3357 |
{
|
| 3358 |
"epoch": 2.86,
|
| 3359 |
"learning_rate": 0.0009548521017125064,
|
| 3360 |
-
"loss": 1.
|
| 3361 |
"step": 5520
|
| 3362 |
},
|
| 3363 |
{
|
| 3364 |
"epoch": 2.87,
|
| 3365 |
"learning_rate": 0.0009565819062445944,
|
| 3366 |
-
"loss": 1.
|
| 3367 |
"step": 5530
|
| 3368 |
},
|
| 3369 |
{
|
| 3370 |
"epoch": 2.87,
|
| 3371 |
"learning_rate": 0.0009583117107766822,
|
| 3372 |
-
"loss": 1.
|
| 3373 |
"step": 5540
|
| 3374 |
},
|
| 3375 |
{
|
| 3376 |
"epoch": 2.88,
|
| 3377 |
"learning_rate": 0.0009600415153087702,
|
| 3378 |
-
"loss": 1.
|
| 3379 |
"step": 5550
|
| 3380 |
},
|
| 3381 |
{
|
| 3382 |
"epoch": 2.89,
|
| 3383 |
"learning_rate": 0.0009617713198408579,
|
| 3384 |
-
"loss": 1.
|
| 3385 |
"step": 5560
|
| 3386 |
},
|
| 3387 |
{
|
| 3388 |
"epoch": 2.89,
|
| 3389 |
"learning_rate": 0.0009635011243729459,
|
| 3390 |
-
"loss": 1.
|
| 3391 |
"step": 5570
|
| 3392 |
},
|
| 3393 |
{
|
| 3394 |
"epoch": 2.9,
|
| 3395 |
"learning_rate": 0.0009652309289050337,
|
| 3396 |
-
"loss": 1.
|
| 3397 |
"step": 5580
|
| 3398 |
},
|
| 3399 |
{
|
| 3400 |
"epoch": 2.9,
|
| 3401 |
"learning_rate": 0.0009669607334371217,
|
| 3402 |
-
"loss":
|
| 3403 |
"step": 5590
|
| 3404 |
},
|
| 3405 |
{
|
| 3406 |
"epoch": 2.91,
|
| 3407 |
"learning_rate": 0.0009686905379692095,
|
| 3408 |
-
"loss": 1.
|
| 3409 |
"step": 5600
|
| 3410 |
},
|
| 3411 |
{
|
| 3412 |
"epoch": 2.91,
|
| 3413 |
"learning_rate": 0.0009704203425012974,
|
| 3414 |
-
"loss":
|
| 3415 |
"step": 5610
|
| 3416 |
},
|
| 3417 |
{
|
| 3418 |
"epoch": 2.92,
|
| 3419 |
"learning_rate": 0.0009721501470333852,
|
| 3420 |
-
"loss": 1.
|
| 3421 |
"step": 5620
|
| 3422 |
},
|
| 3423 |
{
|
| 3424 |
"epoch": 2.92,
|
| 3425 |
"learning_rate": 0.0009738799515654732,
|
| 3426 |
-
"loss": 1.
|
| 3427 |
"step": 5630
|
| 3428 |
},
|
| 3429 |
{
|
| 3430 |
"epoch": 2.93,
|
| 3431 |
"learning_rate": 0.000975609756097561,
|
| 3432 |
-
"loss": 1.
|
| 3433 |
"step": 5640
|
| 3434 |
},
|
| 3435 |
{
|
| 3436 |
"epoch": 2.93,
|
| 3437 |
"learning_rate": 0.0009773395606296489,
|
| 3438 |
-
"loss": 1.
|
| 3439 |
"step": 5650
|
| 3440 |
},
|
| 3441 |
{
|
| 3442 |
"epoch": 2.94,
|
| 3443 |
"learning_rate": 0.0009790693651617367,
|
| 3444 |
-
"loss": 1.
|
| 3445 |
"step": 5660
|
| 3446 |
},
|
| 3447 |
{
|
| 3448 |
"epoch": 2.94,
|
| 3449 |
"learning_rate": 0.0009807991696938246,
|
| 3450 |
-
"loss": 1.
|
| 3451 |
"step": 5670
|
| 3452 |
},
|
| 3453 |
{
|
| 3454 |
"epoch": 2.95,
|
| 3455 |
"learning_rate": 0.0009825289742259125,
|
| 3456 |
-
"loss": 1.
|
| 3457 |
"step": 5680
|
| 3458 |
},
|
| 3459 |
{
|
| 3460 |
"epoch": 2.95,
|
| 3461 |
"learning_rate": 0.0009842587787580004,
|
| 3462 |
-
"loss": 1.
|
| 3463 |
"step": 5690
|
| 3464 |
},
|
| 3465 |
{
|
| 3466 |
"epoch": 2.96,
|
| 3467 |
"learning_rate": 0.0009859885832900882,
|
| 3468 |
-
"loss": 1.
|
| 3469 |
"step": 5700
|
| 3470 |
},
|
| 3471 |
{
|
| 3472 |
"epoch": 2.96,
|
| 3473 |
"learning_rate": 0.0009877183878221761,
|
| 3474 |
-
"loss": 1.
|
| 3475 |
"step": 5710
|
| 3476 |
},
|
| 3477 |
{
|
| 3478 |
"epoch": 2.97,
|
| 3479 |
"learning_rate": 0.000989448192354264,
|
| 3480 |
-
"loss": 1.
|
| 3481 |
"step": 5720
|
| 3482 |
},
|
| 3483 |
{
|
| 3484 |
"epoch": 2.97,
|
| 3485 |
"learning_rate": 0.0009911779968863519,
|
| 3486 |
-
"loss": 1.
|
| 3487 |
"step": 5730
|
| 3488 |
},
|
| 3489 |
{
|
| 3490 |
"epoch": 2.98,
|
| 3491 |
"learning_rate": 0.0009929078014184398,
|
| 3492 |
-
"loss": 1.
|
| 3493 |
"step": 5740
|
| 3494 |
},
|
| 3495 |
{
|
| 3496 |
"epoch": 2.98,
|
| 3497 |
"learning_rate": 0.0009946376059505276,
|
| 3498 |
-
"loss": 1.
|
| 3499 |
"step": 5750
|
| 3500 |
},
|
| 3501 |
{
|
| 3502 |
"epoch": 2.99,
|
| 3503 |
"learning_rate": 0.0009963674104826155,
|
| 3504 |
-
"loss": 1.
|
| 3505 |
"step": 5760
|
| 3506 |
},
|
| 3507 |
{
|
| 3508 |
"epoch": 2.99,
|
| 3509 |
"learning_rate": 0.0009980972150147034,
|
| 3510 |
-
"loss": 1.
|
| 3511 |
"step": 5770
|
| 3512 |
},
|
| 3513 |
{
|
| 3514 |
"epoch": 3.0,
|
| 3515 |
"learning_rate": 0.0009998270195467913,
|
| 3516 |
-
"loss": 1.
|
| 3517 |
"step": 5780
|
| 3518 |
},
|
| 3519 |
{
|
| 3520 |
"epoch": 3.0,
|
| 3521 |
"eval_accuracy": {
|
| 3522 |
-
"accuracy": 0.
|
| 3523 |
},
|
| 3524 |
"eval_f1": {
|
| 3525 |
-
"f1": 0.
|
| 3526 |
},
|
| 3527 |
-
"eval_loss": 1.
|
| 3528 |
"eval_precision": {
|
| 3529 |
-
"precision": 0.
|
| 3530 |
},
|
| 3531 |
"eval_recall": {
|
| 3532 |
-
"recall": 0.
|
| 3533 |
},
|
| 3534 |
-
"eval_runtime": 118.
|
| 3535 |
-
"eval_samples_per_second": 32.
|
| 3536 |
-
"eval_steps_per_second": 16.
|
| 3537 |
"step": 5781
|
| 3538 |
}
|
| 3539 |
],
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.9808588624000549,
|
| 3 |
"best_model_checkpoint": "swinv2-base-patch4-window12-192-22k-ConcreteClassifier-PVT\\checkpoint-1927",
|
| 4 |
"epoch": 3.0,
|
| 5 |
"eval_steps": 500,
|
|
|
|
| 11 |
{
|
| 12 |
"epoch": 0.01,
|
| 13 |
"learning_rate": 1.7298045320878742e-06,
|
| 14 |
+
"loss": 2.3066,
|
| 15 |
"step": 10
|
| 16 |
},
|
| 17 |
{
|
| 18 |
"epoch": 0.01,
|
| 19 |
"learning_rate": 3.4596090641757485e-06,
|
| 20 |
+
"loss": 2.131,
|
| 21 |
"step": 20
|
| 22 |
},
|
| 23 |
{
|
| 24 |
"epoch": 0.02,
|
| 25 |
"learning_rate": 5.1894135962636225e-06,
|
| 26 |
+
"loss": 2.0104,
|
| 27 |
"step": 30
|
| 28 |
},
|
| 29 |
{
|
| 30 |
"epoch": 0.02,
|
| 31 |
"learning_rate": 6.919218128351497e-06,
|
| 32 |
+
"loss": 1.8938,
|
| 33 |
"step": 40
|
| 34 |
},
|
| 35 |
{
|
| 36 |
"epoch": 0.03,
|
| 37 |
"learning_rate": 8.649022660439371e-06,
|
| 38 |
+
"loss": 1.7743,
|
| 39 |
"step": 50
|
| 40 |
},
|
| 41 |
{
|
| 42 |
"epoch": 0.03,
|
| 43 |
"learning_rate": 1.0378827192527245e-05,
|
| 44 |
+
"loss": 1.73,
|
| 45 |
"step": 60
|
| 46 |
},
|
| 47 |
{
|
| 48 |
"epoch": 0.04,
|
| 49 |
"learning_rate": 1.2108631724615119e-05,
|
| 50 |
+
"loss": 1.6552,
|
| 51 |
"step": 70
|
| 52 |
},
|
| 53 |
{
|
| 54 |
"epoch": 0.04,
|
| 55 |
"learning_rate": 1.3838436256702994e-05,
|
| 56 |
+
"loss": 1.4119,
|
| 57 |
"step": 80
|
| 58 |
},
|
| 59 |
{
|
| 60 |
"epoch": 0.05,
|
| 61 |
"learning_rate": 1.5568240788790867e-05,
|
| 62 |
+
"loss": 1.1716,
|
| 63 |
"step": 90
|
| 64 |
},
|
| 65 |
{
|
| 66 |
"epoch": 0.05,
|
| 67 |
"learning_rate": 1.7298045320878743e-05,
|
| 68 |
+
"loss": 1.2431,
|
| 69 |
"step": 100
|
| 70 |
},
|
| 71 |
{
|
| 72 |
"epoch": 0.06,
|
| 73 |
"learning_rate": 1.9027849852966615e-05,
|
| 74 |
+
"loss": 1.1115,
|
| 75 |
"step": 110
|
| 76 |
},
|
| 77 |
{
|
| 78 |
"epoch": 0.06,
|
| 79 |
"learning_rate": 2.075765438505449e-05,
|
| 80 |
+
"loss": 1.0365,
|
| 81 |
"step": 120
|
| 82 |
},
|
| 83 |
{
|
| 84 |
"epoch": 0.07,
|
| 85 |
"learning_rate": 2.2487458917142365e-05,
|
| 86 |
+
"loss": 1.2179,
|
| 87 |
"step": 130
|
| 88 |
},
|
| 89 |
{
|
| 90 |
"epoch": 0.07,
|
| 91 |
"learning_rate": 2.4217263449230237e-05,
|
| 92 |
+
"loss": 1.1192,
|
| 93 |
"step": 140
|
| 94 |
},
|
| 95 |
{
|
| 96 |
"epoch": 0.08,
|
| 97 |
"learning_rate": 2.5947067981318112e-05,
|
| 98 |
+
"loss": 0.8697,
|
| 99 |
"step": 150
|
| 100 |
},
|
| 101 |
{
|
| 102 |
"epoch": 0.08,
|
| 103 |
"learning_rate": 2.7676872513405988e-05,
|
| 104 |
+
"loss": 1.0286,
|
| 105 |
"step": 160
|
| 106 |
},
|
| 107 |
{
|
| 108 |
"epoch": 0.09,
|
| 109 |
"learning_rate": 2.940667704549386e-05,
|
| 110 |
+
"loss": 0.9186,
|
| 111 |
"step": 170
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"epoch": 0.09,
|
| 115 |
"learning_rate": 3.1136481577581735e-05,
|
| 116 |
+
"loss": 1.0909,
|
| 117 |
"step": 180
|
| 118 |
},
|
| 119 |
{
|
| 120 |
"epoch": 0.1,
|
| 121 |
"learning_rate": 3.286628610966961e-05,
|
| 122 |
+
"loss": 1.0519,
|
| 123 |
"step": 190
|
| 124 |
},
|
| 125 |
{
|
| 126 |
"epoch": 0.1,
|
| 127 |
"learning_rate": 3.4596090641757486e-05,
|
| 128 |
+
"loss": 1.1043,
|
| 129 |
"step": 200
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.11,
|
| 133 |
"learning_rate": 3.632589517384536e-05,
|
| 134 |
+
"loss": 0.8634,
|
| 135 |
"step": 210
|
| 136 |
},
|
| 137 |
{
|
| 138 |
"epoch": 0.11,
|
| 139 |
"learning_rate": 3.805569970593323e-05,
|
| 140 |
+
"loss": 0.86,
|
| 141 |
"step": 220
|
| 142 |
},
|
| 143 |
{
|
| 144 |
"epoch": 0.12,
|
| 145 |
"learning_rate": 3.97855042380211e-05,
|
| 146 |
+
"loss": 0.8325,
|
| 147 |
"step": 230
|
| 148 |
},
|
| 149 |
{
|
| 150 |
"epoch": 0.12,
|
| 151 |
"learning_rate": 4.151530877010898e-05,
|
| 152 |
+
"loss": 1.1083,
|
| 153 |
"step": 240
|
| 154 |
},
|
| 155 |
{
|
| 156 |
"epoch": 0.13,
|
| 157 |
"learning_rate": 4.324511330219685e-05,
|
| 158 |
+
"loss": 1.0759,
|
| 159 |
"step": 250
|
| 160 |
},
|
| 161 |
{
|
| 162 |
"epoch": 0.13,
|
| 163 |
"learning_rate": 4.497491783428473e-05,
|
| 164 |
+
"loss": 0.9178,
|
| 165 |
"step": 260
|
| 166 |
},
|
| 167 |
{
|
| 168 |
"epoch": 0.14,
|
| 169 |
"learning_rate": 4.67047223663726e-05,
|
| 170 |
+
"loss": 1.1132,
|
| 171 |
"step": 270
|
| 172 |
},
|
| 173 |
{
|
| 174 |
"epoch": 0.15,
|
| 175 |
"learning_rate": 4.8434526898460474e-05,
|
| 176 |
+
"loss": 1.0484,
|
| 177 |
"step": 280
|
| 178 |
},
|
| 179 |
{
|
| 180 |
"epoch": 0.15,
|
| 181 |
"learning_rate": 5.0164331430548346e-05,
|
| 182 |
+
"loss": 0.8629,
|
| 183 |
"step": 290
|
| 184 |
},
|
| 185 |
{
|
| 186 |
"epoch": 0.16,
|
| 187 |
"learning_rate": 5.1894135962636225e-05,
|
| 188 |
+
"loss": 0.6426,
|
| 189 |
"step": 300
|
| 190 |
},
|
| 191 |
{
|
| 192 |
"epoch": 0.16,
|
| 193 |
"learning_rate": 5.36239404947241e-05,
|
| 194 |
+
"loss": 1.0117,
|
| 195 |
"step": 310
|
| 196 |
},
|
| 197 |
{
|
| 198 |
"epoch": 0.17,
|
| 199 |
"learning_rate": 5.5353745026811976e-05,
|
| 200 |
+
"loss": 0.9769,
|
| 201 |
"step": 320
|
| 202 |
},
|
| 203 |
{
|
| 204 |
"epoch": 0.17,
|
| 205 |
"learning_rate": 5.708354955889985e-05,
|
| 206 |
+
"loss": 0.9459,
|
| 207 |
"step": 330
|
| 208 |
},
|
| 209 |
{
|
| 210 |
"epoch": 0.18,
|
| 211 |
"learning_rate": 5.881335409098772e-05,
|
| 212 |
+
"loss": 1.3208,
|
| 213 |
"step": 340
|
| 214 |
},
|
| 215 |
{
|
| 216 |
"epoch": 0.18,
|
| 217 |
"learning_rate": 6.054315862307559e-05,
|
| 218 |
+
"loss": 0.9631,
|
| 219 |
"step": 350
|
| 220 |
},
|
| 221 |
{
|
| 222 |
"epoch": 0.19,
|
| 223 |
"learning_rate": 6.227296315516347e-05,
|
| 224 |
+
"loss": 1.3533,
|
| 225 |
"step": 360
|
| 226 |
},
|
| 227 |
{
|
| 228 |
"epoch": 0.19,
|
| 229 |
"learning_rate": 6.400276768725135e-05,
|
| 230 |
+
"loss": 1.2228,
|
| 231 |
"step": 370
|
| 232 |
},
|
| 233 |
{
|
| 234 |
"epoch": 0.2,
|
| 235 |
"learning_rate": 6.573257221933921e-05,
|
| 236 |
+
"loss": 0.9195,
|
| 237 |
"step": 380
|
| 238 |
},
|
| 239 |
{
|
| 240 |
"epoch": 0.2,
|
| 241 |
"learning_rate": 6.746237675142709e-05,
|
| 242 |
+
"loss": 1.0466,
|
| 243 |
"step": 390
|
| 244 |
},
|
| 245 |
{
|
| 246 |
"epoch": 0.21,
|
| 247 |
"learning_rate": 6.919218128351497e-05,
|
| 248 |
+
"loss": 0.6537,
|
| 249 |
"step": 400
|
| 250 |
},
|
| 251 |
{
|
| 252 |
"epoch": 0.21,
|
| 253 |
"learning_rate": 7.092198581560285e-05,
|
| 254 |
+
"loss": 1.0766,
|
| 255 |
"step": 410
|
| 256 |
},
|
| 257 |
{
|
| 258 |
"epoch": 0.22,
|
| 259 |
"learning_rate": 7.265179034769071e-05,
|
| 260 |
+
"loss": 1.3196,
|
| 261 |
"step": 420
|
| 262 |
},
|
| 263 |
{
|
| 264 |
"epoch": 0.22,
|
| 265 |
"learning_rate": 7.438159487977858e-05,
|
| 266 |
+
"loss": 1.1902,
|
| 267 |
"step": 430
|
| 268 |
},
|
| 269 |
{
|
| 270 |
"epoch": 0.23,
|
| 271 |
"learning_rate": 7.611139941186646e-05,
|
| 272 |
+
"loss": 0.8251,
|
| 273 |
"step": 440
|
| 274 |
},
|
| 275 |
{
|
| 276 |
"epoch": 0.23,
|
| 277 |
"learning_rate": 7.784120394395434e-05,
|
| 278 |
+
"loss": 0.6196,
|
| 279 |
"step": 450
|
| 280 |
},
|
| 281 |
{
|
| 282 |
"epoch": 0.24,
|
| 283 |
"learning_rate": 7.95710084760422e-05,
|
| 284 |
+
"loss": 1.1941,
|
| 285 |
"step": 460
|
| 286 |
},
|
| 287 |
{
|
| 288 |
"epoch": 0.24,
|
| 289 |
"learning_rate": 8.130081300813008e-05,
|
| 290 |
+
"loss": 0.9509,
|
| 291 |
"step": 470
|
| 292 |
},
|
| 293 |
{
|
| 294 |
"epoch": 0.25,
|
| 295 |
"learning_rate": 8.303061754021796e-05,
|
| 296 |
+
"loss": 1.0669,
|
| 297 |
"step": 480
|
| 298 |
},
|
| 299 |
{
|
| 300 |
"epoch": 0.25,
|
| 301 |
"learning_rate": 8.476042207230583e-05,
|
| 302 |
+
"loss": 0.8493,
|
| 303 |
"step": 490
|
| 304 |
},
|
| 305 |
{
|
| 306 |
"epoch": 0.26,
|
| 307 |
"learning_rate": 8.64902266043937e-05,
|
| 308 |
+
"loss": 1.0381,
|
| 309 |
"step": 500
|
| 310 |
},
|
| 311 |
{
|
| 312 |
"epoch": 0.26,
|
| 313 |
"learning_rate": 8.822003113648158e-05,
|
| 314 |
+
"loss": 0.6853,
|
| 315 |
"step": 510
|
| 316 |
},
|
| 317 |
{
|
| 318 |
"epoch": 0.27,
|
| 319 |
"learning_rate": 8.994983566856946e-05,
|
| 320 |
+
"loss": 1.0025,
|
| 321 |
"step": 520
|
| 322 |
},
|
| 323 |
{
|
| 324 |
"epoch": 0.28,
|
| 325 |
"learning_rate": 9.167964020065733e-05,
|
| 326 |
+
"loss": 0.9758,
|
| 327 |
"step": 530
|
| 328 |
},
|
| 329 |
{
|
| 330 |
"epoch": 0.28,
|
| 331 |
"learning_rate": 9.34094447327452e-05,
|
| 332 |
+
"loss": 1.2263,
|
| 333 |
"step": 540
|
| 334 |
},
|
| 335 |
{
|
| 336 |
"epoch": 0.29,
|
| 337 |
"learning_rate": 9.513924926483307e-05,
|
| 338 |
+
"loss": 1.0245,
|
| 339 |
"step": 550
|
| 340 |
},
|
| 341 |
{
|
| 342 |
"epoch": 0.29,
|
| 343 |
"learning_rate": 9.686905379692095e-05,
|
| 344 |
+
"loss": 0.9136,
|
| 345 |
"step": 560
|
| 346 |
},
|
| 347 |
{
|
| 348 |
"epoch": 0.3,
|
| 349 |
"learning_rate": 9.859885832900881e-05,
|
| 350 |
+
"loss": 1.0586,
|
| 351 |
"step": 570
|
| 352 |
},
|
| 353 |
{
|
| 354 |
"epoch": 0.3,
|
| 355 |
"learning_rate": 0.00010032866286109669,
|
| 356 |
+
"loss": 1.029,
|
| 357 |
"step": 580
|
| 358 |
},
|
| 359 |
{
|
| 360 |
"epoch": 0.31,
|
| 361 |
"learning_rate": 0.00010205846739318457,
|
| 362 |
+
"loss": 0.9973,
|
| 363 |
"step": 590
|
| 364 |
},
|
| 365 |
{
|
| 366 |
"epoch": 0.31,
|
| 367 |
"learning_rate": 0.00010378827192527245,
|
| 368 |
+
"loss": 0.7652,
|
| 369 |
"step": 600
|
| 370 |
},
|
| 371 |
{
|
| 372 |
"epoch": 0.32,
|
| 373 |
"learning_rate": 0.00010551807645736032,
|
| 374 |
+
"loss": 1.2198,
|
| 375 |
"step": 610
|
| 376 |
},
|
| 377 |
{
|
| 378 |
"epoch": 0.32,
|
| 379 |
"learning_rate": 0.0001072478809894482,
|
| 380 |
+
"loss": 1.3727,
|
| 381 |
"step": 620
|
| 382 |
},
|
| 383 |
{
|
| 384 |
"epoch": 0.33,
|
| 385 |
"learning_rate": 0.00010897768552153607,
|
| 386 |
+
"loss": 0.9157,
|
| 387 |
"step": 630
|
| 388 |
},
|
| 389 |
{
|
| 390 |
"epoch": 0.33,
|
| 391 |
"learning_rate": 0.00011070749005362395,
|
| 392 |
+
"loss": 1.4063,
|
| 393 |
"step": 640
|
| 394 |
},
|
| 395 |
{
|
| 396 |
"epoch": 0.34,
|
| 397 |
"learning_rate": 0.00011243729458571182,
|
| 398 |
+
"loss": 0.9697,
|
| 399 |
"step": 650
|
| 400 |
},
|
| 401 |
{
|
| 402 |
"epoch": 0.34,
|
| 403 |
"learning_rate": 0.0001141670991177997,
|
| 404 |
+
"loss": 1.0678,
|
| 405 |
"step": 660
|
| 406 |
},
|
| 407 |
{
|
| 408 |
"epoch": 0.35,
|
| 409 |
"learning_rate": 0.00011589690364988756,
|
| 410 |
+
"loss": 1.1373,
|
| 411 |
"step": 670
|
| 412 |
},
|
| 413 |
{
|
| 414 |
"epoch": 0.35,
|
| 415 |
"learning_rate": 0.00011762670818197544,
|
| 416 |
+
"loss": 1.1127,
|
| 417 |
"step": 680
|
| 418 |
},
|
| 419 |
{
|
| 420 |
"epoch": 0.36,
|
| 421 |
"learning_rate": 0.0001193565127140633,
|
| 422 |
+
"loss": 0.8828,
|
| 423 |
"step": 690
|
| 424 |
},
|
| 425 |
{
|
| 426 |
"epoch": 0.36,
|
| 427 |
"learning_rate": 0.00012108631724615118,
|
| 428 |
+
"loss": 1.0497,
|
| 429 |
"step": 700
|
| 430 |
},
|
| 431 |
{
|
| 432 |
"epoch": 0.37,
|
| 433 |
"learning_rate": 0.00012281612177823906,
|
| 434 |
+
"loss": 1.0252,
|
| 435 |
"step": 710
|
| 436 |
},
|
| 437 |
{
|
| 438 |
"epoch": 0.37,
|
| 439 |
"learning_rate": 0.00012454592631032694,
|
| 440 |
+
"loss": 1.0175,
|
| 441 |
"step": 720
|
| 442 |
},
|
| 443 |
{
|
| 444 |
"epoch": 0.38,
|
| 445 |
"learning_rate": 0.00012627573084241482,
|
| 446 |
+
"loss": 1.1303,
|
| 447 |
"step": 730
|
| 448 |
},
|
| 449 |
{
|
| 450 |
"epoch": 0.38,
|
| 451 |
"learning_rate": 0.0001280055353745027,
|
| 452 |
+
"loss": 0.8751,
|
| 453 |
"step": 740
|
| 454 |
},
|
| 455 |
{
|
| 456 |
"epoch": 0.39,
|
| 457 |
"learning_rate": 0.00012973533990659055,
|
| 458 |
+
"loss": 0.9646,
|
| 459 |
"step": 750
|
| 460 |
},
|
| 461 |
{
|
| 462 |
"epoch": 0.39,
|
| 463 |
"learning_rate": 0.00013146514443867843,
|
| 464 |
+
"loss": 1.0412,
|
| 465 |
"step": 760
|
| 466 |
},
|
| 467 |
{
|
| 468 |
"epoch": 0.4,
|
| 469 |
"learning_rate": 0.0001331949489707663,
|
| 470 |
+
"loss": 1.3243,
|
| 471 |
"step": 770
|
| 472 |
},
|
| 473 |
{
|
| 474 |
"epoch": 0.4,
|
| 475 |
"learning_rate": 0.00013492475350285418,
|
| 476 |
+
"loss": 1.0182,
|
| 477 |
"step": 780
|
| 478 |
},
|
| 479 |
{
|
| 480 |
"epoch": 0.41,
|
| 481 |
"learning_rate": 0.00013665455803494206,
|
| 482 |
+
"loss": 1.3814,
|
| 483 |
"step": 790
|
| 484 |
},
|
| 485 |
{
|
| 486 |
"epoch": 0.42,
|
| 487 |
"learning_rate": 0.00013838436256702994,
|
| 488 |
+
"loss": 0.9131,
|
| 489 |
"step": 800
|
| 490 |
},
|
| 491 |
{
|
| 492 |
"epoch": 0.42,
|
| 493 |
"learning_rate": 0.00014011416709911782,
|
| 494 |
+
"loss": 1.0305,
|
| 495 |
"step": 810
|
| 496 |
},
|
| 497 |
{
|
| 498 |
"epoch": 0.43,
|
| 499 |
"learning_rate": 0.0001418439716312057,
|
| 500 |
+
"loss": 0.9296,
|
| 501 |
"step": 820
|
| 502 |
},
|
| 503 |
{
|
| 504 |
"epoch": 0.43,
|
| 505 |
"learning_rate": 0.00014357377616329355,
|
| 506 |
+
"loss": 1.1304,
|
| 507 |
"step": 830
|
| 508 |
},
|
| 509 |
{
|
| 510 |
"epoch": 0.44,
|
| 511 |
"learning_rate": 0.00014530358069538143,
|
| 512 |
+
"loss": 0.964,
|
| 513 |
"step": 840
|
| 514 |
},
|
| 515 |
{
|
| 516 |
"epoch": 0.44,
|
| 517 |
"learning_rate": 0.0001470333852274693,
|
| 518 |
+
"loss": 0.9223,
|
| 519 |
"step": 850
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.45,
|
| 523 |
"learning_rate": 0.00014876318975955716,
|
| 524 |
+
"loss": 0.8419,
|
| 525 |
"step": 860
|
| 526 |
},
|
| 527 |
{
|
| 528 |
"epoch": 0.45,
|
| 529 |
"learning_rate": 0.00015049299429164504,
|
| 530 |
+
"loss": 0.7896,
|
| 531 |
"step": 870
|
| 532 |
},
|
| 533 |
{
|
| 534 |
"epoch": 0.46,
|
| 535 |
"learning_rate": 0.00015222279882373292,
|
| 536 |
+
"loss": 1.2753,
|
| 537 |
"step": 880
|
| 538 |
},
|
| 539 |
{
|
| 540 |
"epoch": 0.46,
|
| 541 |
"learning_rate": 0.0001539526033558208,
|
| 542 |
+
"loss": 1.01,
|
| 543 |
"step": 890
|
| 544 |
},
|
| 545 |
{
|
| 546 |
"epoch": 0.47,
|
| 547 |
"learning_rate": 0.00015568240788790867,
|
| 548 |
+
"loss": 0.8729,
|
| 549 |
"step": 900
|
| 550 |
},
|
| 551 |
{
|
| 552 |
"epoch": 0.47,
|
| 553 |
"learning_rate": 0.00015741221241999653,
|
| 554 |
+
"loss": 1.2337,
|
| 555 |
"step": 910
|
| 556 |
},
|
| 557 |
{
|
| 558 |
"epoch": 0.48,
|
| 559 |
"learning_rate": 0.0001591420169520844,
|
| 560 |
+
"loss": 1.0598,
|
| 561 |
"step": 920
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.48,
|
| 565 |
"learning_rate": 0.00016087182148417228,
|
| 566 |
+
"loss": 0.9216,
|
| 567 |
"step": 930
|
| 568 |
},
|
| 569 |
{
|
| 570 |
"epoch": 0.49,
|
| 571 |
"learning_rate": 0.00016260162601626016,
|
| 572 |
+
"loss": 1.2614,
|
| 573 |
"step": 940
|
| 574 |
},
|
| 575 |
{
|
| 576 |
"epoch": 0.49,
|
| 577 |
"learning_rate": 0.00016433143054834804,
|
| 578 |
+
"loss": 0.9602,
|
| 579 |
"step": 950
|
| 580 |
},
|
| 581 |
{
|
| 582 |
"epoch": 0.5,
|
| 583 |
"learning_rate": 0.00016606123508043592,
|
| 584 |
+
"loss": 1.1126,
|
| 585 |
"step": 960
|
| 586 |
},
|
| 587 |
{
|
| 588 |
"epoch": 0.5,
|
| 589 |
"learning_rate": 0.0001677910396125238,
|
| 590 |
+
"loss": 0.7103,
|
| 591 |
"step": 970
|
| 592 |
},
|
| 593 |
{
|
| 594 |
"epoch": 0.51,
|
| 595 |
"learning_rate": 0.00016952084414461165,
|
| 596 |
+
"loss": 1.0462,
|
| 597 |
"step": 980
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.51,
|
| 601 |
"learning_rate": 0.00017125064867669953,
|
| 602 |
+
"loss": 0.9843,
|
| 603 |
"step": 990
|
| 604 |
},
|
| 605 |
{
|
| 606 |
"epoch": 0.52,
|
| 607 |
"learning_rate": 0.0001729804532087874,
|
| 608 |
+
"loss": 1.2419,
|
| 609 |
"step": 1000
|
| 610 |
},
|
| 611 |
{
|
| 612 |
"epoch": 0.52,
|
| 613 |
"learning_rate": 0.00017471025774087529,
|
| 614 |
+
"loss": 1.3823,
|
| 615 |
"step": 1010
|
| 616 |
},
|
| 617 |
{
|
| 618 |
"epoch": 0.53,
|
| 619 |
"learning_rate": 0.00017644006227296316,
|
| 620 |
+
"loss": 0.9726,
|
| 621 |
"step": 1020
|
| 622 |
},
|
| 623 |
{
|
| 624 |
"epoch": 0.53,
|
| 625 |
"learning_rate": 0.00017816986680505104,
|
| 626 |
+
"loss": 1.1238,
|
| 627 |
"step": 1030
|
| 628 |
},
|
| 629 |
{
|
| 630 |
"epoch": 0.54,
|
| 631 |
"learning_rate": 0.00017989967133713892,
|
| 632 |
+
"loss": 0.934,
|
| 633 |
"step": 1040
|
| 634 |
},
|
| 635 |
{
|
| 636 |
"epoch": 0.54,
|
| 637 |
"learning_rate": 0.0001816294758692268,
|
| 638 |
+
"loss": 1.5093,
|
| 639 |
"step": 1050
|
| 640 |
},
|
| 641 |
{
|
| 642 |
"epoch": 0.55,
|
| 643 |
"learning_rate": 0.00018335928040131465,
|
| 644 |
+
"loss": 1.0234,
|
| 645 |
"step": 1060
|
| 646 |
},
|
| 647 |
{
|
| 648 |
"epoch": 0.56,
|
| 649 |
"learning_rate": 0.00018508908493340253,
|
| 650 |
+
"loss": 0.7998,
|
| 651 |
"step": 1070
|
| 652 |
},
|
| 653 |
{
|
| 654 |
"epoch": 0.56,
|
| 655 |
"learning_rate": 0.0001868188894654904,
|
| 656 |
+
"loss": 1.0191,
|
| 657 |
"step": 1080
|
| 658 |
},
|
| 659 |
{
|
| 660 |
"epoch": 0.57,
|
| 661 |
"learning_rate": 0.0001885486939975783,
|
| 662 |
+
"loss": 1.1171,
|
| 663 |
"step": 1090
|
| 664 |
},
|
| 665 |
{
|
| 666 |
"epoch": 0.57,
|
| 667 |
"learning_rate": 0.00019027849852966614,
|
| 668 |
+
"loss": 1.1417,
|
| 669 |
"step": 1100
|
| 670 |
},
|
| 671 |
{
|
| 672 |
"epoch": 0.58,
|
| 673 |
"learning_rate": 0.00019200830306175402,
|
| 674 |
+
"loss": 0.9539,
|
| 675 |
"step": 1110
|
| 676 |
},
|
| 677 |
{
|
| 678 |
"epoch": 0.58,
|
| 679 |
"learning_rate": 0.0001937381075938419,
|
| 680 |
+
"loss": 1.0955,
|
| 681 |
"step": 1120
|
| 682 |
},
|
| 683 |
{
|
| 684 |
"epoch": 0.59,
|
| 685 |
"learning_rate": 0.00019546791212592978,
|
| 686 |
+
"loss": 1.3748,
|
| 687 |
"step": 1130
|
| 688 |
},
|
| 689 |
{
|
| 690 |
"epoch": 0.59,
|
| 691 |
"learning_rate": 0.00019719771665801763,
|
| 692 |
+
"loss": 1.1126,
|
| 693 |
"step": 1140
|
| 694 |
},
|
| 695 |
{
|
| 696 |
"epoch": 0.6,
|
| 697 |
"learning_rate": 0.0001989275211901055,
|
| 698 |
+
"loss": 1.0674,
|
| 699 |
"step": 1150
|
| 700 |
},
|
| 701 |
{
|
| 702 |
"epoch": 0.6,
|
| 703 |
"learning_rate": 0.00020065732572219339,
|
| 704 |
+
"loss": 1.1236,
|
| 705 |
"step": 1160
|
| 706 |
},
|
| 707 |
{
|
| 708 |
"epoch": 0.61,
|
| 709 |
"learning_rate": 0.00020238713025428126,
|
| 710 |
+
"loss": 1.0824,
|
| 711 |
"step": 1170
|
| 712 |
},
|
| 713 |
{
|
| 714 |
"epoch": 0.61,
|
| 715 |
"learning_rate": 0.00020411693478636914,
|
| 716 |
+
"loss": 1.1181,
|
| 717 |
"step": 1180
|
| 718 |
},
|
| 719 |
{
|
| 720 |
"epoch": 0.62,
|
| 721 |
"learning_rate": 0.00020584673931845702,
|
| 722 |
+
"loss": 1.0881,
|
| 723 |
"step": 1190
|
| 724 |
},
|
| 725 |
{
|
| 726 |
"epoch": 0.62,
|
| 727 |
"learning_rate": 0.0002075765438505449,
|
| 728 |
+
"loss": 1.1874,
|
| 729 |
"step": 1200
|
| 730 |
},
|
| 731 |
{
|
| 732 |
"epoch": 0.63,
|
| 733 |
"learning_rate": 0.00020930634838263278,
|
| 734 |
+
"loss": 1.0377,
|
| 735 |
"step": 1210
|
| 736 |
},
|
| 737 |
{
|
| 738 |
"epoch": 0.63,
|
| 739 |
"learning_rate": 0.00021103615291472063,
|
| 740 |
+
"loss": 0.8027,
|
| 741 |
"step": 1220
|
| 742 |
},
|
| 743 |
{
|
| 744 |
"epoch": 0.64,
|
| 745 |
"learning_rate": 0.0002127659574468085,
|
| 746 |
+
"loss": 0.8027,
|
| 747 |
"step": 1230
|
| 748 |
},
|
| 749 |
{
|
| 750 |
"epoch": 0.64,
|
| 751 |
"learning_rate": 0.0002144957619788964,
|
| 752 |
+
"loss": 1.2303,
|
| 753 |
"step": 1240
|
| 754 |
},
|
| 755 |
{
|
| 756 |
"epoch": 0.65,
|
| 757 |
"learning_rate": 0.00021622556651098427,
|
| 758 |
+
"loss": 1.2914,
|
| 759 |
"step": 1250
|
| 760 |
},
|
| 761 |
{
|
| 762 |
"epoch": 0.65,
|
| 763 |
"learning_rate": 0.00021795537104307214,
|
| 764 |
+
"loss": 1.2018,
|
| 765 |
"step": 1260
|
| 766 |
},
|
| 767 |
{
|
| 768 |
"epoch": 0.66,
|
| 769 |
"learning_rate": 0.00021968517557516002,
|
| 770 |
+
"loss": 0.994,
|
| 771 |
"step": 1270
|
| 772 |
},
|
| 773 |
{
|
| 774 |
"epoch": 0.66,
|
| 775 |
"learning_rate": 0.0002214149801072479,
|
| 776 |
+
"loss": 1.1917,
|
| 777 |
"step": 1280
|
| 778 |
},
|
| 779 |
{
|
| 780 |
"epoch": 0.67,
|
| 781 |
"learning_rate": 0.00022314478463933578,
|
| 782 |
+
"loss": 0.9846,
|
| 783 |
"step": 1290
|
| 784 |
},
|
| 785 |
{
|
| 786 |
"epoch": 0.67,
|
| 787 |
"learning_rate": 0.00022487458917142363,
|
| 788 |
+
"loss": 1.1561,
|
| 789 |
"step": 1300
|
| 790 |
},
|
| 791 |
{
|
| 792 |
"epoch": 0.68,
|
| 793 |
"learning_rate": 0.0002266043937035115,
|
| 794 |
+
"loss": 1.3345,
|
| 795 |
"step": 1310
|
| 796 |
},
|
| 797 |
{
|
| 798 |
"epoch": 0.69,
|
| 799 |
"learning_rate": 0.0002283341982355994,
|
| 800 |
+
"loss": 1.1663,
|
| 801 |
"step": 1320
|
| 802 |
},
|
| 803 |
{
|
| 804 |
"epoch": 0.69,
|
| 805 |
"learning_rate": 0.00023006400276768727,
|
| 806 |
+
"loss": 1.0486,
|
| 807 |
"step": 1330
|
| 808 |
},
|
| 809 |
{
|
| 810 |
"epoch": 0.7,
|
| 811 |
"learning_rate": 0.00023179380729977512,
|
| 812 |
+
"loss": 1.2478,
|
| 813 |
"step": 1340
|
| 814 |
},
|
| 815 |
{
|
| 816 |
"epoch": 0.7,
|
| 817 |
"learning_rate": 0.000233523611831863,
|
| 818 |
+
"loss": 1.1334,
|
| 819 |
"step": 1350
|
| 820 |
},
|
| 821 |
{
|
| 822 |
"epoch": 0.71,
|
| 823 |
"learning_rate": 0.00023525341636395088,
|
| 824 |
+
"loss": 1.1248,
|
| 825 |
"step": 1360
|
| 826 |
},
|
| 827 |
{
|
| 828 |
"epoch": 0.71,
|
| 829 |
"learning_rate": 0.00023698322089603876,
|
| 830 |
+
"loss": 1.2183,
|
| 831 |
"step": 1370
|
| 832 |
},
|
| 833 |
{
|
| 834 |
"epoch": 0.72,
|
| 835 |
"learning_rate": 0.0002387130254281266,
|
| 836 |
+
"loss": 1.4618,
|
| 837 |
"step": 1380
|
| 838 |
},
|
| 839 |
{
|
| 840 |
"epoch": 0.72,
|
| 841 |
"learning_rate": 0.00024044282996021449,
|
| 842 |
+
"loss": 1.5796,
|
| 843 |
"step": 1390
|
| 844 |
},
|
| 845 |
{
|
| 846 |
"epoch": 0.73,
|
| 847 |
"learning_rate": 0.00024217263449230237,
|
| 848 |
+
"loss": 1.4877,
|
| 849 |
"step": 1400
|
| 850 |
},
|
| 851 |
{
|
| 852 |
"epoch": 0.73,
|
| 853 |
"learning_rate": 0.00024390243902439024,
|
| 854 |
+
"loss": 0.9612,
|
| 855 |
"step": 1410
|
| 856 |
},
|
| 857 |
{
|
| 858 |
"epoch": 0.74,
|
| 859 |
"learning_rate": 0.0002456322435564781,
|
| 860 |
+
"loss": 1.3634,
|
| 861 |
"step": 1420
|
| 862 |
},
|
| 863 |
{
|
| 864 |
"epoch": 0.74,
|
| 865 |
"learning_rate": 0.000247362048088566,
|
| 866 |
+
"loss": 1.1377,
|
| 867 |
"step": 1430
|
| 868 |
},
|
| 869 |
{
|
| 870 |
"epoch": 0.75,
|
| 871 |
"learning_rate": 0.0002490918526206539,
|
| 872 |
+
"loss": 1.1365,
|
| 873 |
"step": 1440
|
| 874 |
},
|
| 875 |
{
|
| 876 |
"epoch": 0.75,
|
| 877 |
"learning_rate": 0.00025082165715274176,
|
| 878 |
+
"loss": 1.2433,
|
| 879 |
"step": 1450
|
| 880 |
},
|
| 881 |
{
|
| 882 |
"epoch": 0.76,
|
| 883 |
"learning_rate": 0.00025255146168482964,
|
| 884 |
+
"loss": 1.2164,
|
| 885 |
"step": 1460
|
| 886 |
},
|
| 887 |
{
|
| 888 |
"epoch": 0.76,
|
| 889 |
"learning_rate": 0.0002542812662169175,
|
| 890 |
+
"loss": 1.0668,
|
| 891 |
"step": 1470
|
| 892 |
},
|
| 893 |
{
|
| 894 |
"epoch": 0.77,
|
| 895 |
"learning_rate": 0.0002560110707490054,
|
| 896 |
+
"loss": 1.174,
|
| 897 |
"step": 1480
|
| 898 |
},
|
| 899 |
{
|
| 900 |
"epoch": 0.77,
|
| 901 |
"learning_rate": 0.0002577408752810932,
|
| 902 |
+
"loss": 1.1307,
|
| 903 |
"step": 1490
|
| 904 |
},
|
| 905 |
{
|
| 906 |
"epoch": 0.78,
|
| 907 |
"learning_rate": 0.0002594706798131811,
|
| 908 |
+
"loss": 1.2406,
|
| 909 |
"step": 1500
|
| 910 |
},
|
| 911 |
{
|
| 912 |
"epoch": 0.78,
|
| 913 |
"learning_rate": 0.000261200484345269,
|
| 914 |
+
"loss": 1.2542,
|
| 915 |
"step": 1510
|
| 916 |
},
|
| 917 |
{
|
| 918 |
"epoch": 0.79,
|
| 919 |
"learning_rate": 0.00026293028887735685,
|
| 920 |
+
"loss": 0.9126,
|
| 921 |
"step": 1520
|
| 922 |
},
|
| 923 |
{
|
| 924 |
"epoch": 0.79,
|
| 925 |
"learning_rate": 0.00026466009340944473,
|
| 926 |
+
"loss": 1.1105,
|
| 927 |
"step": 1530
|
| 928 |
},
|
| 929 |
{
|
| 930 |
"epoch": 0.8,
|
| 931 |
"learning_rate": 0.0002663898979415326,
|
| 932 |
+
"loss": 0.8491,
|
| 933 |
"step": 1540
|
| 934 |
},
|
| 935 |
{
|
| 936 |
"epoch": 0.8,
|
| 937 |
"learning_rate": 0.0002681197024736205,
|
| 938 |
+
"loss": 1.4498,
|
| 939 |
"step": 1550
|
| 940 |
},
|
| 941 |
{
|
| 942 |
"epoch": 0.81,
|
| 943 |
"learning_rate": 0.00026984950700570837,
|
| 944 |
+
"loss": 1.3299,
|
| 945 |
"step": 1560
|
| 946 |
},
|
| 947 |
{
|
| 948 |
"epoch": 0.81,
|
| 949 |
"learning_rate": 0.00027157931153779625,
|
| 950 |
+
"loss": 1.2222,
|
| 951 |
"step": 1570
|
| 952 |
},
|
| 953 |
{
|
| 954 |
"epoch": 0.82,
|
| 955 |
"learning_rate": 0.0002733091160698841,
|
| 956 |
+
"loss": 0.9812,
|
| 957 |
"step": 1580
|
| 958 |
},
|
| 959 |
{
|
| 960 |
"epoch": 0.83,
|
| 961 |
"learning_rate": 0.000275038920601972,
|
| 962 |
+
"loss": 1.2824,
|
| 963 |
"step": 1590
|
| 964 |
},
|
| 965 |
{
|
| 966 |
"epoch": 0.83,
|
| 967 |
"learning_rate": 0.0002767687251340599,
|
| 968 |
+
"loss": 1.1105,
|
| 969 |
"step": 1600
|
| 970 |
},
|
| 971 |
{
|
| 972 |
"epoch": 0.84,
|
| 973 |
"learning_rate": 0.00027849852966614776,
|
| 974 |
+
"loss": 1.0978,
|
| 975 |
"step": 1610
|
| 976 |
},
|
| 977 |
{
|
| 978 |
"epoch": 0.84,
|
| 979 |
"learning_rate": 0.00028022833419823564,
|
| 980 |
+
"loss": 1.1159,
|
| 981 |
"step": 1620
|
| 982 |
},
|
| 983 |
{
|
| 984 |
"epoch": 0.85,
|
| 985 |
"learning_rate": 0.0002819581387303235,
|
| 986 |
+
"loss": 1.1441,
|
| 987 |
"step": 1630
|
| 988 |
},
|
| 989 |
{
|
| 990 |
"epoch": 0.85,
|
| 991 |
"learning_rate": 0.0002836879432624114,
|
| 992 |
+
"loss": 1.2566,
|
| 993 |
"step": 1640
|
| 994 |
},
|
| 995 |
{
|
| 996 |
"epoch": 0.86,
|
| 997 |
"learning_rate": 0.0002854177477944992,
|
| 998 |
+
"loss": 1.1327,
|
| 999 |
"step": 1650
|
| 1000 |
},
|
| 1001 |
{
|
| 1002 |
"epoch": 0.86,
|
| 1003 |
"learning_rate": 0.0002871475523265871,
|
| 1004 |
+
"loss": 1.232,
|
| 1005 |
"step": 1660
|
| 1006 |
},
|
| 1007 |
{
|
| 1008 |
"epoch": 0.87,
|
| 1009 |
"learning_rate": 0.000288877356858675,
|
| 1010 |
+
"loss": 1.0974,
|
| 1011 |
"step": 1670
|
| 1012 |
},
|
| 1013 |
{
|
| 1014 |
"epoch": 0.87,
|
| 1015 |
"learning_rate": 0.00029060716139076286,
|
| 1016 |
+
"loss": 1.0083,
|
| 1017 |
"step": 1680
|
| 1018 |
},
|
| 1019 |
{
|
| 1020 |
"epoch": 0.88,
|
| 1021 |
"learning_rate": 0.00029233696592285074,
|
| 1022 |
+
"loss": 1.5513,
|
| 1023 |
"step": 1690
|
| 1024 |
},
|
| 1025 |
{
|
| 1026 |
"epoch": 0.88,
|
| 1027 |
"learning_rate": 0.0002940667704549386,
|
| 1028 |
+
"loss": 1.3731,
|
| 1029 |
"step": 1700
|
| 1030 |
},
|
| 1031 |
{
|
| 1032 |
"epoch": 0.89,
|
| 1033 |
"learning_rate": 0.00029579657498702644,
|
| 1034 |
+
"loss": 1.1877,
|
| 1035 |
"step": 1710
|
| 1036 |
},
|
| 1037 |
{
|
| 1038 |
"epoch": 0.89,
|
| 1039 |
"learning_rate": 0.0002975263795191143,
|
| 1040 |
+
"loss": 1.199,
|
| 1041 |
"step": 1720
|
| 1042 |
},
|
| 1043 |
{
|
| 1044 |
"epoch": 0.9,
|
| 1045 |
"learning_rate": 0.0002992561840512022,
|
| 1046 |
+
"loss": 1.4768,
|
| 1047 |
"step": 1730
|
| 1048 |
},
|
| 1049 |
{
|
| 1050 |
"epoch": 0.9,
|
| 1051 |
"learning_rate": 0.0003009859885832901,
|
| 1052 |
+
"loss": 1.7352,
|
| 1053 |
"step": 1740
|
| 1054 |
},
|
| 1055 |
{
|
| 1056 |
"epoch": 0.91,
|
| 1057 |
"learning_rate": 0.00030271579311537796,
|
| 1058 |
+
"loss": 1.3652,
|
| 1059 |
"step": 1750
|
| 1060 |
},
|
| 1061 |
{
|
| 1062 |
"epoch": 0.91,
|
| 1063 |
"learning_rate": 0.00030444559764746583,
|
| 1064 |
+
"loss": 1.1335,
|
| 1065 |
"step": 1760
|
| 1066 |
},
|
| 1067 |
{
|
| 1068 |
"epoch": 0.92,
|
| 1069 |
"learning_rate": 0.0003061754021795537,
|
| 1070 |
+
"loss": 1.2259,
|
| 1071 |
"step": 1770
|
| 1072 |
},
|
| 1073 |
{
|
| 1074 |
"epoch": 0.92,
|
| 1075 |
"learning_rate": 0.0003079052067116416,
|
| 1076 |
+
"loss": 1.3024,
|
| 1077 |
"step": 1780
|
| 1078 |
},
|
| 1079 |
{
|
| 1080 |
"epoch": 0.93,
|
| 1081 |
"learning_rate": 0.00030963501124372947,
|
| 1082 |
+
"loss": 1.4433,
|
| 1083 |
"step": 1790
|
| 1084 |
},
|
| 1085 |
{
|
| 1086 |
"epoch": 0.93,
|
| 1087 |
"learning_rate": 0.00031136481577581735,
|
| 1088 |
+
"loss": 1.1668,
|
| 1089 |
"step": 1800
|
| 1090 |
},
|
| 1091 |
{
|
| 1092 |
"epoch": 0.94,
|
| 1093 |
"learning_rate": 0.0003130946203079052,
|
| 1094 |
+
"loss": 1.3298,
|
| 1095 |
"step": 1810
|
| 1096 |
},
|
| 1097 |
{
|
| 1098 |
"epoch": 0.94,
|
| 1099 |
"learning_rate": 0.00031482442483999305,
|
| 1100 |
+
"loss": 1.2601,
|
| 1101 |
"step": 1820
|
| 1102 |
},
|
| 1103 |
{
|
| 1104 |
"epoch": 0.95,
|
| 1105 |
"learning_rate": 0.00031655422937208093,
|
| 1106 |
+
"loss": 1.1204,
|
| 1107 |
"step": 1830
|
| 1108 |
},
|
| 1109 |
{
|
| 1110 |
"epoch": 0.95,
|
| 1111 |
"learning_rate": 0.0003182840339041688,
|
| 1112 |
+
"loss": 1.2142,
|
| 1113 |
"step": 1840
|
| 1114 |
},
|
| 1115 |
{
|
| 1116 |
"epoch": 0.96,
|
| 1117 |
"learning_rate": 0.0003200138384362567,
|
| 1118 |
+
"loss": 0.93,
|
| 1119 |
"step": 1850
|
| 1120 |
},
|
| 1121 |
{
|
| 1122 |
"epoch": 0.97,
|
| 1123 |
"learning_rate": 0.00032174364296834457,
|
| 1124 |
+
"loss": 1.4335,
|
| 1125 |
"step": 1860
|
| 1126 |
},
|
| 1127 |
{
|
| 1128 |
"epoch": 0.97,
|
| 1129 |
"learning_rate": 0.00032347344750043245,
|
| 1130 |
+
"loss": 1.3081,
|
| 1131 |
"step": 1870
|
| 1132 |
},
|
| 1133 |
{
|
| 1134 |
"epoch": 0.98,
|
| 1135 |
"learning_rate": 0.0003252032520325203,
|
| 1136 |
+
"loss": 1.2692,
|
| 1137 |
"step": 1880
|
| 1138 |
},
|
| 1139 |
{
|
| 1140 |
"epoch": 0.98,
|
| 1141 |
"learning_rate": 0.0003269330565646082,
|
| 1142 |
+
"loss": 1.2172,
|
| 1143 |
"step": 1890
|
| 1144 |
},
|
| 1145 |
{
|
| 1146 |
"epoch": 0.99,
|
| 1147 |
"learning_rate": 0.0003286628610966961,
|
| 1148 |
+
"loss": 1.2254,
|
| 1149 |
"step": 1900
|
| 1150 |
},
|
| 1151 |
{
|
| 1152 |
"epoch": 0.99,
|
| 1153 |
"learning_rate": 0.00033039266562878396,
|
| 1154 |
+
"loss": 1.1448,
|
| 1155 |
"step": 1910
|
| 1156 |
},
|
| 1157 |
{
|
| 1158 |
"epoch": 1.0,
|
| 1159 |
"learning_rate": 0.00033212247016087184,
|
| 1160 |
+
"loss": 1.2473,
|
| 1161 |
"step": 1920
|
| 1162 |
},
|
| 1163 |
{
|
| 1164 |
"epoch": 1.0,
|
| 1165 |
"eval_accuracy": {
|
| 1166 |
+
"accuracy": 0.5909208819714656
|
| 1167 |
},
|
| 1168 |
"eval_f1": {
|
| 1169 |
+
"f1": 0.5910435702573883
|
| 1170 |
},
|
| 1171 |
+
"eval_loss": 0.9808588624000549,
|
| 1172 |
"eval_precision": {
|
| 1173 |
+
"precision": 0.6661654194259725
|
| 1174 |
},
|
| 1175 |
"eval_recall": {
|
| 1176 |
+
"recall": 0.6016064778916844
|
| 1177 |
},
|
| 1178 |
+
"eval_runtime": 243.6104,
|
| 1179 |
+
"eval_samples_per_second": 15.824,
|
| 1180 |
+
"eval_steps_per_second": 7.914,
|
| 1181 |
"step": 1927
|
| 1182 |
},
|
| 1183 |
{
|
| 1184 |
"epoch": 1.0,
|
| 1185 |
"learning_rate": 0.0003338522746929597,
|
| 1186 |
+
"loss": 1.1687,
|
| 1187 |
"step": 1930
|
| 1188 |
},
|
| 1189 |
{
|
| 1190 |
"epoch": 1.01,
|
| 1191 |
"learning_rate": 0.0003355820792250476,
|
| 1192 |
+
"loss": 1.4294,
|
| 1193 |
"step": 1940
|
| 1194 |
},
|
| 1195 |
{
|
| 1196 |
"epoch": 1.01,
|
| 1197 |
"learning_rate": 0.0003373118837571355,
|
| 1198 |
+
"loss": 1.2866,
|
| 1199 |
"step": 1950
|
| 1200 |
},
|
| 1201 |
{
|
| 1202 |
"epoch": 1.02,
|
| 1203 |
"learning_rate": 0.0003390416882892233,
|
| 1204 |
+
"loss": 1.0858,
|
| 1205 |
"step": 1960
|
| 1206 |
},
|
| 1207 |
{
|
| 1208 |
"epoch": 1.02,
|
| 1209 |
"learning_rate": 0.0003407714928213112,
|
| 1210 |
+
"loss": 1.1465,
|
| 1211 |
"step": 1970
|
| 1212 |
},
|
| 1213 |
{
|
| 1214 |
"epoch": 1.03,
|
| 1215 |
"learning_rate": 0.00034250129735339906,
|
| 1216 |
+
"loss": 1.2876,
|
| 1217 |
"step": 1980
|
| 1218 |
},
|
| 1219 |
{
|
| 1220 |
"epoch": 1.03,
|
| 1221 |
"learning_rate": 0.00034423110188548694,
|
| 1222 |
+
"loss": 1.0266,
|
| 1223 |
"step": 1990
|
| 1224 |
},
|
| 1225 |
{
|
| 1226 |
"epoch": 1.04,
|
| 1227 |
"learning_rate": 0.0003459609064175748,
|
| 1228 |
+
"loss": 1.1494,
|
| 1229 |
"step": 2000
|
| 1230 |
},
|
| 1231 |
{
|
| 1232 |
"epoch": 1.04,
|
| 1233 |
"learning_rate": 0.0003476907109496627,
|
| 1234 |
+
"loss": 1.5705,
|
| 1235 |
"step": 2010
|
| 1236 |
},
|
| 1237 |
{
|
| 1238 |
"epoch": 1.05,
|
| 1239 |
"learning_rate": 0.00034942051548175057,
|
| 1240 |
+
"loss": 0.9916,
|
| 1241 |
"step": 2020
|
| 1242 |
},
|
| 1243 |
{
|
| 1244 |
"epoch": 1.05,
|
| 1245 |
"learning_rate": 0.00035115032001383845,
|
| 1246 |
+
"loss": 1.2183,
|
| 1247 |
"step": 2030
|
| 1248 |
},
|
| 1249 |
{
|
| 1250 |
"epoch": 1.06,
|
| 1251 |
"learning_rate": 0.00035288012454592633,
|
| 1252 |
+
"loss": 1.1186,
|
| 1253 |
"step": 2040
|
| 1254 |
},
|
| 1255 |
{
|
| 1256 |
"epoch": 1.06,
|
| 1257 |
"learning_rate": 0.0003546099290780142,
|
| 1258 |
+
"loss": 1.1386,
|
| 1259 |
"step": 2050
|
| 1260 |
},
|
| 1261 |
{
|
| 1262 |
"epoch": 1.07,
|
| 1263 |
"learning_rate": 0.0003563397336101021,
|
| 1264 |
+
"loss": 1.413,
|
| 1265 |
"step": 2060
|
| 1266 |
},
|
| 1267 |
{
|
| 1268 |
"epoch": 1.07,
|
| 1269 |
"learning_rate": 0.00035806953814218997,
|
| 1270 |
+
"loss": 1.4434,
|
| 1271 |
"step": 2070
|
| 1272 |
},
|
| 1273 |
{
|
| 1274 |
"epoch": 1.08,
|
| 1275 |
"learning_rate": 0.00035979934267427784,
|
| 1276 |
+
"loss": 1.3144,
|
| 1277 |
"step": 2080
|
| 1278 |
},
|
| 1279 |
{
|
| 1280 |
"epoch": 1.08,
|
| 1281 |
"learning_rate": 0.0003615291472063657,
|
| 1282 |
+
"loss": 1.4108,
|
| 1283 |
"step": 2090
|
| 1284 |
},
|
| 1285 |
{
|
| 1286 |
"epoch": 1.09,
|
| 1287 |
"learning_rate": 0.0003632589517384536,
|
| 1288 |
+
"loss": 1.2305,
|
| 1289 |
"step": 2100
|
| 1290 |
},
|
| 1291 |
{
|
| 1292 |
"epoch": 1.09,
|
| 1293 |
"learning_rate": 0.0003649887562705415,
|
| 1294 |
+
"loss": 1.2335,
|
| 1295 |
"step": 2110
|
| 1296 |
},
|
| 1297 |
{
|
| 1298 |
"epoch": 1.1,
|
| 1299 |
"learning_rate": 0.0003667185608026293,
|
| 1300 |
+
"loss": 1.3617,
|
| 1301 |
"step": 2120
|
| 1302 |
},
|
| 1303 |
{
|
| 1304 |
"epoch": 1.11,
|
| 1305 |
"learning_rate": 0.0003684483653347172,
|
| 1306 |
+
"loss": 1.1457,
|
| 1307 |
"step": 2130
|
| 1308 |
},
|
| 1309 |
{
|
| 1310 |
"epoch": 1.11,
|
| 1311 |
"learning_rate": 0.00037017816986680506,
|
| 1312 |
+
"loss": 1.0253,
|
| 1313 |
"step": 2140
|
| 1314 |
},
|
| 1315 |
{
|
| 1316 |
"epoch": 1.12,
|
| 1317 |
"learning_rate": 0.00037190797439889294,
|
| 1318 |
+
"loss": 1.1486,
|
| 1319 |
"step": 2150
|
| 1320 |
},
|
| 1321 |
{
|
| 1322 |
"epoch": 1.12,
|
| 1323 |
"learning_rate": 0.0003736377789309808,
|
| 1324 |
+
"loss": 1.2603,
|
| 1325 |
"step": 2160
|
| 1326 |
},
|
| 1327 |
{
|
| 1328 |
"epoch": 1.13,
|
| 1329 |
"learning_rate": 0.0003753675834630687,
|
| 1330 |
+
"loss": 1.25,
|
| 1331 |
"step": 2170
|
| 1332 |
},
|
| 1333 |
{
|
| 1334 |
"epoch": 1.13,
|
| 1335 |
"learning_rate": 0.0003770973879951566,
|
| 1336 |
+
"loss": 1.5531,
|
| 1337 |
"step": 2180
|
| 1338 |
},
|
| 1339 |
{
|
| 1340 |
"epoch": 1.14,
|
| 1341 |
"learning_rate": 0.0003788271925272444,
|
| 1342 |
+
"loss": 1.3442,
|
| 1343 |
"step": 2190
|
| 1344 |
},
|
| 1345 |
{
|
| 1346 |
"epoch": 1.14,
|
| 1347 |
"learning_rate": 0.0003805569970593323,
|
| 1348 |
+
"loss": 1.1317,
|
| 1349 |
"step": 2200
|
| 1350 |
},
|
| 1351 |
{
|
| 1352 |
"epoch": 1.15,
|
| 1353 |
"learning_rate": 0.00038228680159142016,
|
| 1354 |
+
"loss": 1.3271,
|
| 1355 |
"step": 2210
|
| 1356 |
},
|
| 1357 |
{
|
| 1358 |
"epoch": 1.15,
|
| 1359 |
"learning_rate": 0.00038401660612350804,
|
| 1360 |
+
"loss": 1.3165,
|
| 1361 |
"step": 2220
|
| 1362 |
},
|
| 1363 |
{
|
| 1364 |
"epoch": 1.16,
|
| 1365 |
"learning_rate": 0.0003857464106555959,
|
| 1366 |
+
"loss": 1.2377,
|
| 1367 |
"step": 2230
|
| 1368 |
},
|
| 1369 |
{
|
| 1370 |
"epoch": 1.16,
|
| 1371 |
"learning_rate": 0.0003874762151876838,
|
| 1372 |
+
"loss": 1.2677,
|
| 1373 |
"step": 2240
|
| 1374 |
},
|
| 1375 |
{
|
| 1376 |
"epoch": 1.17,
|
| 1377 |
"learning_rate": 0.0003892060197197717,
|
| 1378 |
+
"loss": 1.2756,
|
| 1379 |
"step": 2250
|
| 1380 |
},
|
| 1381 |
{
|
| 1382 |
"epoch": 1.17,
|
| 1383 |
"learning_rate": 0.00039093582425185955,
|
| 1384 |
+
"loss": 1.0193,
|
| 1385 |
"step": 2260
|
| 1386 |
},
|
| 1387 |
{
|
| 1388 |
"epoch": 1.18,
|
| 1389 |
"learning_rate": 0.00039266562878394743,
|
| 1390 |
+
"loss": 1.239,
|
| 1391 |
"step": 2270
|
| 1392 |
},
|
| 1393 |
{
|
| 1394 |
"epoch": 1.18,
|
| 1395 |
"learning_rate": 0.00039439543331603526,
|
| 1396 |
+
"loss": 1.6044,
|
| 1397 |
"step": 2280
|
| 1398 |
},
|
| 1399 |
{
|
| 1400 |
"epoch": 1.19,
|
| 1401 |
"learning_rate": 0.00039612523784812313,
|
| 1402 |
+
"loss": 1.2396,
|
| 1403 |
"step": 2290
|
| 1404 |
},
|
| 1405 |
{
|
| 1406 |
"epoch": 1.19,
|
| 1407 |
"learning_rate": 0.000397855042380211,
|
| 1408 |
+
"loss": 1.0345,
|
| 1409 |
"step": 2300
|
| 1410 |
},
|
| 1411 |
{
|
| 1412 |
"epoch": 1.2,
|
| 1413 |
"learning_rate": 0.0003995848469122989,
|
| 1414 |
+
"loss": 1.3125,
|
| 1415 |
"step": 2310
|
| 1416 |
},
|
| 1417 |
{
|
| 1418 |
"epoch": 1.2,
|
| 1419 |
"learning_rate": 0.00040131465144438677,
|
| 1420 |
+
"loss": 1.4382,
|
| 1421 |
"step": 2320
|
| 1422 |
},
|
| 1423 |
{
|
| 1424 |
"epoch": 1.21,
|
| 1425 |
"learning_rate": 0.00040304445597647465,
|
| 1426 |
+
"loss": 1.1232,
|
| 1427 |
"step": 2330
|
| 1428 |
},
|
| 1429 |
{
|
| 1430 |
"epoch": 1.21,
|
| 1431 |
"learning_rate": 0.00040477426050856253,
|
| 1432 |
+
"loss": 1.3415,
|
| 1433 |
"step": 2340
|
| 1434 |
},
|
| 1435 |
{
|
| 1436 |
"epoch": 1.22,
|
| 1437 |
"learning_rate": 0.0004065040650406504,
|
| 1438 |
+
"loss": 1.3517,
|
| 1439 |
"step": 2350
|
| 1440 |
},
|
| 1441 |
{
|
| 1442 |
"epoch": 1.22,
|
| 1443 |
"learning_rate": 0.0004082338695727383,
|
| 1444 |
+
"loss": 1.1871,
|
| 1445 |
"step": 2360
|
| 1446 |
},
|
| 1447 |
{
|
| 1448 |
"epoch": 1.23,
|
| 1449 |
"learning_rate": 0.00040996367410482616,
|
| 1450 |
+
"loss": 1.541,
|
| 1451 |
"step": 2370
|
| 1452 |
},
|
| 1453 |
{
|
| 1454 |
"epoch": 1.24,
|
| 1455 |
"learning_rate": 0.00041169347863691404,
|
| 1456 |
+
"loss": 1.0871,
|
| 1457 |
"step": 2380
|
| 1458 |
},
|
| 1459 |
{
|
| 1460 |
"epoch": 1.24,
|
| 1461 |
"learning_rate": 0.0004134232831690019,
|
| 1462 |
+
"loss": 1.0727,
|
| 1463 |
"step": 2390
|
| 1464 |
},
|
| 1465 |
{
|
| 1466 |
"epoch": 1.25,
|
| 1467 |
"learning_rate": 0.0004151530877010898,
|
| 1468 |
+
"loss": 1.2664,
|
| 1469 |
"step": 2400
|
| 1470 |
},
|
| 1471 |
{
|
| 1472 |
"epoch": 1.25,
|
| 1473 |
"learning_rate": 0.0004168828922331777,
|
| 1474 |
+
"loss": 0.9518,
|
| 1475 |
"step": 2410
|
| 1476 |
},
|
| 1477 |
{
|
| 1478 |
"epoch": 1.26,
|
| 1479 |
"learning_rate": 0.00041861269676526556,
|
| 1480 |
+
"loss": 1.5851,
|
| 1481 |
"step": 2420
|
| 1482 |
},
|
| 1483 |
{
|
| 1484 |
"epoch": 1.26,
|
| 1485 |
"learning_rate": 0.0004203425012973534,
|
| 1486 |
+
"loss": 1.3433,
|
| 1487 |
"step": 2430
|
| 1488 |
},
|
| 1489 |
{
|
| 1490 |
"epoch": 1.27,
|
| 1491 |
"learning_rate": 0.00042207230582944126,
|
| 1492 |
+
"loss": 1.1357,
|
| 1493 |
"step": 2440
|
| 1494 |
},
|
| 1495 |
{
|
| 1496 |
"epoch": 1.27,
|
| 1497 |
"learning_rate": 0.00042380211036152914,
|
| 1498 |
+
"loss": 1.3124,
|
| 1499 |
"step": 2450
|
| 1500 |
},
|
| 1501 |
{
|
| 1502 |
"epoch": 1.28,
|
| 1503 |
"learning_rate": 0.000425531914893617,
|
| 1504 |
+
"loss": 1.3198,
|
| 1505 |
"step": 2460
|
| 1506 |
},
|
| 1507 |
{
|
| 1508 |
"epoch": 1.28,
|
| 1509 |
"learning_rate": 0.0004272617194257049,
|
| 1510 |
+
"loss": 1.31,
|
| 1511 |
"step": 2470
|
| 1512 |
},
|
| 1513 |
{
|
| 1514 |
"epoch": 1.29,
|
| 1515 |
"learning_rate": 0.0004289915239577928,
|
| 1516 |
+
"loss": 1.2563,
|
| 1517 |
"step": 2480
|
| 1518 |
},
|
| 1519 |
{
|
| 1520 |
"epoch": 1.29,
|
| 1521 |
"learning_rate": 0.00043072132848988065,
|
| 1522 |
+
"loss": 1.2553,
|
| 1523 |
"step": 2490
|
| 1524 |
},
|
| 1525 |
{
|
| 1526 |
"epoch": 1.3,
|
| 1527 |
"learning_rate": 0.00043245113302196853,
|
| 1528 |
+
"loss": 1.3314,
|
| 1529 |
"step": 2500
|
| 1530 |
},
|
| 1531 |
{
|
| 1532 |
"epoch": 1.3,
|
| 1533 |
"learning_rate": 0.0004341809375540564,
|
| 1534 |
+
"loss": 1.8077,
|
| 1535 |
"step": 2510
|
| 1536 |
},
|
| 1537 |
{
|
| 1538 |
"epoch": 1.31,
|
| 1539 |
"learning_rate": 0.0004359107420861443,
|
| 1540 |
+
"loss": 1.7739,
|
| 1541 |
"step": 2520
|
| 1542 |
},
|
| 1543 |
{
|
| 1544 |
"epoch": 1.31,
|
| 1545 |
"learning_rate": 0.00043764054661823217,
|
| 1546 |
+
"loss": 1.3835,
|
| 1547 |
"step": 2530
|
| 1548 |
},
|
| 1549 |
{
|
| 1550 |
"epoch": 1.32,
|
| 1551 |
"learning_rate": 0.00043937035115032005,
|
| 1552 |
+
"loss": 1.3689,
|
| 1553 |
"step": 2540
|
| 1554 |
},
|
| 1555 |
{
|
| 1556 |
"epoch": 1.32,
|
| 1557 |
"learning_rate": 0.0004411001556824079,
|
| 1558 |
+
"loss": 1.4259,
|
| 1559 |
"step": 2550
|
| 1560 |
},
|
| 1561 |
{
|
| 1562 |
"epoch": 1.33,
|
| 1563 |
"learning_rate": 0.0004428299602144958,
|
| 1564 |
+
"loss": 1.0997,
|
| 1565 |
"step": 2560
|
| 1566 |
},
|
| 1567 |
{
|
| 1568 |
"epoch": 1.33,
|
| 1569 |
"learning_rate": 0.0004445597647465837,
|
| 1570 |
+
"loss": 0.944,
|
| 1571 |
"step": 2570
|
| 1572 |
},
|
| 1573 |
{
|
| 1574 |
"epoch": 1.34,
|
| 1575 |
"learning_rate": 0.00044628956927867156,
|
| 1576 |
+
"loss": 1.431,
|
| 1577 |
"step": 2580
|
| 1578 |
},
|
| 1579 |
{
|
| 1580 |
"epoch": 1.34,
|
| 1581 |
"learning_rate": 0.0004480193738107594,
|
| 1582 |
+
"loss": 1.492,
|
| 1583 |
"step": 2590
|
| 1584 |
},
|
| 1585 |
{
|
| 1586 |
"epoch": 1.35,
|
| 1587 |
"learning_rate": 0.00044974917834284726,
|
| 1588 |
+
"loss": 1.0205,
|
| 1589 |
"step": 2600
|
| 1590 |
},
|
| 1591 |
{
|
| 1592 |
"epoch": 1.35,
|
| 1593 |
"learning_rate": 0.00045147898287493514,
|
| 1594 |
+
"loss": 1.2334,
|
| 1595 |
"step": 2610
|
| 1596 |
},
|
| 1597 |
{
|
| 1598 |
"epoch": 1.36,
|
| 1599 |
"learning_rate": 0.000453208787407023,
|
| 1600 |
+
"loss": 1.0434,
|
| 1601 |
"step": 2620
|
| 1602 |
},
|
| 1603 |
{
|
| 1604 |
"epoch": 1.36,
|
| 1605 |
"learning_rate": 0.0004549385919391109,
|
| 1606 |
+
"loss": 1.4912,
|
| 1607 |
"step": 2630
|
| 1608 |
},
|
| 1609 |
{
|
| 1610 |
"epoch": 1.37,
|
| 1611 |
"learning_rate": 0.0004566683964711988,
|
| 1612 |
+
"loss": 1.2939,
|
| 1613 |
"step": 2640
|
| 1614 |
},
|
| 1615 |
{
|
| 1616 |
"epoch": 1.38,
|
| 1617 |
"learning_rate": 0.00045839820100328666,
|
| 1618 |
+
"loss": 1.0389,
|
| 1619 |
"step": 2650
|
| 1620 |
},
|
| 1621 |
{
|
| 1622 |
"epoch": 1.38,
|
| 1623 |
"learning_rate": 0.00046012800553537454,
|
| 1624 |
+
"loss": 1.1925,
|
| 1625 |
"step": 2660
|
| 1626 |
},
|
| 1627 |
{
|
| 1628 |
"epoch": 1.39,
|
| 1629 |
"learning_rate": 0.0004618578100674624,
|
| 1630 |
+
"loss": 1.5455,
|
| 1631 |
"step": 2670
|
| 1632 |
},
|
| 1633 |
{
|
| 1634 |
"epoch": 1.39,
|
| 1635 |
"learning_rate": 0.00046358761459955024,
|
| 1636 |
+
"loss": 1.2916,
|
| 1637 |
"step": 2680
|
| 1638 |
},
|
| 1639 |
{
|
| 1640 |
"epoch": 1.4,
|
| 1641 |
"learning_rate": 0.0004653174191316381,
|
| 1642 |
+
"loss": 1.253,
|
| 1643 |
"step": 2690
|
| 1644 |
},
|
| 1645 |
{
|
| 1646 |
"epoch": 1.4,
|
| 1647 |
"learning_rate": 0.000467047223663726,
|
| 1648 |
+
"loss": 1.377,
|
| 1649 |
"step": 2700
|
| 1650 |
},
|
| 1651 |
{
|
| 1652 |
"epoch": 1.41,
|
| 1653 |
"learning_rate": 0.0004687770281958139,
|
| 1654 |
+
"loss": 1.4184,
|
| 1655 |
"step": 2710
|
| 1656 |
},
|
| 1657 |
{
|
| 1658 |
"epoch": 1.41,
|
| 1659 |
"learning_rate": 0.00047050683272790175,
|
| 1660 |
+
"loss": 1.3838,
|
| 1661 |
"step": 2720
|
| 1662 |
},
|
| 1663 |
{
|
| 1664 |
"epoch": 1.42,
|
| 1665 |
"learning_rate": 0.00047223663725998963,
|
| 1666 |
+
"loss": 1.1912,
|
| 1667 |
"step": 2730
|
| 1668 |
},
|
| 1669 |
{
|
| 1670 |
"epoch": 1.42,
|
| 1671 |
"learning_rate": 0.0004739664417920775,
|
| 1672 |
+
"loss": 1.0304,
|
| 1673 |
"step": 2740
|
| 1674 |
},
|
| 1675 |
{
|
| 1676 |
"epoch": 1.43,
|
| 1677 |
"learning_rate": 0.00047569624632416534,
|
| 1678 |
+
"loss": 1.3583,
|
| 1679 |
"step": 2750
|
| 1680 |
},
|
| 1681 |
{
|
| 1682 |
"epoch": 1.43,
|
| 1683 |
"learning_rate": 0.0004774260508562532,
|
| 1684 |
+
"loss": 1.3856,
|
| 1685 |
"step": 2760
|
| 1686 |
},
|
| 1687 |
{
|
| 1688 |
"epoch": 1.44,
|
| 1689 |
"learning_rate": 0.0004791558553883411,
|
| 1690 |
+
"loss": 1.5254,
|
| 1691 |
"step": 2770
|
| 1692 |
},
|
| 1693 |
{
|
| 1694 |
"epoch": 1.44,
|
| 1695 |
"learning_rate": 0.00048088565992042897,
|
| 1696 |
+
"loss": 1.2068,
|
| 1697 |
"step": 2780
|
| 1698 |
},
|
| 1699 |
{
|
| 1700 |
"epoch": 1.45,
|
| 1701 |
"learning_rate": 0.00048261546445251685,
|
| 1702 |
+
"loss": 1.3795,
|
| 1703 |
"step": 2790
|
| 1704 |
},
|
| 1705 |
{
|
| 1706 |
"epoch": 1.45,
|
| 1707 |
"learning_rate": 0.00048434526898460473,
|
| 1708 |
+
"loss": 1.5103,
|
| 1709 |
"step": 2800
|
| 1710 |
},
|
| 1711 |
{
|
| 1712 |
"epoch": 1.46,
|
| 1713 |
"learning_rate": 0.0004860750735166926,
|
| 1714 |
+
"loss": 1.6531,
|
| 1715 |
"step": 2810
|
| 1716 |
},
|
| 1717 |
{
|
| 1718 |
"epoch": 1.46,
|
| 1719 |
"learning_rate": 0.0004878048780487805,
|
| 1720 |
+
"loss": 1.3675,
|
| 1721 |
"step": 2820
|
| 1722 |
},
|
| 1723 |
{
|
| 1724 |
"epoch": 1.47,
|
| 1725 |
"learning_rate": 0.0004895346825808684,
|
| 1726 |
+
"loss": 1.6244,
|
| 1727 |
"step": 2830
|
| 1728 |
},
|
| 1729 |
{
|
| 1730 |
"epoch": 1.47,
|
| 1731 |
"learning_rate": 0.0004912644871129562,
|
| 1732 |
+
"loss": 1.8395,
|
| 1733 |
"step": 2840
|
| 1734 |
},
|
| 1735 |
{
|
| 1736 |
"epoch": 1.48,
|
| 1737 |
"learning_rate": 0.0004929942916450441,
|
| 1738 |
+
"loss": 1.6552,
|
| 1739 |
"step": 2850
|
| 1740 |
},
|
| 1741 |
{
|
| 1742 |
"epoch": 1.48,
|
| 1743 |
"learning_rate": 0.000494724096177132,
|
| 1744 |
+
"loss": 1.6432,
|
| 1745 |
"step": 2860
|
| 1746 |
},
|
| 1747 |
{
|
| 1748 |
"epoch": 1.49,
|
| 1749 |
"learning_rate": 0.0004964539007092199,
|
| 1750 |
+
"loss": 1.4839,
|
| 1751 |
"step": 2870
|
| 1752 |
},
|
| 1753 |
{
|
| 1754 |
"epoch": 1.49,
|
| 1755 |
"learning_rate": 0.0004981837052413078,
|
| 1756 |
+
"loss": 1.403,
|
| 1757 |
"step": 2880
|
| 1758 |
},
|
| 1759 |
{
|
| 1760 |
"epoch": 1.5,
|
| 1761 |
"learning_rate": 0.0004999135097733956,
|
| 1762 |
+
"loss": 1.8624,
|
| 1763 |
"step": 2890
|
| 1764 |
},
|
| 1765 |
{
|
| 1766 |
"epoch": 1.5,
|
| 1767 |
"learning_rate": 0.0005016433143054835,
|
| 1768 |
+
"loss": 1.4894,
|
| 1769 |
"step": 2900
|
| 1770 |
},
|
| 1771 |
{
|
| 1772 |
"epoch": 1.51,
|
| 1773 |
"learning_rate": 0.0005033731188375714,
|
| 1774 |
+
"loss": 2.1144,
|
| 1775 |
"step": 2910
|
| 1776 |
},
|
| 1777 |
{
|
| 1778 |
"epoch": 1.52,
|
| 1779 |
"learning_rate": 0.0005051029233696593,
|
| 1780 |
+
"loss": 1.7605,
|
| 1781 |
"step": 2920
|
| 1782 |
},
|
| 1783 |
{
|
| 1784 |
"epoch": 1.52,
|
| 1785 |
"learning_rate": 0.000506832727901747,
|
| 1786 |
+
"loss": 1.6352,
|
| 1787 |
"step": 2930
|
| 1788 |
},
|
| 1789 |
{
|
| 1790 |
"epoch": 1.53,
|
| 1791 |
"learning_rate": 0.000508562532433835,
|
| 1792 |
+
"loss": 1.6826,
|
| 1793 |
"step": 2940
|
| 1794 |
},
|
| 1795 |
{
|
| 1796 |
"epoch": 1.53,
|
| 1797 |
"learning_rate": 0.0005102923369659228,
|
| 1798 |
+
"loss": 1.8023,
|
| 1799 |
"step": 2950
|
| 1800 |
},
|
| 1801 |
{
|
| 1802 |
"epoch": 1.54,
|
| 1803 |
"learning_rate": 0.0005120221414980108,
|
| 1804 |
+
"loss": 1.6957,
|
| 1805 |
"step": 2960
|
| 1806 |
},
|
| 1807 |
{
|
| 1808 |
"epoch": 1.54,
|
| 1809 |
"learning_rate": 0.0005137519460300986,
|
| 1810 |
+
"loss": 2.1667,
|
| 1811 |
"step": 2970
|
| 1812 |
},
|
| 1813 |
{
|
| 1814 |
"epoch": 1.55,
|
| 1815 |
"learning_rate": 0.0005154817505621864,
|
| 1816 |
+
"loss": 2.1662,
|
| 1817 |
"step": 2980
|
| 1818 |
},
|
| 1819 |
{
|
| 1820 |
"epoch": 1.55,
|
| 1821 |
"learning_rate": 0.0005172115550942743,
|
| 1822 |
+
"loss": 1.8724,
|
| 1823 |
"step": 2990
|
| 1824 |
},
|
| 1825 |
{
|
| 1826 |
"epoch": 1.56,
|
| 1827 |
"learning_rate": 0.0005189413596263622,
|
| 1828 |
+
"loss": 1.7312,
|
| 1829 |
"step": 3000
|
| 1830 |
},
|
| 1831 |
{
|
| 1832 |
"epoch": 1.56,
|
| 1833 |
"learning_rate": 0.0005206711641584501,
|
| 1834 |
+
"loss": 2.0953,
|
| 1835 |
"step": 3010
|
| 1836 |
},
|
| 1837 |
{
|
| 1838 |
"epoch": 1.57,
|
| 1839 |
"learning_rate": 0.000522400968690538,
|
| 1840 |
+
"loss": 1.8218,
|
| 1841 |
"step": 3020
|
| 1842 |
},
|
| 1843 |
{
|
| 1844 |
"epoch": 1.57,
|
| 1845 |
"learning_rate": 0.0005241307732226258,
|
| 1846 |
+
"loss": 1.9732,
|
| 1847 |
"step": 3030
|
| 1848 |
},
|
| 1849 |
{
|
| 1850 |
"epoch": 1.58,
|
| 1851 |
"learning_rate": 0.0005258605777547137,
|
| 1852 |
+
"loss": 2.229,
|
| 1853 |
"step": 3040
|
| 1854 |
},
|
| 1855 |
{
|
| 1856 |
"epoch": 1.58,
|
| 1857 |
"learning_rate": 0.0005275903822868016,
|
| 1858 |
+
"loss": 1.9946,
|
| 1859 |
"step": 3050
|
| 1860 |
},
|
| 1861 |
{
|
| 1862 |
"epoch": 1.59,
|
| 1863 |
"learning_rate": 0.0005293201868188895,
|
| 1864 |
+
"loss": 2.162,
|
| 1865 |
"step": 3060
|
| 1866 |
},
|
| 1867 |
{
|
| 1868 |
"epoch": 1.59,
|
| 1869 |
"learning_rate": 0.0005310499913509773,
|
| 1870 |
+
"loss": 2.0647,
|
| 1871 |
"step": 3070
|
| 1872 |
},
|
| 1873 |
{
|
| 1874 |
"epoch": 1.6,
|
| 1875 |
"learning_rate": 0.0005327797958830652,
|
| 1876 |
+
"loss": 2.0291,
|
| 1877 |
"step": 3080
|
| 1878 |
},
|
| 1879 |
{
|
| 1880 |
"epoch": 1.6,
|
| 1881 |
"learning_rate": 0.0005345096004151531,
|
| 1882 |
+
"loss": 2.0622,
|
| 1883 |
"step": 3090
|
| 1884 |
},
|
| 1885 |
{
|
| 1886 |
"epoch": 1.61,
|
| 1887 |
"learning_rate": 0.000536239404947241,
|
| 1888 |
+
"loss": 1.9414,
|
| 1889 |
"step": 3100
|
| 1890 |
},
|
| 1891 |
{
|
| 1892 |
"epoch": 1.61,
|
| 1893 |
"learning_rate": 0.0005379692094793289,
|
| 1894 |
+
"loss": 2.0906,
|
| 1895 |
"step": 3110
|
| 1896 |
},
|
| 1897 |
{
|
| 1898 |
"epoch": 1.62,
|
| 1899 |
"learning_rate": 0.0005396990140114167,
|
| 1900 |
+
"loss": 2.0018,
|
| 1901 |
"step": 3120
|
| 1902 |
},
|
| 1903 |
{
|
| 1904 |
"epoch": 1.62,
|
| 1905 |
"learning_rate": 0.0005414288185435046,
|
| 1906 |
+
"loss": 1.9754,
|
| 1907 |
"step": 3130
|
| 1908 |
},
|
| 1909 |
{
|
| 1910 |
"epoch": 1.63,
|
| 1911 |
"learning_rate": 0.0005431586230755925,
|
| 1912 |
+
"loss": 1.949,
|
| 1913 |
"step": 3140
|
| 1914 |
},
|
| 1915 |
{
|
| 1916 |
"epoch": 1.63,
|
| 1917 |
"learning_rate": 0.0005448884276076803,
|
| 1918 |
+
"loss": 1.918,
|
| 1919 |
"step": 3150
|
| 1920 |
},
|
| 1921 |
{
|
| 1922 |
"epoch": 1.64,
|
| 1923 |
"learning_rate": 0.0005466182321397683,
|
| 1924 |
+
"loss": 2.1205,
|
| 1925 |
"step": 3160
|
| 1926 |
},
|
| 1927 |
{
|
| 1928 |
"epoch": 1.65,
|
| 1929 |
"learning_rate": 0.000548348036671856,
|
| 1930 |
+
"loss": 1.8933,
|
| 1931 |
"step": 3170
|
| 1932 |
},
|
| 1933 |
{
|
| 1934 |
"epoch": 1.65,
|
| 1935 |
"learning_rate": 0.000550077841203944,
|
| 1936 |
+
"loss": 1.9666,
|
| 1937 |
"step": 3180
|
| 1938 |
},
|
| 1939 |
{
|
| 1940 |
"epoch": 1.66,
|
| 1941 |
"learning_rate": 0.0005518076457360318,
|
| 1942 |
+
"loss": 1.9286,
|
| 1943 |
"step": 3190
|
| 1944 |
},
|
| 1945 |
{
|
| 1946 |
"epoch": 1.66,
|
| 1947 |
"learning_rate": 0.0005535374502681198,
|
| 1948 |
+
"loss": 1.9417,
|
| 1949 |
"step": 3200
|
| 1950 |
},
|
| 1951 |
{
|
| 1952 |
"epoch": 1.67,
|
| 1953 |
"learning_rate": 0.0005552672548002075,
|
| 1954 |
+
"loss": 1.8296,
|
| 1955 |
"step": 3210
|
| 1956 |
},
|
| 1957 |
{
|
| 1958 |
"epoch": 1.67,
|
| 1959 |
"learning_rate": 0.0005569970593322955,
|
| 1960 |
+
"loss": 1.9005,
|
| 1961 |
"step": 3220
|
| 1962 |
},
|
| 1963 |
{
|
| 1964 |
"epoch": 1.68,
|
| 1965 |
"learning_rate": 0.0005587268638643833,
|
| 1966 |
+
"loss": 1.954,
|
| 1967 |
"step": 3230
|
| 1968 |
},
|
| 1969 |
{
|
| 1970 |
"epoch": 1.68,
|
| 1971 |
"learning_rate": 0.0005604566683964713,
|
| 1972 |
+
"loss": 1.9252,
|
| 1973 |
"step": 3240
|
| 1974 |
},
|
| 1975 |
{
|
| 1976 |
"epoch": 1.69,
|
| 1977 |
"learning_rate": 0.000562186472928559,
|
| 1978 |
+
"loss": 1.9827,
|
| 1979 |
"step": 3250
|
| 1980 |
},
|
| 1981 |
{
|
| 1982 |
"epoch": 1.69,
|
| 1983 |
"learning_rate": 0.000563916277460647,
|
| 1984 |
+
"loss": 1.7931,
|
| 1985 |
"step": 3260
|
| 1986 |
},
|
| 1987 |
{
|
| 1988 |
"epoch": 1.7,
|
| 1989 |
"learning_rate": 0.0005656460819927348,
|
| 1990 |
+
"loss": 1.8842,
|
| 1991 |
"step": 3270
|
| 1992 |
},
|
| 1993 |
{
|
| 1994 |
"epoch": 1.7,
|
| 1995 |
"learning_rate": 0.0005673758865248228,
|
| 1996 |
+
"loss": 1.8838,
|
| 1997 |
"step": 3280
|
| 1998 |
},
|
| 1999 |
{
|
| 2000 |
"epoch": 1.71,
|
| 2001 |
"learning_rate": 0.0005691056910569106,
|
| 2002 |
+
"loss": 1.7902,
|
| 2003 |
"step": 3290
|
| 2004 |
},
|
| 2005 |
{
|
| 2006 |
"epoch": 1.71,
|
| 2007 |
"learning_rate": 0.0005708354955889984,
|
| 2008 |
+
"loss": 1.9054,
|
| 2009 |
"step": 3300
|
| 2010 |
},
|
| 2011 |
{
|
| 2012 |
"epoch": 1.72,
|
| 2013 |
"learning_rate": 0.0005725653001210863,
|
| 2014 |
+
"loss": 1.972,
|
| 2015 |
"step": 3310
|
| 2016 |
},
|
| 2017 |
{
|
| 2018 |
"epoch": 1.72,
|
| 2019 |
"learning_rate": 0.0005742951046531742,
|
| 2020 |
+
"loss": 1.999,
|
| 2021 |
"step": 3320
|
| 2022 |
},
|
| 2023 |
{
|
| 2024 |
"epoch": 1.73,
|
| 2025 |
"learning_rate": 0.0005760249091852621,
|
| 2026 |
+
"loss": 1.9566,
|
| 2027 |
"step": 3330
|
| 2028 |
},
|
| 2029 |
{
|
| 2030 |
"epoch": 1.73,
|
| 2031 |
"learning_rate": 0.00057775471371735,
|
| 2032 |
+
"loss": 1.8273,
|
| 2033 |
"step": 3340
|
| 2034 |
},
|
| 2035 |
{
|
| 2036 |
"epoch": 1.74,
|
| 2037 |
"learning_rate": 0.0005794845182494378,
|
| 2038 |
+
"loss": 1.9554,
|
| 2039 |
"step": 3350
|
| 2040 |
},
|
| 2041 |
{
|
| 2042 |
"epoch": 1.74,
|
| 2043 |
"learning_rate": 0.0005812143227815257,
|
| 2044 |
+
"loss": 2.0318,
|
| 2045 |
"step": 3360
|
| 2046 |
},
|
| 2047 |
{
|
| 2048 |
"epoch": 1.75,
|
| 2049 |
"learning_rate": 0.0005829441273136136,
|
| 2050 |
+
"loss": 1.9838,
|
| 2051 |
"step": 3370
|
| 2052 |
},
|
| 2053 |
{
|
| 2054 |
"epoch": 1.75,
|
| 2055 |
"learning_rate": 0.0005846739318457015,
|
| 2056 |
+
"loss": 1.9402,
|
| 2057 |
"step": 3380
|
| 2058 |
},
|
| 2059 |
{
|
| 2060 |
"epoch": 1.76,
|
| 2061 |
"learning_rate": 0.0005864037363777894,
|
| 2062 |
+
"loss": 1.9904,
|
| 2063 |
"step": 3390
|
| 2064 |
},
|
| 2065 |
{
|
| 2066 |
"epoch": 1.76,
|
| 2067 |
"learning_rate": 0.0005881335409098772,
|
| 2068 |
+
"loss": 1.9198,
|
| 2069 |
"step": 3400
|
| 2070 |
},
|
| 2071 |
{
|
| 2072 |
"epoch": 1.77,
|
| 2073 |
"learning_rate": 0.0005898633454419651,
|
| 2074 |
+
"loss": 1.9405,
|
| 2075 |
"step": 3410
|
| 2076 |
},
|
| 2077 |
{
|
| 2078 |
"epoch": 1.77,
|
| 2079 |
"learning_rate": 0.0005915931499740529,
|
| 2080 |
+
"loss": 1.8445,
|
| 2081 |
"step": 3420
|
| 2082 |
},
|
| 2083 |
{
|
| 2084 |
"epoch": 1.78,
|
| 2085 |
"learning_rate": 0.0005933229545061409,
|
| 2086 |
+
"loss": 1.9201,
|
| 2087 |
"step": 3430
|
| 2088 |
},
|
| 2089 |
{
|
| 2090 |
"epoch": 1.79,
|
| 2091 |
"learning_rate": 0.0005950527590382286,
|
| 2092 |
+
"loss": 1.8237,
|
| 2093 |
"step": 3440
|
| 2094 |
},
|
| 2095 |
{
|
| 2096 |
"epoch": 1.79,
|
| 2097 |
"learning_rate": 0.0005967825635703165,
|
| 2098 |
+
"loss": 1.7255,
|
| 2099 |
"step": 3450
|
| 2100 |
},
|
| 2101 |
{
|
| 2102 |
"epoch": 1.8,
|
| 2103 |
"learning_rate": 0.0005985123681024044,
|
| 2104 |
+
"loss": 1.844,
|
| 2105 |
"step": 3460
|
| 2106 |
},
|
| 2107 |
{
|
| 2108 |
"epoch": 1.8,
|
| 2109 |
"learning_rate": 0.0006002421726344923,
|
| 2110 |
+
"loss": 2.0433,
|
| 2111 |
"step": 3470
|
| 2112 |
},
|
| 2113 |
{
|
| 2114 |
"epoch": 1.81,
|
| 2115 |
"learning_rate": 0.0006019719771665802,
|
| 2116 |
+
"loss": 1.7753,
|
| 2117 |
"step": 3480
|
| 2118 |
},
|
| 2119 |
{
|
| 2120 |
"epoch": 1.81,
|
| 2121 |
"learning_rate": 0.000603701781698668,
|
| 2122 |
+
"loss": 1.858,
|
| 2123 |
"step": 3490
|
| 2124 |
},
|
| 2125 |
{
|
| 2126 |
"epoch": 1.82,
|
| 2127 |
"learning_rate": 0.0006054315862307559,
|
| 2128 |
+
"loss": 1.9348,
|
| 2129 |
"step": 3500
|
| 2130 |
},
|
| 2131 |
{
|
| 2132 |
"epoch": 1.82,
|
| 2133 |
"learning_rate": 0.0006071613907628438,
|
| 2134 |
+
"loss": 1.8196,
|
| 2135 |
"step": 3510
|
| 2136 |
},
|
| 2137 |
{
|
| 2138 |
"epoch": 1.83,
|
| 2139 |
"learning_rate": 0.0006088911952949317,
|
| 2140 |
+
"loss": 2.0001,
|
| 2141 |
"step": 3520
|
| 2142 |
},
|
| 2143 |
{
|
| 2144 |
"epoch": 1.83,
|
| 2145 |
"learning_rate": 0.0006106209998270195,
|
| 2146 |
+
"loss": 1.806,
|
| 2147 |
"step": 3530
|
| 2148 |
},
|
| 2149 |
{
|
| 2150 |
"epoch": 1.84,
|
| 2151 |
"learning_rate": 0.0006123508043591074,
|
| 2152 |
+
"loss": 1.8369,
|
| 2153 |
"step": 3540
|
| 2154 |
},
|
| 2155 |
{
|
| 2156 |
"epoch": 1.84,
|
| 2157 |
"learning_rate": 0.0006140806088911953,
|
| 2158 |
+
"loss": 1.8824,
|
| 2159 |
"step": 3550
|
| 2160 |
},
|
| 2161 |
{
|
| 2162 |
"epoch": 1.85,
|
| 2163 |
"learning_rate": 0.0006158104134232832,
|
| 2164 |
+
"loss": 1.8376,
|
| 2165 |
"step": 3560
|
| 2166 |
},
|
| 2167 |
{
|
| 2168 |
"epoch": 1.85,
|
| 2169 |
"learning_rate": 0.0006175402179553711,
|
| 2170 |
+
"loss": 1.7554,
|
| 2171 |
"step": 3570
|
| 2172 |
},
|
| 2173 |
{
|
| 2174 |
"epoch": 1.86,
|
| 2175 |
"learning_rate": 0.0006192700224874589,
|
| 2176 |
+
"loss": 1.835,
|
| 2177 |
"step": 3580
|
| 2178 |
},
|
| 2179 |
{
|
| 2180 |
"epoch": 1.86,
|
| 2181 |
"learning_rate": 0.0006209998270195468,
|
| 2182 |
+
"loss": 1.7473,
|
| 2183 |
"step": 3590
|
| 2184 |
},
|
| 2185 |
{
|
| 2186 |
"epoch": 1.87,
|
| 2187 |
"learning_rate": 0.0006227296315516347,
|
| 2188 |
+
"loss": 1.8827,
|
| 2189 |
"step": 3600
|
| 2190 |
},
|
| 2191 |
{
|
| 2192 |
"epoch": 1.87,
|
| 2193 |
"learning_rate": 0.0006244594360837226,
|
| 2194 |
+
"loss": 1.7536,
|
| 2195 |
"step": 3610
|
| 2196 |
},
|
| 2197 |
{
|
| 2198 |
"epoch": 1.88,
|
| 2199 |
"learning_rate": 0.0006261892406158103,
|
| 2200 |
+
"loss": 1.8997,
|
| 2201 |
"step": 3620
|
| 2202 |
},
|
| 2203 |
{
|
| 2204 |
"epoch": 1.88,
|
| 2205 |
"learning_rate": 0.0006279190451478983,
|
| 2206 |
+
"loss": 1.813,
|
| 2207 |
"step": 3630
|
| 2208 |
},
|
| 2209 |
{
|
| 2210 |
"epoch": 1.89,
|
| 2211 |
"learning_rate": 0.0006296488496799861,
|
| 2212 |
+
"loss": 1.8474,
|
| 2213 |
"step": 3640
|
| 2214 |
},
|
| 2215 |
{
|
| 2216 |
"epoch": 1.89,
|
| 2217 |
"learning_rate": 0.0006313786542120741,
|
| 2218 |
+
"loss": 1.8263,
|
| 2219 |
"step": 3650
|
| 2220 |
},
|
| 2221 |
{
|
| 2222 |
"epoch": 1.9,
|
| 2223 |
"learning_rate": 0.0006331084587441619,
|
| 2224 |
+
"loss": 1.8817,
|
| 2225 |
"step": 3660
|
| 2226 |
},
|
| 2227 |
{
|
| 2228 |
"epoch": 1.9,
|
| 2229 |
"learning_rate": 0.0006348382632762499,
|
| 2230 |
+
"loss": 1.9347,
|
| 2231 |
"step": 3670
|
| 2232 |
},
|
| 2233 |
{
|
| 2234 |
"epoch": 1.91,
|
| 2235 |
"learning_rate": 0.0006365680678083376,
|
| 2236 |
+
"loss": 1.786,
|
| 2237 |
"step": 3680
|
| 2238 |
},
|
| 2239 |
{
|
| 2240 |
"epoch": 1.91,
|
| 2241 |
"learning_rate": 0.0006382978723404256,
|
| 2242 |
+
"loss": 2.197,
|
| 2243 |
"step": 3690
|
| 2244 |
},
|
| 2245 |
{
|
| 2246 |
"epoch": 1.92,
|
| 2247 |
"learning_rate": 0.0006400276768725134,
|
| 2248 |
+
"loss": 1.9695,
|
| 2249 |
"step": 3700
|
| 2250 |
},
|
| 2251 |
{
|
| 2252 |
"epoch": 1.93,
|
| 2253 |
"learning_rate": 0.0006417574814046014,
|
| 2254 |
+
"loss": 1.8939,
|
| 2255 |
"step": 3710
|
| 2256 |
},
|
| 2257 |
{
|
| 2258 |
"epoch": 1.93,
|
| 2259 |
"learning_rate": 0.0006434872859366891,
|
| 2260 |
+
"loss": 1.8662,
|
| 2261 |
"step": 3720
|
| 2262 |
},
|
| 2263 |
{
|
| 2264 |
"epoch": 1.94,
|
| 2265 |
"learning_rate": 0.0006452170904687771,
|
| 2266 |
+
"loss": 1.8054,
|
| 2267 |
"step": 3730
|
| 2268 |
},
|
| 2269 |
{
|
| 2270 |
"epoch": 1.94,
|
| 2271 |
"learning_rate": 0.0006469468950008649,
|
| 2272 |
+
"loss": 1.8351,
|
| 2273 |
"step": 3740
|
| 2274 |
},
|
| 2275 |
{
|
| 2276 |
"epoch": 1.95,
|
| 2277 |
"learning_rate": 0.0006486766995329529,
|
| 2278 |
+
"loss": 1.9102,
|
| 2279 |
"step": 3750
|
| 2280 |
},
|
| 2281 |
{
|
| 2282 |
"epoch": 1.95,
|
| 2283 |
"learning_rate": 0.0006504065040650406,
|
| 2284 |
+
"loss": 1.8462,
|
| 2285 |
"step": 3760
|
| 2286 |
},
|
| 2287 |
{
|
| 2288 |
"epoch": 1.96,
|
| 2289 |
"learning_rate": 0.0006521363085971285,
|
| 2290 |
+
"loss": 1.894,
|
| 2291 |
"step": 3770
|
| 2292 |
},
|
| 2293 |
{
|
| 2294 |
"epoch": 1.96,
|
| 2295 |
"learning_rate": 0.0006538661131292164,
|
| 2296 |
+
"loss": 1.7559,
|
| 2297 |
"step": 3780
|
| 2298 |
},
|
| 2299 |
{
|
| 2300 |
"epoch": 1.97,
|
| 2301 |
"learning_rate": 0.0006555959176613043,
|
| 2302 |
+
"loss": 1.8255,
|
| 2303 |
"step": 3790
|
| 2304 |
},
|
| 2305 |
{
|
| 2306 |
"epoch": 1.97,
|
| 2307 |
"learning_rate": 0.0006573257221933922,
|
| 2308 |
+
"loss": 1.7118,
|
| 2309 |
"step": 3800
|
| 2310 |
},
|
| 2311 |
{
|
| 2312 |
"epoch": 1.98,
|
| 2313 |
"learning_rate": 0.00065905552672548,
|
| 2314 |
+
"loss": 1.8048,
|
| 2315 |
"step": 3810
|
| 2316 |
},
|
| 2317 |
{
|
| 2318 |
"epoch": 1.98,
|
| 2319 |
"learning_rate": 0.0006607853312575679,
|
| 2320 |
+
"loss": 1.7759,
|
| 2321 |
"step": 3820
|
| 2322 |
},
|
| 2323 |
{
|
| 2324 |
"epoch": 1.99,
|
| 2325 |
"learning_rate": 0.0006625151357896558,
|
| 2326 |
+
"loss": 1.7974,
|
| 2327 |
"step": 3830
|
| 2328 |
},
|
| 2329 |
{
|
| 2330 |
"epoch": 1.99,
|
| 2331 |
"learning_rate": 0.0006642449403217437,
|
| 2332 |
+
"loss": 1.7673,
|
| 2333 |
"step": 3840
|
| 2334 |
},
|
| 2335 |
{
|
| 2336 |
"epoch": 2.0,
|
| 2337 |
"learning_rate": 0.0006659747448538316,
|
| 2338 |
+
"loss": 1.7959,
|
| 2339 |
"step": 3850
|
| 2340 |
},
|
| 2341 |
{
|
| 2342 |
"epoch": 2.0,
|
| 2343 |
"eval_accuracy": {
|
| 2344 |
+
"accuracy": 0.26848249027237353
|
| 2345 |
},
|
| 2346 |
"eval_f1": {
|
| 2347 |
+
"f1": 0.17635435227079216
|
| 2348 |
},
|
| 2349 |
+
"eval_loss": 1.8579649925231934,
|
| 2350 |
"eval_precision": {
|
| 2351 |
+
"precision": 0.15600300783137744
|
| 2352 |
},
|
| 2353 |
"eval_recall": {
|
| 2354 |
+
"recall": 0.2748312652287413
|
| 2355 |
},
|
| 2356 |
+
"eval_runtime": 115.151,
|
| 2357 |
+
"eval_samples_per_second": 33.478,
|
| 2358 |
+
"eval_steps_per_second": 16.743,
|
| 2359 |
"step": 3854
|
| 2360 |
},
|
| 2361 |
{
|
| 2362 |
"epoch": 2.0,
|
| 2363 |
"learning_rate": 0.0006677045493859194,
|
| 2364 |
+
"loss": 1.9251,
|
| 2365 |
"step": 3860
|
| 2366 |
},
|
| 2367 |
{
|
| 2368 |
"epoch": 2.01,
|
| 2369 |
"learning_rate": 0.0006694343539180073,
|
| 2370 |
+
"loss": 1.7743,
|
| 2371 |
"step": 3870
|
| 2372 |
},
|
| 2373 |
{
|
| 2374 |
"epoch": 2.01,
|
| 2375 |
"learning_rate": 0.0006711641584500952,
|
| 2376 |
+
"loss": 1.8658,
|
| 2377 |
"step": 3880
|
| 2378 |
},
|
| 2379 |
{
|
| 2380 |
"epoch": 2.02,
|
| 2381 |
"learning_rate": 0.000672893962982183,
|
| 2382 |
+
"loss": 1.8682,
|
| 2383 |
"step": 3890
|
| 2384 |
},
|
| 2385 |
{
|
| 2386 |
"epoch": 2.02,
|
| 2387 |
"learning_rate": 0.000674623767514271,
|
| 2388 |
+
"loss": 1.9279,
|
| 2389 |
"step": 3900
|
| 2390 |
},
|
| 2391 |
{
|
| 2392 |
"epoch": 2.03,
|
| 2393 |
"learning_rate": 0.0006763535720463587,
|
| 2394 |
+
"loss": 2.0145,
|
| 2395 |
"step": 3910
|
| 2396 |
},
|
| 2397 |
{
|
| 2398 |
"epoch": 2.03,
|
| 2399 |
"learning_rate": 0.0006780833765784466,
|
| 2400 |
+
"loss": 2.0323,
|
| 2401 |
"step": 3920
|
| 2402 |
},
|
| 2403 |
{
|
| 2404 |
"epoch": 2.04,
|
| 2405 |
"learning_rate": 0.0006798131811105345,
|
| 2406 |
+
"loss": 1.9978,
|
| 2407 |
"step": 3930
|
| 2408 |
},
|
| 2409 |
{
|
| 2410 |
"epoch": 2.04,
|
| 2411 |
"learning_rate": 0.0006815429856426224,
|
| 2412 |
+
"loss": 1.9258,
|
| 2413 |
"step": 3940
|
| 2414 |
},
|
| 2415 |
{
|
| 2416 |
"epoch": 2.05,
|
| 2417 |
"learning_rate": 0.0006832727901747102,
|
| 2418 |
+
"loss": 1.9777,
|
| 2419 |
"step": 3950
|
| 2420 |
},
|
| 2421 |
{
|
| 2422 |
"epoch": 2.06,
|
| 2423 |
"learning_rate": 0.0006850025947067981,
|
| 2424 |
+
"loss": 1.9442,
|
| 2425 |
"step": 3960
|
| 2426 |
},
|
| 2427 |
{
|
| 2428 |
"epoch": 2.06,
|
| 2429 |
"learning_rate": 0.000686732399238886,
|
| 2430 |
+
"loss": 1.9772,
|
| 2431 |
"step": 3970
|
| 2432 |
},
|
| 2433 |
{
|
| 2434 |
"epoch": 2.07,
|
| 2435 |
"learning_rate": 0.0006884622037709739,
|
| 2436 |
+
"loss": 1.9368,
|
| 2437 |
"step": 3980
|
| 2438 |
},
|
| 2439 |
{
|
| 2440 |
"epoch": 2.07,
|
| 2441 |
"learning_rate": 0.0006901920083030618,
|
| 2442 |
+
"loss": 1.8948,
|
| 2443 |
"step": 3990
|
| 2444 |
},
|
| 2445 |
{
|
| 2446 |
"epoch": 2.08,
|
| 2447 |
"learning_rate": 0.0006919218128351496,
|
| 2448 |
+
"loss": 1.9485,
|
| 2449 |
"step": 4000
|
| 2450 |
},
|
| 2451 |
{
|
| 2452 |
"epoch": 2.08,
|
| 2453 |
"learning_rate": 0.0006936516173672375,
|
| 2454 |
+
"loss": 1.8472,
|
| 2455 |
"step": 4010
|
| 2456 |
},
|
| 2457 |
{
|
| 2458 |
"epoch": 2.09,
|
| 2459 |
"learning_rate": 0.0006953814218993254,
|
| 2460 |
+
"loss": 1.7604,
|
| 2461 |
"step": 4020
|
| 2462 |
},
|
| 2463 |
{
|
| 2464 |
"epoch": 2.09,
|
| 2465 |
"learning_rate": 0.0006971112264314133,
|
| 2466 |
+
"loss": 1.8371,
|
| 2467 |
"step": 4030
|
| 2468 |
},
|
| 2469 |
{
|
| 2470 |
"epoch": 2.1,
|
| 2471 |
"learning_rate": 0.0006988410309635011,
|
| 2472 |
+
"loss": 1.7566,
|
| 2473 |
"step": 4040
|
| 2474 |
},
|
| 2475 |
{
|
| 2476 |
"epoch": 2.1,
|
| 2477 |
"learning_rate": 0.000700570835495589,
|
| 2478 |
+
"loss": 1.9021,
|
| 2479 |
"step": 4050
|
| 2480 |
},
|
| 2481 |
{
|
| 2482 |
"epoch": 2.11,
|
| 2483 |
"learning_rate": 0.0007023006400276769,
|
| 2484 |
+
"loss": 1.744,
|
| 2485 |
"step": 4060
|
| 2486 |
},
|
| 2487 |
{
|
| 2488 |
"epoch": 2.11,
|
| 2489 |
"learning_rate": 0.0007040304445597648,
|
| 2490 |
+
"loss": 1.8639,
|
| 2491 |
"step": 4070
|
| 2492 |
},
|
| 2493 |
{
|
| 2494 |
"epoch": 2.12,
|
| 2495 |
"learning_rate": 0.0007057602490918527,
|
| 2496 |
+
"loss": 1.857,
|
| 2497 |
"step": 4080
|
| 2498 |
},
|
| 2499 |
{
|
| 2500 |
"epoch": 2.12,
|
| 2501 |
"learning_rate": 0.0007074900536239404,
|
| 2502 |
+
"loss": 1.8229,
|
| 2503 |
"step": 4090
|
| 2504 |
},
|
| 2505 |
{
|
| 2506 |
"epoch": 2.13,
|
| 2507 |
"learning_rate": 0.0007092198581560284,
|
| 2508 |
+
"loss": 1.8733,
|
| 2509 |
"step": 4100
|
| 2510 |
},
|
| 2511 |
{
|
| 2512 |
"epoch": 2.13,
|
| 2513 |
"learning_rate": 0.0007109496626881162,
|
| 2514 |
+
"loss": 1.8442,
|
| 2515 |
"step": 4110
|
| 2516 |
},
|
| 2517 |
{
|
| 2518 |
"epoch": 2.14,
|
| 2519 |
"learning_rate": 0.0007126794672202042,
|
| 2520 |
+
"loss": 1.8209,
|
| 2521 |
"step": 4120
|
| 2522 |
},
|
| 2523 |
{
|
| 2524 |
"epoch": 2.14,
|
| 2525 |
"learning_rate": 0.0007144092717522919,
|
| 2526 |
+
"loss": 1.7018,
|
| 2527 |
"step": 4130
|
| 2528 |
},
|
| 2529 |
{
|
| 2530 |
"epoch": 2.15,
|
| 2531 |
"learning_rate": 0.0007161390762843799,
|
| 2532 |
+
"loss": 1.8013,
|
| 2533 |
"step": 4140
|
| 2534 |
},
|
| 2535 |
{
|
| 2536 |
"epoch": 2.15,
|
| 2537 |
"learning_rate": 0.0007178688808164677,
|
| 2538 |
+
"loss": 1.9447,
|
| 2539 |
"step": 4150
|
| 2540 |
},
|
| 2541 |
{
|
| 2542 |
"epoch": 2.16,
|
| 2543 |
"learning_rate": 0.0007195986853485557,
|
| 2544 |
+
"loss": 1.8768,
|
| 2545 |
"step": 4160
|
| 2546 |
},
|
| 2547 |
{
|
| 2548 |
"epoch": 2.16,
|
| 2549 |
"learning_rate": 0.0007213284898806435,
|
| 2550 |
+
"loss": 1.822,
|
| 2551 |
"step": 4170
|
| 2552 |
},
|
| 2553 |
{
|
| 2554 |
"epoch": 2.17,
|
| 2555 |
"learning_rate": 0.0007230582944127314,
|
| 2556 |
+
"loss": 1.77,
|
| 2557 |
"step": 4180
|
| 2558 |
},
|
| 2559 |
{
|
| 2560 |
"epoch": 2.17,
|
| 2561 |
"learning_rate": 0.0007247880989448192,
|
| 2562 |
+
"loss": 1.8653,
|
| 2563 |
"step": 4190
|
| 2564 |
},
|
| 2565 |
{
|
| 2566 |
"epoch": 2.18,
|
| 2567 |
"learning_rate": 0.0007265179034769072,
|
| 2568 |
+
"loss": 1.7695,
|
| 2569 |
"step": 4200
|
| 2570 |
},
|
| 2571 |
{
|
| 2572 |
"epoch": 2.18,
|
| 2573 |
"learning_rate": 0.000728247708008995,
|
| 2574 |
+
"loss": 1.8426,
|
| 2575 |
"step": 4210
|
| 2576 |
},
|
| 2577 |
{
|
| 2578 |
"epoch": 2.19,
|
| 2579 |
"learning_rate": 0.000729977512541083,
|
| 2580 |
+
"loss": 1.8379,
|
| 2581 |
"step": 4220
|
| 2582 |
},
|
| 2583 |
{
|
| 2584 |
"epoch": 2.2,
|
| 2585 |
"learning_rate": 0.0007317073170731707,
|
| 2586 |
+
"loss": 1.8384,
|
| 2587 |
"step": 4230
|
| 2588 |
},
|
| 2589 |
{
|
| 2590 |
"epoch": 2.2,
|
| 2591 |
"learning_rate": 0.0007334371216052586,
|
| 2592 |
+
"loss": 1.9245,
|
| 2593 |
"step": 4240
|
| 2594 |
},
|
| 2595 |
{
|
| 2596 |
"epoch": 2.21,
|
| 2597 |
"learning_rate": 0.0007351669261373465,
|
| 2598 |
+
"loss": 1.8017,
|
| 2599 |
"step": 4250
|
| 2600 |
},
|
| 2601 |
{
|
| 2602 |
"epoch": 2.21,
|
| 2603 |
"learning_rate": 0.0007368967306694344,
|
| 2604 |
+
"loss": 1.8151,
|
| 2605 |
"step": 4260
|
| 2606 |
},
|
| 2607 |
{
|
| 2608 |
"epoch": 2.22,
|
| 2609 |
"learning_rate": 0.0007386265352015222,
|
| 2610 |
+
"loss": 1.8356,
|
| 2611 |
"step": 4270
|
| 2612 |
},
|
| 2613 |
{
|
| 2614 |
"epoch": 2.22,
|
| 2615 |
"learning_rate": 0.0007403563397336101,
|
| 2616 |
+
"loss": 1.9064,
|
| 2617 |
"step": 4280
|
| 2618 |
},
|
| 2619 |
{
|
| 2620 |
"epoch": 2.23,
|
| 2621 |
"learning_rate": 0.000742086144265698,
|
| 2622 |
+
"loss": 1.8325,
|
| 2623 |
"step": 4290
|
| 2624 |
},
|
| 2625 |
{
|
| 2626 |
"epoch": 2.23,
|
| 2627 |
"learning_rate": 0.0007438159487977859,
|
| 2628 |
+
"loss": 1.8079,
|
| 2629 |
"step": 4300
|
| 2630 |
},
|
| 2631 |
{
|
| 2632 |
"epoch": 2.24,
|
| 2633 |
"learning_rate": 0.0007455457533298738,
|
| 2634 |
+
"loss": 1.7581,
|
| 2635 |
"step": 4310
|
| 2636 |
},
|
| 2637 |
{
|
| 2638 |
"epoch": 2.24,
|
| 2639 |
"learning_rate": 0.0007472755578619616,
|
| 2640 |
+
"loss": 1.9346,
|
| 2641 |
"step": 4320
|
| 2642 |
},
|
| 2643 |
{
|
| 2644 |
"epoch": 2.25,
|
| 2645 |
"learning_rate": 0.0007490053623940495,
|
| 2646 |
+
"loss": 1.8706,
|
| 2647 |
"step": 4330
|
| 2648 |
},
|
| 2649 |
{
|
| 2650 |
"epoch": 2.25,
|
| 2651 |
"learning_rate": 0.0007507351669261374,
|
| 2652 |
+
"loss": 1.8314,
|
| 2653 |
"step": 4340
|
| 2654 |
},
|
| 2655 |
{
|
| 2656 |
"epoch": 2.26,
|
| 2657 |
"learning_rate": 0.0007524649714582253,
|
| 2658 |
+
"loss": 1.8881,
|
| 2659 |
"step": 4350
|
| 2660 |
},
|
| 2661 |
{
|
| 2662 |
"epoch": 2.26,
|
| 2663 |
"learning_rate": 0.0007541947759903132,
|
| 2664 |
+
"loss": 1.8785,
|
| 2665 |
"step": 4360
|
| 2666 |
},
|
| 2667 |
{
|
| 2668 |
"epoch": 2.27,
|
| 2669 |
"learning_rate": 0.000755924580522401,
|
| 2670 |
+
"loss": 1.7916,
|
| 2671 |
"step": 4370
|
| 2672 |
},
|
| 2673 |
{
|
| 2674 |
"epoch": 2.27,
|
| 2675 |
"learning_rate": 0.0007576543850544888,
|
| 2676 |
+
"loss": 1.8089,
|
| 2677 |
"step": 4380
|
| 2678 |
},
|
| 2679 |
{
|
| 2680 |
"epoch": 2.28,
|
| 2681 |
"learning_rate": 0.0007593841895865767,
|
| 2682 |
+
"loss": 1.7736,
|
| 2683 |
"step": 4390
|
| 2684 |
},
|
| 2685 |
{
|
| 2686 |
"epoch": 2.28,
|
| 2687 |
"learning_rate": 0.0007611139941186646,
|
| 2688 |
+
"loss": 1.8184,
|
| 2689 |
"step": 4400
|
| 2690 |
},
|
| 2691 |
{
|
| 2692 |
"epoch": 2.29,
|
| 2693 |
"learning_rate": 0.0007628437986507524,
|
| 2694 |
+
"loss": 1.8084,
|
| 2695 |
"step": 4410
|
| 2696 |
},
|
| 2697 |
{
|
| 2698 |
"epoch": 2.29,
|
| 2699 |
"learning_rate": 0.0007645736031828403,
|
| 2700 |
+
"loss": 2.0347,
|
| 2701 |
"step": 4420
|
| 2702 |
},
|
| 2703 |
{
|
| 2704 |
"epoch": 2.3,
|
| 2705 |
"learning_rate": 0.0007663034077149282,
|
| 2706 |
+
"loss": 1.7984,
|
| 2707 |
"step": 4430
|
| 2708 |
},
|
| 2709 |
{
|
| 2710 |
"epoch": 2.3,
|
| 2711 |
"learning_rate": 0.0007680332122470161,
|
| 2712 |
+
"loss": 1.7799,
|
| 2713 |
"step": 4440
|
| 2714 |
},
|
| 2715 |
{
|
| 2716 |
"epoch": 2.31,
|
| 2717 |
"learning_rate": 0.000769763016779104,
|
| 2718 |
+
"loss": 1.7603,
|
| 2719 |
"step": 4450
|
| 2720 |
},
|
| 2721 |
{
|
| 2722 |
"epoch": 2.31,
|
| 2723 |
"learning_rate": 0.0007714928213111918,
|
| 2724 |
+
"loss": 1.6663,
|
| 2725 |
"step": 4460
|
| 2726 |
},
|
| 2727 |
{
|
| 2728 |
"epoch": 2.32,
|
| 2729 |
"learning_rate": 0.0007732226258432797,
|
| 2730 |
+
"loss": 1.9232,
|
| 2731 |
"step": 4470
|
| 2732 |
},
|
| 2733 |
{
|
| 2734 |
"epoch": 2.32,
|
| 2735 |
"learning_rate": 0.0007749524303753676,
|
| 2736 |
+
"loss": 1.8086,
|
| 2737 |
"step": 4480
|
| 2738 |
},
|
| 2739 |
{
|
| 2740 |
"epoch": 2.33,
|
| 2741 |
"learning_rate": 0.0007766822349074555,
|
| 2742 |
+
"loss": 1.7616,
|
| 2743 |
"step": 4490
|
| 2744 |
},
|
| 2745 |
{
|
| 2746 |
"epoch": 2.34,
|
| 2747 |
"learning_rate": 0.0007784120394395433,
|
| 2748 |
+
"loss": 1.8018,
|
| 2749 |
"step": 4500
|
| 2750 |
},
|
| 2751 |
{
|
| 2752 |
"epoch": 2.34,
|
| 2753 |
"learning_rate": 0.0007801418439716312,
|
| 2754 |
+
"loss": 1.8413,
|
| 2755 |
"step": 4510
|
| 2756 |
},
|
| 2757 |
{
|
| 2758 |
"epoch": 2.35,
|
| 2759 |
"learning_rate": 0.0007818716485037191,
|
| 2760 |
+
"loss": 2.1126,
|
| 2761 |
"step": 4520
|
| 2762 |
},
|
| 2763 |
{
|
| 2764 |
"epoch": 2.35,
|
| 2765 |
"learning_rate": 0.000783601453035807,
|
| 2766 |
+
"loss": 1.9105,
|
| 2767 |
"step": 4530
|
| 2768 |
},
|
| 2769 |
{
|
| 2770 |
"epoch": 2.36,
|
| 2771 |
"learning_rate": 0.0007853312575678949,
|
| 2772 |
+
"loss": 1.8485,
|
| 2773 |
"step": 4540
|
| 2774 |
},
|
| 2775 |
{
|
| 2776 |
"epoch": 2.36,
|
| 2777 |
"learning_rate": 0.0007870610620999827,
|
| 2778 |
+
"loss": 1.8835,
|
| 2779 |
"step": 4550
|
| 2780 |
},
|
| 2781 |
{
|
| 2782 |
"epoch": 2.37,
|
| 2783 |
"learning_rate": 0.0007887908666320705,
|
| 2784 |
+
"loss": 1.7881,
|
| 2785 |
"step": 4560
|
| 2786 |
},
|
| 2787 |
{
|
| 2788 |
"epoch": 2.37,
|
| 2789 |
"learning_rate": 0.0007905206711641585,
|
| 2790 |
+
"loss": 1.7013,
|
| 2791 |
"step": 4570
|
| 2792 |
},
|
| 2793 |
{
|
| 2794 |
"epoch": 2.38,
|
| 2795 |
"learning_rate": 0.0007922504756962463,
|
| 2796 |
+
"loss": 1.7657,
|
| 2797 |
"step": 4580
|
| 2798 |
},
|
| 2799 |
{
|
| 2800 |
"epoch": 2.38,
|
| 2801 |
"learning_rate": 0.0007939802802283343,
|
| 2802 |
+
"loss": 1.9326,
|
| 2803 |
"step": 4590
|
| 2804 |
},
|
| 2805 |
{
|
| 2806 |
"epoch": 2.39,
|
| 2807 |
"learning_rate": 0.000795710084760422,
|
| 2808 |
+
"loss": 1.6888,
|
| 2809 |
"step": 4600
|
| 2810 |
},
|
| 2811 |
{
|
| 2812 |
"epoch": 2.39,
|
| 2813 |
"learning_rate": 0.00079743988929251,
|
| 2814 |
+
"loss": 1.8114,
|
| 2815 |
"step": 4610
|
| 2816 |
},
|
| 2817 |
{
|
| 2818 |
"epoch": 2.4,
|
| 2819 |
"learning_rate": 0.0007991696938245978,
|
| 2820 |
+
"loss": 1.9245,
|
| 2821 |
"step": 4620
|
| 2822 |
},
|
| 2823 |
{
|
| 2824 |
"epoch": 2.4,
|
| 2825 |
"learning_rate": 0.0008008994983566858,
|
| 2826 |
+
"loss": 1.8287,
|
| 2827 |
"step": 4630
|
| 2828 |
},
|
| 2829 |
{
|
| 2830 |
"epoch": 2.41,
|
| 2831 |
"learning_rate": 0.0008026293028887735,
|
| 2832 |
+
"loss": 1.8763,
|
| 2833 |
"step": 4640
|
| 2834 |
},
|
| 2835 |
{
|
| 2836 |
"epoch": 2.41,
|
| 2837 |
"learning_rate": 0.0008043591074208615,
|
| 2838 |
+
"loss": 1.6896,
|
| 2839 |
"step": 4650
|
| 2840 |
},
|
| 2841 |
{
|
| 2842 |
"epoch": 2.42,
|
| 2843 |
"learning_rate": 0.0008060889119529493,
|
| 2844 |
+
"loss": 1.9996,
|
| 2845 |
"step": 4660
|
| 2846 |
},
|
| 2847 |
{
|
| 2848 |
"epoch": 2.42,
|
| 2849 |
"learning_rate": 0.0008078187164850373,
|
| 2850 |
+
"loss": 1.7345,
|
| 2851 |
"step": 4670
|
| 2852 |
},
|
| 2853 |
{
|
| 2854 |
"epoch": 2.43,
|
| 2855 |
"learning_rate": 0.0008095485210171251,
|
| 2856 |
+
"loss": 1.7908,
|
| 2857 |
"step": 4680
|
| 2858 |
},
|
| 2859 |
{
|
| 2860 |
"epoch": 2.43,
|
| 2861 |
"learning_rate": 0.000811278325549213,
|
| 2862 |
+
"loss": 1.8212,
|
| 2863 |
"step": 4690
|
| 2864 |
},
|
| 2865 |
{
|
| 2866 |
"epoch": 2.44,
|
| 2867 |
"learning_rate": 0.0008130081300813008,
|
| 2868 |
+
"loss": 1.8434,
|
| 2869 |
"step": 4700
|
| 2870 |
},
|
| 2871 |
{
|
| 2872 |
"epoch": 2.44,
|
| 2873 |
"learning_rate": 0.0008147379346133887,
|
| 2874 |
+
"loss": 1.9086,
|
| 2875 |
"step": 4710
|
| 2876 |
},
|
| 2877 |
{
|
| 2878 |
"epoch": 2.45,
|
| 2879 |
"learning_rate": 0.0008164677391454766,
|
| 2880 |
+
"loss": 1.9098,
|
| 2881 |
"step": 4720
|
| 2882 |
},
|
| 2883 |
{
|
| 2884 |
"epoch": 2.45,
|
| 2885 |
"learning_rate": 0.0008181975436775644,
|
| 2886 |
+
"loss": 1.7695,
|
| 2887 |
"step": 4730
|
| 2888 |
},
|
| 2889 |
{
|
| 2890 |
"epoch": 2.46,
|
| 2891 |
"learning_rate": 0.0008199273482096523,
|
| 2892 |
+
"loss": 1.7415,
|
| 2893 |
"step": 4740
|
| 2894 |
},
|
| 2895 |
{
|
| 2896 |
"epoch": 2.46,
|
| 2897 |
"learning_rate": 0.0008216571527417402,
|
| 2898 |
+
"loss": 1.8291,
|
| 2899 |
"step": 4750
|
| 2900 |
},
|
| 2901 |
{
|
| 2902 |
"epoch": 2.47,
|
| 2903 |
"learning_rate": 0.0008233869572738281,
|
| 2904 |
+
"loss": 1.7842,
|
| 2905 |
"step": 4760
|
| 2906 |
},
|
| 2907 |
{
|
| 2908 |
"epoch": 2.48,
|
| 2909 |
"learning_rate": 0.000825116761805916,
|
| 2910 |
+
"loss": 1.8745,
|
| 2911 |
"step": 4770
|
| 2912 |
},
|
| 2913 |
{
|
| 2914 |
"epoch": 2.48,
|
| 2915 |
"learning_rate": 0.0008268465663380038,
|
| 2916 |
+
"loss": 1.7075,
|
| 2917 |
"step": 4780
|
| 2918 |
},
|
| 2919 |
{
|
| 2920 |
"epoch": 2.49,
|
| 2921 |
"learning_rate": 0.0008285763708700917,
|
| 2922 |
+
"loss": 1.8133,
|
| 2923 |
"step": 4790
|
| 2924 |
},
|
| 2925 |
{
|
| 2926 |
"epoch": 2.49,
|
| 2927 |
"learning_rate": 0.0008303061754021796,
|
| 2928 |
+
"loss": 1.7318,
|
| 2929 |
"step": 4800
|
| 2930 |
},
|
| 2931 |
{
|
| 2932 |
"epoch": 2.5,
|
| 2933 |
"learning_rate": 0.0008320359799342675,
|
| 2934 |
+
"loss": 1.874,
|
| 2935 |
"step": 4810
|
| 2936 |
},
|
| 2937 |
{
|
| 2938 |
"epoch": 2.5,
|
| 2939 |
"learning_rate": 0.0008337657844663554,
|
| 2940 |
+
"loss": 1.8895,
|
| 2941 |
"step": 4820
|
| 2942 |
},
|
| 2943 |
{
|
| 2944 |
"epoch": 2.51,
|
| 2945 |
"learning_rate": 0.0008354955889984432,
|
| 2946 |
+
"loss": 1.8485,
|
| 2947 |
"step": 4830
|
| 2948 |
},
|
| 2949 |
{
|
| 2950 |
"epoch": 2.51,
|
| 2951 |
"learning_rate": 0.0008372253935305311,
|
| 2952 |
+
"loss": 1.7643,
|
| 2953 |
"step": 4840
|
| 2954 |
},
|
| 2955 |
{
|
| 2956 |
"epoch": 2.52,
|
| 2957 |
"learning_rate": 0.000838955198062619,
|
| 2958 |
+
"loss": 1.8477,
|
| 2959 |
"step": 4850
|
| 2960 |
},
|
| 2961 |
{
|
| 2962 |
"epoch": 2.52,
|
| 2963 |
"learning_rate": 0.0008406850025947068,
|
| 2964 |
+
"loss": 1.8611,
|
| 2965 |
"step": 4860
|
| 2966 |
},
|
| 2967 |
{
|
| 2968 |
"epoch": 2.53,
|
| 2969 |
"learning_rate": 0.0008424148071267946,
|
| 2970 |
+
"loss": 1.8908,
|
| 2971 |
"step": 4870
|
| 2972 |
},
|
| 2973 |
{
|
| 2974 |
"epoch": 2.53,
|
| 2975 |
"learning_rate": 0.0008441446116588825,
|
| 2976 |
+
"loss": 1.8017,
|
| 2977 |
"step": 4880
|
| 2978 |
},
|
| 2979 |
{
|
| 2980 |
"epoch": 2.54,
|
| 2981 |
"learning_rate": 0.0008458744161909704,
|
| 2982 |
+
"loss": 1.9186,
|
| 2983 |
"step": 4890
|
| 2984 |
},
|
| 2985 |
{
|
| 2986 |
"epoch": 2.54,
|
| 2987 |
"learning_rate": 0.0008476042207230583,
|
| 2988 |
+
"loss": 1.8141,
|
| 2989 |
"step": 4900
|
| 2990 |
},
|
| 2991 |
{
|
| 2992 |
"epoch": 2.55,
|
| 2993 |
"learning_rate": 0.0008493340252551462,
|
| 2994 |
+
"loss": 1.8298,
|
| 2995 |
"step": 4910
|
| 2996 |
},
|
| 2997 |
{
|
| 2998 |
"epoch": 2.55,
|
| 2999 |
"learning_rate": 0.000851063829787234,
|
| 3000 |
+
"loss": 1.7711,
|
| 3001 |
"step": 4920
|
| 3002 |
},
|
| 3003 |
{
|
| 3004 |
"epoch": 2.56,
|
| 3005 |
"learning_rate": 0.0008527936343193219,
|
| 3006 |
+
"loss": 1.8265,
|
| 3007 |
"step": 4930
|
| 3008 |
},
|
| 3009 |
{
|
| 3010 |
"epoch": 2.56,
|
| 3011 |
"learning_rate": 0.0008545234388514098,
|
| 3012 |
+
"loss": 1.7649,
|
| 3013 |
"step": 4940
|
| 3014 |
},
|
| 3015 |
{
|
| 3016 |
"epoch": 2.57,
|
| 3017 |
"learning_rate": 0.0008562532433834977,
|
| 3018 |
+
"loss": 1.9159,
|
| 3019 |
"step": 4950
|
| 3020 |
},
|
| 3021 |
{
|
| 3022 |
"epoch": 2.57,
|
| 3023 |
"learning_rate": 0.0008579830479155855,
|
| 3024 |
+
"loss": 1.7968,
|
| 3025 |
"step": 4960
|
| 3026 |
},
|
| 3027 |
{
|
| 3028 |
"epoch": 2.58,
|
| 3029 |
"learning_rate": 0.0008597128524476734,
|
| 3030 |
+
"loss": 1.807,
|
| 3031 |
"step": 4970
|
| 3032 |
},
|
| 3033 |
{
|
| 3034 |
"epoch": 2.58,
|
| 3035 |
"learning_rate": 0.0008614426569797613,
|
| 3036 |
+
"loss": 1.7633,
|
| 3037 |
"step": 4980
|
| 3038 |
},
|
| 3039 |
{
|
| 3040 |
"epoch": 2.59,
|
| 3041 |
"learning_rate": 0.0008631724615118492,
|
| 3042 |
+
"loss": 1.8019,
|
| 3043 |
"step": 4990
|
| 3044 |
},
|
| 3045 |
{
|
| 3046 |
"epoch": 2.59,
|
| 3047 |
"learning_rate": 0.0008649022660439371,
|
| 3048 |
+
"loss": 1.8697,
|
| 3049 |
"step": 5000
|
| 3050 |
},
|
| 3051 |
{
|
| 3052 |
"epoch": 2.6,
|
| 3053 |
"learning_rate": 0.0008666320705760249,
|
| 3054 |
+
"loss": 1.8348,
|
| 3055 |
"step": 5010
|
| 3056 |
},
|
| 3057 |
{
|
| 3058 |
"epoch": 2.61,
|
| 3059 |
"learning_rate": 0.0008683618751081128,
|
| 3060 |
+
"loss": 1.805,
|
| 3061 |
"step": 5020
|
| 3062 |
},
|
| 3063 |
{
|
| 3064 |
"epoch": 2.61,
|
| 3065 |
"learning_rate": 0.0008700916796402006,
|
| 3066 |
+
"loss": 1.7346,
|
| 3067 |
"step": 5030
|
| 3068 |
},
|
| 3069 |
{
|
| 3070 |
"epoch": 2.62,
|
| 3071 |
"learning_rate": 0.0008718214841722886,
|
| 3072 |
+
"loss": 1.9219,
|
| 3073 |
"step": 5040
|
| 3074 |
},
|
| 3075 |
{
|
| 3076 |
"epoch": 2.62,
|
| 3077 |
"learning_rate": 0.0008735512887043763,
|
| 3078 |
+
"loss": 2.0954,
|
| 3079 |
"step": 5050
|
| 3080 |
},
|
| 3081 |
{
|
| 3082 |
"epoch": 2.63,
|
| 3083 |
"learning_rate": 0.0008752810932364643,
|
| 3084 |
+
"loss": 1.9262,
|
| 3085 |
"step": 5060
|
| 3086 |
},
|
| 3087 |
{
|
| 3088 |
"epoch": 2.63,
|
| 3089 |
"learning_rate": 0.0008770108977685521,
|
| 3090 |
+
"loss": 1.8491,
|
| 3091 |
"step": 5070
|
| 3092 |
},
|
| 3093 |
{
|
| 3094 |
"epoch": 2.64,
|
| 3095 |
"learning_rate": 0.0008787407023006401,
|
| 3096 |
+
"loss": 2.0113,
|
| 3097 |
"step": 5080
|
| 3098 |
},
|
| 3099 |
{
|
| 3100 |
"epoch": 2.64,
|
| 3101 |
"learning_rate": 0.0008804705068327279,
|
| 3102 |
+
"loss": 1.9341,
|
| 3103 |
"step": 5090
|
| 3104 |
},
|
| 3105 |
{
|
| 3106 |
"epoch": 2.65,
|
| 3107 |
"learning_rate": 0.0008822003113648159,
|
| 3108 |
+
"loss": 1.8968,
|
| 3109 |
"step": 5100
|
| 3110 |
},
|
| 3111 |
{
|
| 3112 |
"epoch": 2.65,
|
| 3113 |
"learning_rate": 0.0008839301158969036,
|
| 3114 |
+
"loss": 2.0009,
|
| 3115 |
"step": 5110
|
| 3116 |
},
|
| 3117 |
{
|
| 3118 |
"epoch": 2.66,
|
| 3119 |
"learning_rate": 0.0008856599204289916,
|
| 3120 |
+
"loss": 1.9956,
|
| 3121 |
"step": 5120
|
| 3122 |
},
|
| 3123 |
{
|
| 3124 |
"epoch": 2.66,
|
| 3125 |
"learning_rate": 0.0008873897249610794,
|
| 3126 |
+
"loss": 1.9083,
|
| 3127 |
"step": 5130
|
| 3128 |
},
|
| 3129 |
{
|
| 3130 |
"epoch": 2.67,
|
| 3131 |
"learning_rate": 0.0008891195294931674,
|
| 3132 |
+
"loss": 1.9274,
|
| 3133 |
"step": 5140
|
| 3134 |
},
|
| 3135 |
{
|
| 3136 |
"epoch": 2.67,
|
| 3137 |
"learning_rate": 0.0008908493340252551,
|
| 3138 |
+
"loss": 1.8263,
|
| 3139 |
"step": 5150
|
| 3140 |
},
|
| 3141 |
{
|
| 3142 |
"epoch": 2.68,
|
| 3143 |
"learning_rate": 0.0008925791385573431,
|
| 3144 |
+
"loss": 1.8235,
|
| 3145 |
"step": 5160
|
| 3146 |
},
|
| 3147 |
{
|
| 3148 |
"epoch": 2.68,
|
| 3149 |
"learning_rate": 0.0008943089430894309,
|
| 3150 |
+
"loss": 1.8786,
|
| 3151 |
"step": 5170
|
| 3152 |
},
|
| 3153 |
{
|
| 3154 |
"epoch": 2.69,
|
| 3155 |
"learning_rate": 0.0008960387476215188,
|
| 3156 |
+
"loss": 1.8952,
|
| 3157 |
"step": 5180
|
| 3158 |
},
|
| 3159 |
{
|
| 3160 |
"epoch": 2.69,
|
| 3161 |
"learning_rate": 0.0008977685521536067,
|
| 3162 |
+
"loss": 1.7448,
|
| 3163 |
"step": 5190
|
| 3164 |
},
|
| 3165 |
{
|
| 3166 |
"epoch": 2.7,
|
| 3167 |
"learning_rate": 0.0008994983566856945,
|
| 3168 |
+
"loss": 1.9518,
|
| 3169 |
"step": 5200
|
| 3170 |
},
|
| 3171 |
{
|
| 3172 |
"epoch": 2.7,
|
| 3173 |
"learning_rate": 0.0009012281612177824,
|
| 3174 |
+
"loss": 1.8636,
|
| 3175 |
"step": 5210
|
| 3176 |
},
|
| 3177 |
{
|
| 3178 |
"epoch": 2.71,
|
| 3179 |
"learning_rate": 0.0009029579657498703,
|
| 3180 |
+
"loss": 1.9309,
|
| 3181 |
"step": 5220
|
| 3182 |
},
|
| 3183 |
{
|
| 3184 |
"epoch": 2.71,
|
| 3185 |
"learning_rate": 0.0009046877702819582,
|
| 3186 |
+
"loss": 1.8714,
|
| 3187 |
"step": 5230
|
| 3188 |
},
|
| 3189 |
{
|
| 3190 |
"epoch": 2.72,
|
| 3191 |
"learning_rate": 0.000906417574814046,
|
| 3192 |
+
"loss": 1.7624,
|
| 3193 |
"step": 5240
|
| 3194 |
},
|
| 3195 |
{
|
| 3196 |
"epoch": 2.72,
|
| 3197 |
"learning_rate": 0.0009081473793461339,
|
| 3198 |
+
"loss": 1.9421,
|
| 3199 |
"step": 5250
|
| 3200 |
},
|
| 3201 |
{
|
| 3202 |
"epoch": 2.73,
|
| 3203 |
"learning_rate": 0.0009098771838782218,
|
| 3204 |
+
"loss": 1.9204,
|
| 3205 |
"step": 5260
|
| 3206 |
},
|
| 3207 |
{
|
| 3208 |
"epoch": 2.73,
|
| 3209 |
"learning_rate": 0.0009116069884103097,
|
| 3210 |
+
"loss": 1.8561,
|
| 3211 |
"step": 5270
|
| 3212 |
},
|
| 3213 |
{
|
| 3214 |
"epoch": 2.74,
|
| 3215 |
"learning_rate": 0.0009133367929423976,
|
| 3216 |
+
"loss": 1.8766,
|
| 3217 |
"step": 5280
|
| 3218 |
},
|
| 3219 |
{
|
| 3220 |
"epoch": 2.75,
|
| 3221 |
"learning_rate": 0.0009150665974744854,
|
| 3222 |
+
"loss": 1.8905,
|
| 3223 |
"step": 5290
|
| 3224 |
},
|
| 3225 |
{
|
| 3226 |
"epoch": 2.75,
|
| 3227 |
"learning_rate": 0.0009167964020065733,
|
| 3228 |
+
"loss": 1.8466,
|
| 3229 |
"step": 5300
|
| 3230 |
},
|
| 3231 |
{
|
| 3232 |
"epoch": 2.76,
|
| 3233 |
"learning_rate": 0.0009185262065386612,
|
| 3234 |
+
"loss": 1.8735,
|
| 3235 |
"step": 5310
|
| 3236 |
},
|
| 3237 |
{
|
| 3238 |
"epoch": 2.76,
|
| 3239 |
"learning_rate": 0.0009202560110707491,
|
| 3240 |
+
"loss": 2.0758,
|
| 3241 |
"step": 5320
|
| 3242 |
},
|
| 3243 |
{
|
| 3244 |
"epoch": 2.77,
|
| 3245 |
"learning_rate": 0.0009219858156028368,
|
| 3246 |
+
"loss": 1.9582,
|
| 3247 |
"step": 5330
|
| 3248 |
},
|
| 3249 |
{
|
| 3250 |
"epoch": 2.77,
|
| 3251 |
"learning_rate": 0.0009237156201349248,
|
| 3252 |
+
"loss": 1.9317,
|
| 3253 |
"step": 5340
|
| 3254 |
},
|
| 3255 |
{
|
| 3256 |
"epoch": 2.78,
|
| 3257 |
"learning_rate": 0.0009254454246670126,
|
| 3258 |
+
"loss": 1.8719,
|
| 3259 |
"step": 5350
|
| 3260 |
},
|
| 3261 |
{
|
| 3262 |
"epoch": 2.78,
|
| 3263 |
"learning_rate": 0.0009271752291991005,
|
| 3264 |
+
"loss": 1.8377,
|
| 3265 |
"step": 5360
|
| 3266 |
},
|
| 3267 |
{
|
| 3268 |
"epoch": 2.79,
|
| 3269 |
"learning_rate": 0.0009289050337311884,
|
| 3270 |
+
"loss": 1.9785,
|
| 3271 |
"step": 5370
|
| 3272 |
},
|
| 3273 |
{
|
| 3274 |
"epoch": 2.79,
|
| 3275 |
"learning_rate": 0.0009306348382632762,
|
| 3276 |
+
"loss": 1.9259,
|
| 3277 |
"step": 5380
|
| 3278 |
},
|
| 3279 |
{
|
| 3280 |
"epoch": 2.8,
|
| 3281 |
"learning_rate": 0.0009323646427953641,
|
| 3282 |
+
"loss": 1.9097,
|
| 3283 |
"step": 5390
|
| 3284 |
},
|
| 3285 |
{
|
| 3286 |
"epoch": 2.8,
|
| 3287 |
"learning_rate": 0.000934094447327452,
|
| 3288 |
+
"loss": 1.9034,
|
| 3289 |
"step": 5400
|
| 3290 |
},
|
| 3291 |
{
|
| 3292 |
"epoch": 2.81,
|
| 3293 |
"learning_rate": 0.0009358242518595399,
|
| 3294 |
+
"loss": 1.8983,
|
| 3295 |
"step": 5410
|
| 3296 |
},
|
| 3297 |
{
|
| 3298 |
"epoch": 2.81,
|
| 3299 |
"learning_rate": 0.0009375540563916278,
|
| 3300 |
+
"loss": 1.8115,
|
| 3301 |
"step": 5420
|
| 3302 |
},
|
| 3303 |
{
|
| 3304 |
"epoch": 2.82,
|
| 3305 |
"learning_rate": 0.0009392838609237156,
|
| 3306 |
+
"loss": 1.8606,
|
| 3307 |
"step": 5430
|
| 3308 |
},
|
| 3309 |
{
|
| 3310 |
"epoch": 2.82,
|
| 3311 |
"learning_rate": 0.0009410136654558035,
|
| 3312 |
+
"loss": 1.8883,
|
| 3313 |
"step": 5440
|
| 3314 |
},
|
| 3315 |
{
|
| 3316 |
"epoch": 2.83,
|
| 3317 |
"learning_rate": 0.0009427434699878914,
|
| 3318 |
+
"loss": 1.7868,
|
| 3319 |
"step": 5450
|
| 3320 |
},
|
| 3321 |
{
|
| 3322 |
"epoch": 2.83,
|
| 3323 |
"learning_rate": 0.0009444732745199793,
|
| 3324 |
+
"loss": 1.9107,
|
| 3325 |
"step": 5460
|
| 3326 |
},
|
| 3327 |
{
|
| 3328 |
"epoch": 2.84,
|
| 3329 |
"learning_rate": 0.0009462030790520671,
|
| 3330 |
+
"loss": 1.9427,
|
| 3331 |
"step": 5470
|
| 3332 |
},
|
| 3333 |
{
|
| 3334 |
"epoch": 2.84,
|
| 3335 |
"learning_rate": 0.000947932883584155,
|
| 3336 |
+
"loss": 1.8159,
|
| 3337 |
"step": 5480
|
| 3338 |
},
|
| 3339 |
{
|
| 3340 |
"epoch": 2.85,
|
| 3341 |
"learning_rate": 0.0009496626881162429,
|
| 3342 |
+
"loss": 1.909,
|
| 3343 |
"step": 5490
|
| 3344 |
},
|
| 3345 |
{
|
| 3346 |
"epoch": 2.85,
|
| 3347 |
"learning_rate": 0.0009513924926483307,
|
| 3348 |
+
"loss": 1.8716,
|
| 3349 |
"step": 5500
|
| 3350 |
},
|
| 3351 |
{
|
| 3352 |
"epoch": 2.86,
|
| 3353 |
"learning_rate": 0.0009531222971804187,
|
| 3354 |
+
"loss": 1.9116,
|
| 3355 |
"step": 5510
|
| 3356 |
},
|
| 3357 |
{
|
| 3358 |
"epoch": 2.86,
|
| 3359 |
"learning_rate": 0.0009548521017125064,
|
| 3360 |
+
"loss": 1.8846,
|
| 3361 |
"step": 5520
|
| 3362 |
},
|
| 3363 |
{
|
| 3364 |
"epoch": 2.87,
|
| 3365 |
"learning_rate": 0.0009565819062445944,
|
| 3366 |
+
"loss": 1.8776,
|
| 3367 |
"step": 5530
|
| 3368 |
},
|
| 3369 |
{
|
| 3370 |
"epoch": 2.87,
|
| 3371 |
"learning_rate": 0.0009583117107766822,
|
| 3372 |
+
"loss": 1.828,
|
| 3373 |
"step": 5540
|
| 3374 |
},
|
| 3375 |
{
|
| 3376 |
"epoch": 2.88,
|
| 3377 |
"learning_rate": 0.0009600415153087702,
|
| 3378 |
+
"loss": 1.9392,
|
| 3379 |
"step": 5550
|
| 3380 |
},
|
| 3381 |
{
|
| 3382 |
"epoch": 2.89,
|
| 3383 |
"learning_rate": 0.0009617713198408579,
|
| 3384 |
+
"loss": 1.8823,
|
| 3385 |
"step": 5560
|
| 3386 |
},
|
| 3387 |
{
|
| 3388 |
"epoch": 2.89,
|
| 3389 |
"learning_rate": 0.0009635011243729459,
|
| 3390 |
+
"loss": 1.8782,
|
| 3391 |
"step": 5570
|
| 3392 |
},
|
| 3393 |
{
|
| 3394 |
"epoch": 2.9,
|
| 3395 |
"learning_rate": 0.0009652309289050337,
|
| 3396 |
+
"loss": 1.8898,
|
| 3397 |
"step": 5580
|
| 3398 |
},
|
| 3399 |
{
|
| 3400 |
"epoch": 2.9,
|
| 3401 |
"learning_rate": 0.0009669607334371217,
|
| 3402 |
+
"loss": 1.8107,
|
| 3403 |
"step": 5590
|
| 3404 |
},
|
| 3405 |
{
|
| 3406 |
"epoch": 2.91,
|
| 3407 |
"learning_rate": 0.0009686905379692095,
|
| 3408 |
+
"loss": 1.8026,
|
| 3409 |
"step": 5600
|
| 3410 |
},
|
| 3411 |
{
|
| 3412 |
"epoch": 2.91,
|
| 3413 |
"learning_rate": 0.0009704203425012974,
|
| 3414 |
+
"loss": 1.8616,
|
| 3415 |
"step": 5610
|
| 3416 |
},
|
| 3417 |
{
|
| 3418 |
"epoch": 2.92,
|
| 3419 |
"learning_rate": 0.0009721501470333852,
|
| 3420 |
+
"loss": 1.8443,
|
| 3421 |
"step": 5620
|
| 3422 |
},
|
| 3423 |
{
|
| 3424 |
"epoch": 2.92,
|
| 3425 |
"learning_rate": 0.0009738799515654732,
|
| 3426 |
+
"loss": 1.8465,
|
| 3427 |
"step": 5630
|
| 3428 |
},
|
| 3429 |
{
|
| 3430 |
"epoch": 2.93,
|
| 3431 |
"learning_rate": 0.000975609756097561,
|
| 3432 |
+
"loss": 1.8806,
|
| 3433 |
"step": 5640
|
| 3434 |
},
|
| 3435 |
{
|
| 3436 |
"epoch": 2.93,
|
| 3437 |
"learning_rate": 0.0009773395606296489,
|
| 3438 |
+
"loss": 1.8441,
|
| 3439 |
"step": 5650
|
| 3440 |
},
|
| 3441 |
{
|
| 3442 |
"epoch": 2.94,
|
| 3443 |
"learning_rate": 0.0009790693651617367,
|
| 3444 |
+
"loss": 1.8194,
|
| 3445 |
"step": 5660
|
| 3446 |
},
|
| 3447 |
{
|
| 3448 |
"epoch": 2.94,
|
| 3449 |
"learning_rate": 0.0009807991696938246,
|
| 3450 |
+
"loss": 1.9228,
|
| 3451 |
"step": 5670
|
| 3452 |
},
|
| 3453 |
{
|
| 3454 |
"epoch": 2.95,
|
| 3455 |
"learning_rate": 0.0009825289742259125,
|
| 3456 |
+
"loss": 1.792,
|
| 3457 |
"step": 5680
|
| 3458 |
},
|
| 3459 |
{
|
| 3460 |
"epoch": 2.95,
|
| 3461 |
"learning_rate": 0.0009842587787580004,
|
| 3462 |
+
"loss": 1.8211,
|
| 3463 |
"step": 5690
|
| 3464 |
},
|
| 3465 |
{
|
| 3466 |
"epoch": 2.96,
|
| 3467 |
"learning_rate": 0.0009859885832900882,
|
| 3468 |
+
"loss": 1.81,
|
| 3469 |
"step": 5700
|
| 3470 |
},
|
| 3471 |
{
|
| 3472 |
"epoch": 2.96,
|
| 3473 |
"learning_rate": 0.0009877183878221761,
|
| 3474 |
+
"loss": 1.8862,
|
| 3475 |
"step": 5710
|
| 3476 |
},
|
| 3477 |
{
|
| 3478 |
"epoch": 2.97,
|
| 3479 |
"learning_rate": 0.000989448192354264,
|
| 3480 |
+
"loss": 1.8373,
|
| 3481 |
"step": 5720
|
| 3482 |
},
|
| 3483 |
{
|
| 3484 |
"epoch": 2.97,
|
| 3485 |
"learning_rate": 0.0009911779968863519,
|
| 3486 |
+
"loss": 1.8639,
|
| 3487 |
"step": 5730
|
| 3488 |
},
|
| 3489 |
{
|
| 3490 |
"epoch": 2.98,
|
| 3491 |
"learning_rate": 0.0009929078014184398,
|
| 3492 |
+
"loss": 1.8718,
|
| 3493 |
"step": 5740
|
| 3494 |
},
|
| 3495 |
{
|
| 3496 |
"epoch": 2.98,
|
| 3497 |
"learning_rate": 0.0009946376059505276,
|
| 3498 |
+
"loss": 1.8258,
|
| 3499 |
"step": 5750
|
| 3500 |
},
|
| 3501 |
{
|
| 3502 |
"epoch": 2.99,
|
| 3503 |
"learning_rate": 0.0009963674104826155,
|
| 3504 |
+
"loss": 1.8564,
|
| 3505 |
"step": 5760
|
| 3506 |
},
|
| 3507 |
{
|
| 3508 |
"epoch": 2.99,
|
| 3509 |
"learning_rate": 0.0009980972150147034,
|
| 3510 |
+
"loss": 1.9462,
|
| 3511 |
"step": 5770
|
| 3512 |
},
|
| 3513 |
{
|
| 3514 |
"epoch": 3.0,
|
| 3515 |
"learning_rate": 0.0009998270195467913,
|
| 3516 |
+
"loss": 1.8437,
|
| 3517 |
"step": 5780
|
| 3518 |
},
|
| 3519 |
{
|
| 3520 |
"epoch": 3.0,
|
| 3521 |
"eval_accuracy": {
|
| 3522 |
+
"accuracy": 0.1743190661478599
|
| 3523 |
},
|
| 3524 |
"eval_f1": {
|
| 3525 |
+
"f1": 0.1111482743173735
|
| 3526 |
},
|
| 3527 |
+
"eval_loss": 1.9242523908615112,
|
| 3528 |
"eval_precision": {
|
| 3529 |
+
"precision": 0.1516816437642854
|
| 3530 |
},
|
| 3531 |
"eval_recall": {
|
| 3532 |
+
"recall": 0.1887752678966561
|
| 3533 |
},
|
| 3534 |
+
"eval_runtime": 118.0197,
|
| 3535 |
+
"eval_samples_per_second": 32.664,
|
| 3536 |
+
"eval_steps_per_second": 16.336,
|
| 3537 |
"step": 5781
|
| 3538 |
}
|
| 3539 |
],
|
tmp-checkpoint-5781/training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4792
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c4e0514912b5f3d200ad2aec0814727d3064165a4d951379f5c5cc93d9c62d9
|
| 3 |
size 4792
|