Training in progress, epoch 4
Browse files
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 343230128
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3a30970205203a66044e82b0f2d859f75263755860693688948ab78c4df61a7d
|
| 3 |
size 343230128
|
tmp-checkpoint-5573/model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 343230128
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3a30970205203a66044e82b0f2d859f75263755860693688948ab78c4df61a7d
|
| 3 |
size 343230128
|
tmp-checkpoint-5573/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 686581178
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5d48912e184047ff12f593595e16ddfc1c743eaca237d092a6563b5cd10d2047
|
| 3 |
size 686581178
|
tmp-checkpoint-5573/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:ef1f82b7c27d56152be2ffc3305ebd61ca0dd521f605c40bdc495d74bde7812a
|
| 3 |
size 14244
|
tmp-checkpoint-5573/trainer_state.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
"best_model_checkpoint": "vit-base-patch16-224-in21k-crack-detectorVITmain50epochs\\checkpoint-5573",
|
| 4 |
"epoch": 4.999327203408836,
|
| 5 |
"eval_steps": 500,
|
|
@@ -11,3443 +11,3443 @@
|
|
| 11 |
{
|
| 12 |
"epoch": 0.01,
|
| 13 |
"learning_rate": 8.976660682226213e-08,
|
| 14 |
-
"loss": 1.
|
| 15 |
"step": 10
|
| 16 |
},
|
| 17 |
{
|
| 18 |
"epoch": 0.02,
|
| 19 |
"learning_rate": 1.7953321364452426e-07,
|
| 20 |
-
"loss": 1.
|
| 21 |
"step": 20
|
| 22 |
},
|
| 23 |
{
|
| 24 |
"epoch": 0.03,
|
| 25 |
"learning_rate": 2.692998204667864e-07,
|
| 26 |
-
"loss": 1.
|
| 27 |
"step": 30
|
| 28 |
},
|
| 29 |
{
|
| 30 |
"epoch": 0.04,
|
| 31 |
"learning_rate": 3.590664272890485e-07,
|
| 32 |
-
"loss": 1.
|
| 33 |
"step": 40
|
| 34 |
},
|
| 35 |
{
|
| 36 |
"epoch": 0.04,
|
| 37 |
"learning_rate": 4.4883303411131064e-07,
|
| 38 |
-
"loss": 1.
|
| 39 |
"step": 50
|
| 40 |
},
|
| 41 |
{
|
| 42 |
"epoch": 0.05,
|
| 43 |
"learning_rate": 5.385996409335728e-07,
|
| 44 |
-
"loss": 1.
|
| 45 |
"step": 60
|
| 46 |
},
|
| 47 |
{
|
| 48 |
"epoch": 0.06,
|
| 49 |
"learning_rate": 6.283662477558349e-07,
|
| 50 |
-
"loss": 1.
|
| 51 |
"step": 70
|
| 52 |
},
|
| 53 |
{
|
| 54 |
"epoch": 0.07,
|
| 55 |
"learning_rate": 7.18132854578097e-07,
|
| 56 |
-
"loss": 1.
|
| 57 |
"step": 80
|
| 58 |
},
|
| 59 |
{
|
| 60 |
"epoch": 0.08,
|
| 61 |
"learning_rate": 8.078994614003591e-07,
|
| 62 |
-
"loss": 1.
|
| 63 |
"step": 90
|
| 64 |
},
|
| 65 |
{
|
| 66 |
"epoch": 0.09,
|
| 67 |
"learning_rate": 8.976660682226213e-07,
|
| 68 |
-
"loss": 1.
|
| 69 |
"step": 100
|
| 70 |
},
|
| 71 |
{
|
| 72 |
"epoch": 0.1,
|
| 73 |
"learning_rate": 9.874326750448833e-07,
|
| 74 |
-
"loss": 1.
|
| 75 |
"step": 110
|
| 76 |
},
|
| 77 |
{
|
| 78 |
"epoch": 0.11,
|
| 79 |
"learning_rate": 1.0771992818671456e-06,
|
| 80 |
-
"loss": 1.
|
| 81 |
"step": 120
|
| 82 |
},
|
| 83 |
{
|
| 84 |
"epoch": 0.12,
|
| 85 |
"learning_rate": 1.1669658886894075e-06,
|
| 86 |
-
"loss": 1.
|
| 87 |
"step": 130
|
| 88 |
},
|
| 89 |
{
|
| 90 |
"epoch": 0.13,
|
| 91 |
"learning_rate": 1.2567324955116697e-06,
|
| 92 |
-
"loss": 1.
|
| 93 |
"step": 140
|
| 94 |
},
|
| 95 |
{
|
| 96 |
"epoch": 0.13,
|
| 97 |
"learning_rate": 1.3464991023339318e-06,
|
| 98 |
-
"loss": 1.
|
| 99 |
"step": 150
|
| 100 |
},
|
| 101 |
{
|
| 102 |
"epoch": 0.14,
|
| 103 |
"learning_rate": 1.436265709156194e-06,
|
| 104 |
-
"loss": 1.
|
| 105 |
"step": 160
|
| 106 |
},
|
| 107 |
{
|
| 108 |
"epoch": 0.15,
|
| 109 |
"learning_rate": 1.5260323159784561e-06,
|
| 110 |
-
"loss": 1.
|
| 111 |
"step": 170
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"epoch": 0.16,
|
| 115 |
"learning_rate": 1.6157989228007182e-06,
|
| 116 |
-
"loss": 1.
|
| 117 |
"step": 180
|
| 118 |
},
|
| 119 |
{
|
| 120 |
"epoch": 0.17,
|
| 121 |
"learning_rate": 1.7055655296229805e-06,
|
| 122 |
-
"loss": 1.
|
| 123 |
"step": 190
|
| 124 |
},
|
| 125 |
{
|
| 126 |
"epoch": 0.18,
|
| 127 |
"learning_rate": 1.7953321364452425e-06,
|
| 128 |
-
"loss": 1.
|
| 129 |
"step": 200
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.19,
|
| 133 |
"learning_rate": 1.8850987432675046e-06,
|
| 134 |
-
"loss": 1.
|
| 135 |
"step": 210
|
| 136 |
},
|
| 137 |
{
|
| 138 |
"epoch": 0.2,
|
| 139 |
"learning_rate": 1.9748653500897667e-06,
|
| 140 |
-
"loss": 1.
|
| 141 |
"step": 220
|
| 142 |
},
|
| 143 |
{
|
| 144 |
"epoch": 0.21,
|
| 145 |
"learning_rate": 2.064631956912029e-06,
|
| 146 |
-
"loss": 1.
|
| 147 |
"step": 230
|
| 148 |
},
|
| 149 |
{
|
| 150 |
"epoch": 0.22,
|
| 151 |
"learning_rate": 2.1543985637342912e-06,
|
| 152 |
-
"loss": 1.
|
| 153 |
"step": 240
|
| 154 |
},
|
| 155 |
{
|
| 156 |
"epoch": 0.22,
|
| 157 |
"learning_rate": 2.244165170556553e-06,
|
| 158 |
-
"loss": 1.
|
| 159 |
"step": 250
|
| 160 |
},
|
| 161 |
{
|
| 162 |
"epoch": 0.23,
|
| 163 |
"learning_rate": 2.333931777378815e-06,
|
| 164 |
-
"loss": 1.
|
| 165 |
"step": 260
|
| 166 |
},
|
| 167 |
{
|
| 168 |
"epoch": 0.24,
|
| 169 |
"learning_rate": 2.423698384201077e-06,
|
| 170 |
-
"loss": 1.
|
| 171 |
"step": 270
|
| 172 |
},
|
| 173 |
{
|
| 174 |
"epoch": 0.25,
|
| 175 |
"learning_rate": 2.5134649910233395e-06,
|
| 176 |
-
"loss": 1.
|
| 177 |
"step": 280
|
| 178 |
},
|
| 179 |
{
|
| 180 |
"epoch": 0.26,
|
| 181 |
"learning_rate": 2.6032315978456015e-06,
|
| 182 |
-
"loss": 1.
|
| 183 |
"step": 290
|
| 184 |
},
|
| 185 |
{
|
| 186 |
"epoch": 0.27,
|
| 187 |
"learning_rate": 2.6929982046678636e-06,
|
| 188 |
-
"loss": 1.
|
| 189 |
"step": 300
|
| 190 |
},
|
| 191 |
{
|
| 192 |
"epoch": 0.28,
|
| 193 |
"learning_rate": 2.7827648114901257e-06,
|
| 194 |
-
"loss": 1.
|
| 195 |
"step": 310
|
| 196 |
},
|
| 197 |
{
|
| 198 |
"epoch": 0.29,
|
| 199 |
"learning_rate": 2.872531418312388e-06,
|
| 200 |
-
"loss": 1.
|
| 201 |
"step": 320
|
| 202 |
},
|
| 203 |
{
|
| 204 |
"epoch": 0.3,
|
| 205 |
"learning_rate": 2.9622980251346502e-06,
|
| 206 |
-
"loss": 1.
|
| 207 |
"step": 330
|
| 208 |
},
|
| 209 |
{
|
| 210 |
"epoch": 0.31,
|
| 211 |
"learning_rate": 3.0520646319569123e-06,
|
| 212 |
-
"loss": 1.
|
| 213 |
"step": 340
|
| 214 |
},
|
| 215 |
{
|
| 216 |
"epoch": 0.31,
|
| 217 |
"learning_rate": 3.1418312387791743e-06,
|
| 218 |
-
"loss": 1.
|
| 219 |
"step": 350
|
| 220 |
},
|
| 221 |
{
|
| 222 |
"epoch": 0.32,
|
| 223 |
"learning_rate": 3.2315978456014364e-06,
|
| 224 |
-
"loss": 1.
|
| 225 |
"step": 360
|
| 226 |
},
|
| 227 |
{
|
| 228 |
"epoch": 0.33,
|
| 229 |
"learning_rate": 3.321364452423698e-06,
|
| 230 |
-
"loss": 1.
|
| 231 |
"step": 370
|
| 232 |
},
|
| 233 |
{
|
| 234 |
"epoch": 0.34,
|
| 235 |
"learning_rate": 3.411131059245961e-06,
|
| 236 |
-
"loss": 1.
|
| 237 |
"step": 380
|
| 238 |
},
|
| 239 |
{
|
| 240 |
"epoch": 0.35,
|
| 241 |
"learning_rate": 3.500897666068223e-06,
|
| 242 |
-
"loss": 1.
|
| 243 |
"step": 390
|
| 244 |
},
|
| 245 |
{
|
| 246 |
"epoch": 0.36,
|
| 247 |
"learning_rate": 3.590664272890485e-06,
|
| 248 |
-
"loss": 1.
|
| 249 |
"step": 400
|
| 250 |
},
|
| 251 |
{
|
| 252 |
"epoch": 0.37,
|
| 253 |
"learning_rate": 3.680430879712747e-06,
|
| 254 |
-
"loss": 1.
|
| 255 |
"step": 410
|
| 256 |
},
|
| 257 |
{
|
| 258 |
"epoch": 0.38,
|
| 259 |
"learning_rate": 3.770197486535009e-06,
|
| 260 |
-
"loss": 1.
|
| 261 |
"step": 420
|
| 262 |
},
|
| 263 |
{
|
| 264 |
"epoch": 0.39,
|
| 265 |
"learning_rate": 3.859964093357271e-06,
|
| 266 |
-
"loss": 1.
|
| 267 |
"step": 430
|
| 268 |
},
|
| 269 |
{
|
| 270 |
"epoch": 0.39,
|
| 271 |
"learning_rate": 3.949730700179533e-06,
|
| 272 |
-
"loss": 1.
|
| 273 |
"step": 440
|
| 274 |
},
|
| 275 |
{
|
| 276 |
"epoch": 0.4,
|
| 277 |
"learning_rate": 4.039497307001795e-06,
|
| 278 |
-
"loss": 1.
|
| 279 |
"step": 450
|
| 280 |
},
|
| 281 |
{
|
| 282 |
"epoch": 0.41,
|
| 283 |
"learning_rate": 4.129263913824058e-06,
|
| 284 |
-
"loss": 1.
|
| 285 |
"step": 460
|
| 286 |
},
|
| 287 |
{
|
| 288 |
"epoch": 0.42,
|
| 289 |
"learning_rate": 4.21903052064632e-06,
|
| 290 |
-
"loss":
|
| 291 |
"step": 470
|
| 292 |
},
|
| 293 |
{
|
| 294 |
"epoch": 0.43,
|
| 295 |
"learning_rate": 4.3087971274685824e-06,
|
| 296 |
-
"loss": 0.
|
| 297 |
"step": 480
|
| 298 |
},
|
| 299 |
{
|
| 300 |
"epoch": 0.44,
|
| 301 |
"learning_rate": 4.3985637342908445e-06,
|
| 302 |
-
"loss": 0.
|
| 303 |
"step": 490
|
| 304 |
},
|
| 305 |
{
|
| 306 |
"epoch": 0.45,
|
| 307 |
"learning_rate": 4.488330341113106e-06,
|
| 308 |
-
"loss": 0.
|
| 309 |
"step": 500
|
| 310 |
},
|
| 311 |
{
|
| 312 |
"epoch": 0.46,
|
| 313 |
"learning_rate": 4.578096947935368e-06,
|
| 314 |
-
"loss": 0.
|
| 315 |
"step": 510
|
| 316 |
},
|
| 317 |
{
|
| 318 |
"epoch": 0.47,
|
| 319 |
"learning_rate": 4.66786355475763e-06,
|
| 320 |
-
"loss": 0.
|
| 321 |
"step": 520
|
| 322 |
},
|
| 323 |
{
|
| 324 |
"epoch": 0.48,
|
| 325 |
"learning_rate": 4.757630161579892e-06,
|
| 326 |
-
"loss": 0.
|
| 327 |
"step": 530
|
| 328 |
},
|
| 329 |
{
|
| 330 |
"epoch": 0.48,
|
| 331 |
"learning_rate": 4.847396768402154e-06,
|
| 332 |
-
"loss": 0.
|
| 333 |
"step": 540
|
| 334 |
},
|
| 335 |
{
|
| 336 |
"epoch": 0.49,
|
| 337 |
"learning_rate": 4.937163375224417e-06,
|
| 338 |
-
"loss": 0.
|
| 339 |
"step": 550
|
| 340 |
},
|
| 341 |
{
|
| 342 |
"epoch": 0.5,
|
| 343 |
"learning_rate": 5.026929982046679e-06,
|
| 344 |
-
"loss": 0.
|
| 345 |
"step": 560
|
| 346 |
},
|
| 347 |
{
|
| 348 |
"epoch": 0.51,
|
| 349 |
"learning_rate": 5.116696588868941e-06,
|
| 350 |
-
"loss": 0.
|
| 351 |
"step": 570
|
| 352 |
},
|
| 353 |
{
|
| 354 |
"epoch": 0.52,
|
| 355 |
"learning_rate": 5.206463195691203e-06,
|
| 356 |
-
"loss": 0.
|
| 357 |
"step": 580
|
| 358 |
},
|
| 359 |
{
|
| 360 |
"epoch": 0.53,
|
| 361 |
"learning_rate": 5.296229802513465e-06,
|
| 362 |
-
"loss": 0.
|
| 363 |
"step": 590
|
| 364 |
},
|
| 365 |
{
|
| 366 |
"epoch": 0.54,
|
| 367 |
"learning_rate": 5.385996409335727e-06,
|
| 368 |
-
"loss": 0.
|
| 369 |
"step": 600
|
| 370 |
},
|
| 371 |
{
|
| 372 |
"epoch": 0.55,
|
| 373 |
"learning_rate": 5.475763016157989e-06,
|
| 374 |
-
"loss": 0.
|
| 375 |
"step": 610
|
| 376 |
},
|
| 377 |
{
|
| 378 |
"epoch": 0.56,
|
| 379 |
"learning_rate": 5.565529622980251e-06,
|
| 380 |
-
"loss": 0.
|
| 381 |
"step": 620
|
| 382 |
},
|
| 383 |
{
|
| 384 |
"epoch": 0.57,
|
| 385 |
"learning_rate": 5.655296229802514e-06,
|
| 386 |
-
"loss": 0.
|
| 387 |
"step": 630
|
| 388 |
},
|
| 389 |
{
|
| 390 |
"epoch": 0.57,
|
| 391 |
"learning_rate": 5.745062836624776e-06,
|
| 392 |
-
"loss": 0.
|
| 393 |
"step": 640
|
| 394 |
},
|
| 395 |
{
|
| 396 |
"epoch": 0.58,
|
| 397 |
"learning_rate": 5.834829443447038e-06,
|
| 398 |
-
"loss": 0.
|
| 399 |
"step": 650
|
| 400 |
},
|
| 401 |
{
|
| 402 |
"epoch": 0.59,
|
| 403 |
"learning_rate": 5.9245960502693004e-06,
|
| 404 |
-
"loss": 0.
|
| 405 |
"step": 660
|
| 406 |
},
|
| 407 |
{
|
| 408 |
"epoch": 0.6,
|
| 409 |
"learning_rate": 6.0143626570915625e-06,
|
| 410 |
-
"loss": 0.
|
| 411 |
"step": 670
|
| 412 |
},
|
| 413 |
{
|
| 414 |
"epoch": 0.61,
|
| 415 |
"learning_rate": 6.1041292639138246e-06,
|
| 416 |
-
"loss": 0.
|
| 417 |
"step": 680
|
| 418 |
},
|
| 419 |
{
|
| 420 |
"epoch": 0.62,
|
| 421 |
"learning_rate": 6.193895870736087e-06,
|
| 422 |
-
"loss": 0.
|
| 423 |
"step": 690
|
| 424 |
},
|
| 425 |
{
|
| 426 |
"epoch": 0.63,
|
| 427 |
"learning_rate": 6.283662477558349e-06,
|
| 428 |
-
"loss": 0.
|
| 429 |
"step": 700
|
| 430 |
},
|
| 431 |
{
|
| 432 |
"epoch": 0.64,
|
| 433 |
"learning_rate": 6.373429084380611e-06,
|
| 434 |
-
"loss": 0.
|
| 435 |
"step": 710
|
| 436 |
},
|
| 437 |
{
|
| 438 |
"epoch": 0.65,
|
| 439 |
"learning_rate": 6.463195691202873e-06,
|
| 440 |
-
"loss": 0.
|
| 441 |
"step": 720
|
| 442 |
},
|
| 443 |
{
|
| 444 |
"epoch": 0.65,
|
| 445 |
"learning_rate": 6.552962298025135e-06,
|
| 446 |
-
"loss": 0.
|
| 447 |
"step": 730
|
| 448 |
},
|
| 449 |
{
|
| 450 |
"epoch": 0.66,
|
| 451 |
"learning_rate": 6.642728904847396e-06,
|
| 452 |
-
"loss": 0.
|
| 453 |
"step": 740
|
| 454 |
},
|
| 455 |
{
|
| 456 |
"epoch": 0.67,
|
| 457 |
"learning_rate": 6.732495511669659e-06,
|
| 458 |
-
"loss": 0.
|
| 459 |
"step": 750
|
| 460 |
},
|
| 461 |
{
|
| 462 |
"epoch": 0.68,
|
| 463 |
"learning_rate": 6.822262118491922e-06,
|
| 464 |
-
"loss": 0.
|
| 465 |
"step": 760
|
| 466 |
},
|
| 467 |
{
|
| 468 |
"epoch": 0.69,
|
| 469 |
"learning_rate": 6.912028725314183e-06,
|
| 470 |
-
"loss": 0.
|
| 471 |
"step": 770
|
| 472 |
},
|
| 473 |
{
|
| 474 |
"epoch": 0.7,
|
| 475 |
"learning_rate": 7.001795332136446e-06,
|
| 476 |
-
"loss": 0.
|
| 477 |
"step": 780
|
| 478 |
},
|
| 479 |
{
|
| 480 |
"epoch": 0.71,
|
| 481 |
"learning_rate": 7.091561938958707e-06,
|
| 482 |
-
"loss": 0.
|
| 483 |
"step": 790
|
| 484 |
},
|
| 485 |
{
|
| 486 |
"epoch": 0.72,
|
| 487 |
"learning_rate": 7.18132854578097e-06,
|
| 488 |
-
"loss": 0.
|
| 489 |
"step": 800
|
| 490 |
},
|
| 491 |
{
|
| 492 |
"epoch": 0.73,
|
| 493 |
"learning_rate": 7.271095152603231e-06,
|
| 494 |
-
"loss": 0.
|
| 495 |
"step": 810
|
| 496 |
},
|
| 497 |
{
|
| 498 |
"epoch": 0.74,
|
| 499 |
"learning_rate": 7.360861759425494e-06,
|
| 500 |
-
"loss": 0.
|
| 501 |
"step": 820
|
| 502 |
},
|
| 503 |
{
|
| 504 |
"epoch": 0.74,
|
| 505 |
"learning_rate": 7.4506283662477555e-06,
|
| 506 |
-
"loss": 0.
|
| 507 |
"step": 830
|
| 508 |
},
|
| 509 |
{
|
| 510 |
"epoch": 0.75,
|
| 511 |
"learning_rate": 7.540394973070018e-06,
|
| 512 |
-
"loss": 0.
|
| 513 |
"step": 840
|
| 514 |
},
|
| 515 |
{
|
| 516 |
"epoch": 0.76,
|
| 517 |
"learning_rate": 7.630161579892281e-06,
|
| 518 |
-
"loss": 0.
|
| 519 |
"step": 850
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.77,
|
| 523 |
"learning_rate": 7.719928186714543e-06,
|
| 524 |
-
"loss": 0.
|
| 525 |
"step": 860
|
| 526 |
},
|
| 527 |
{
|
| 528 |
"epoch": 0.78,
|
| 529 |
"learning_rate": 7.809694793536805e-06,
|
| 530 |
-
"loss": 0.
|
| 531 |
"step": 870
|
| 532 |
},
|
| 533 |
{
|
| 534 |
"epoch": 0.79,
|
| 535 |
"learning_rate": 7.899461400359067e-06,
|
| 536 |
-
"loss": 0.
|
| 537 |
"step": 880
|
| 538 |
},
|
| 539 |
{
|
| 540 |
"epoch": 0.8,
|
| 541 |
"learning_rate": 7.98922800718133e-06,
|
| 542 |
-
"loss": 0.
|
| 543 |
"step": 890
|
| 544 |
},
|
| 545 |
{
|
| 546 |
"epoch": 0.81,
|
| 547 |
"learning_rate": 8.07899461400359e-06,
|
| 548 |
-
"loss": 0.
|
| 549 |
"step": 900
|
| 550 |
},
|
| 551 |
{
|
| 552 |
"epoch": 0.82,
|
| 553 |
"learning_rate": 8.168761220825854e-06,
|
| 554 |
-
"loss": 0.
|
| 555 |
"step": 910
|
| 556 |
},
|
| 557 |
{
|
| 558 |
"epoch": 0.83,
|
| 559 |
"learning_rate": 8.258527827648117e-06,
|
| 560 |
-
"loss": 0.
|
| 561 |
"step": 920
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.83,
|
| 565 |
"learning_rate": 8.348294434470378e-06,
|
| 566 |
-
"loss": 0.
|
| 567 |
"step": 930
|
| 568 |
},
|
| 569 |
{
|
| 570 |
"epoch": 0.84,
|
| 571 |
"learning_rate": 8.43806104129264e-06,
|
| 572 |
-
"loss": 0.
|
| 573 |
"step": 940
|
| 574 |
},
|
| 575 |
{
|
| 576 |
"epoch": 0.85,
|
| 577 |
"learning_rate": 8.527827648114902e-06,
|
| 578 |
-
"loss": 0.
|
| 579 |
"step": 950
|
| 580 |
},
|
| 581 |
{
|
| 582 |
"epoch": 0.86,
|
| 583 |
"learning_rate": 8.617594254937165e-06,
|
| 584 |
-
"loss": 0.
|
| 585 |
"step": 960
|
| 586 |
},
|
| 587 |
{
|
| 588 |
"epoch": 0.87,
|
| 589 |
"learning_rate": 8.707360861759426e-06,
|
| 590 |
-
"loss": 0.
|
| 591 |
"step": 970
|
| 592 |
},
|
| 593 |
{
|
| 594 |
"epoch": 0.88,
|
| 595 |
"learning_rate": 8.797127468581689e-06,
|
| 596 |
-
"loss": 0.
|
| 597 |
"step": 980
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.89,
|
| 601 |
"learning_rate": 8.88689407540395e-06,
|
| 602 |
-
"loss": 0.
|
| 603 |
"step": 990
|
| 604 |
},
|
| 605 |
{
|
| 606 |
"epoch": 0.9,
|
| 607 |
"learning_rate": 8.976660682226211e-06,
|
| 608 |
-
"loss": 0.
|
| 609 |
"step": 1000
|
| 610 |
},
|
| 611 |
{
|
| 612 |
"epoch": 0.91,
|
| 613 |
"learning_rate": 9.066427289048474e-06,
|
| 614 |
-
"loss": 0.
|
| 615 |
"step": 1010
|
| 616 |
},
|
| 617 |
{
|
| 618 |
"epoch": 0.92,
|
| 619 |
"learning_rate": 9.156193895870736e-06,
|
| 620 |
-
"loss": 0.
|
| 621 |
"step": 1020
|
| 622 |
},
|
| 623 |
{
|
| 624 |
"epoch": 0.92,
|
| 625 |
"learning_rate": 9.245960502692998e-06,
|
| 626 |
-
"loss": 0.
|
| 627 |
"step": 1030
|
| 628 |
},
|
| 629 |
{
|
| 630 |
"epoch": 0.93,
|
| 631 |
"learning_rate": 9.33572710951526e-06,
|
| 632 |
-
"loss": 0.
|
| 633 |
"step": 1040
|
| 634 |
},
|
| 635 |
{
|
| 636 |
"epoch": 0.94,
|
| 637 |
"learning_rate": 9.425493716337523e-06,
|
| 638 |
-
"loss": 0.
|
| 639 |
"step": 1050
|
| 640 |
},
|
| 641 |
{
|
| 642 |
"epoch": 0.95,
|
| 643 |
"learning_rate": 9.515260323159784e-06,
|
| 644 |
-
"loss": 0.
|
| 645 |
"step": 1060
|
| 646 |
},
|
| 647 |
{
|
| 648 |
"epoch": 0.96,
|
| 649 |
"learning_rate": 9.605026929982047e-06,
|
| 650 |
-
"loss": 0.
|
| 651 |
"step": 1070
|
| 652 |
},
|
| 653 |
{
|
| 654 |
"epoch": 0.97,
|
| 655 |
"learning_rate": 9.694793536804308e-06,
|
| 656 |
-
"loss": 0.
|
| 657 |
"step": 1080
|
| 658 |
},
|
| 659 |
{
|
| 660 |
"epoch": 0.98,
|
| 661 |
"learning_rate": 9.784560143626571e-06,
|
| 662 |
-
"loss": 0.
|
| 663 |
"step": 1090
|
| 664 |
},
|
| 665 |
{
|
| 666 |
"epoch": 0.99,
|
| 667 |
"learning_rate": 9.874326750448834e-06,
|
| 668 |
-
"loss": 0.
|
| 669 |
"step": 1100
|
| 670 |
},
|
| 671 |
{
|
| 672 |
"epoch": 1.0,
|
| 673 |
"learning_rate": 9.964093357271095e-06,
|
| 674 |
-
"loss": 0.
|
| 675 |
"step": 1110
|
| 676 |
},
|
| 677 |
{
|
| 678 |
"epoch": 1.0,
|
| 679 |
"eval_accuracy": {
|
| 680 |
-
"accuracy": 0.
|
| 681 |
},
|
| 682 |
"eval_f1": {
|
| 683 |
-
"f1": 0.
|
| 684 |
},
|
| 685 |
-
"eval_loss": 0.
|
| 686 |
"eval_precision": {
|
| 687 |
-
"precision": 0.
|
| 688 |
},
|
| 689 |
"eval_recall": {
|
| 690 |
-
"recall": 0.
|
| 691 |
},
|
| 692 |
-
"eval_runtime":
|
| 693 |
-
"eval_samples_per_second":
|
| 694 |
-
"eval_steps_per_second": 4.
|
| 695 |
"step": 1114
|
| 696 |
},
|
| 697 |
{
|
| 698 |
"epoch": 1.0,
|
| 699 |
"learning_rate": 1.0053859964093358e-05,
|
| 700 |
-
"loss": 0.
|
| 701 |
"step": 1120
|
| 702 |
},
|
| 703 |
{
|
| 704 |
"epoch": 1.01,
|
| 705 |
"learning_rate": 1.0143626570915619e-05,
|
| 706 |
-
"loss": 0.
|
| 707 |
"step": 1130
|
| 708 |
},
|
| 709 |
{
|
| 710 |
"epoch": 1.02,
|
| 711 |
"learning_rate": 1.0233393177737882e-05,
|
| 712 |
-
"loss": 0.
|
| 713 |
"step": 1140
|
| 714 |
},
|
| 715 |
{
|
| 716 |
"epoch": 1.03,
|
| 717 |
"learning_rate": 1.0323159784560143e-05,
|
| 718 |
-
"loss": 0.
|
| 719 |
"step": 1150
|
| 720 |
},
|
| 721 |
{
|
| 722 |
"epoch": 1.04,
|
| 723 |
"learning_rate": 1.0412926391382406e-05,
|
| 724 |
-
"loss": 0.
|
| 725 |
"step": 1160
|
| 726 |
},
|
| 727 |
{
|
| 728 |
"epoch": 1.05,
|
| 729 |
"learning_rate": 1.0502692998204669e-05,
|
| 730 |
-
"loss": 0.
|
| 731 |
"step": 1170
|
| 732 |
},
|
| 733 |
{
|
| 734 |
"epoch": 1.06,
|
| 735 |
"learning_rate": 1.059245960502693e-05,
|
| 736 |
-
"loss": 0.
|
| 737 |
"step": 1180
|
| 738 |
},
|
| 739 |
{
|
| 740 |
"epoch": 1.07,
|
| 741 |
"learning_rate": 1.0682226211849193e-05,
|
| 742 |
-
"loss": 0.
|
| 743 |
"step": 1190
|
| 744 |
},
|
| 745 |
{
|
| 746 |
"epoch": 1.08,
|
| 747 |
"learning_rate": 1.0771992818671454e-05,
|
| 748 |
-
"loss": 0.
|
| 749 |
"step": 1200
|
| 750 |
},
|
| 751 |
{
|
| 752 |
"epoch": 1.09,
|
| 753 |
"learning_rate": 1.0861759425493717e-05,
|
| 754 |
-
"loss": 0.
|
| 755 |
"step": 1210
|
| 756 |
},
|
| 757 |
{
|
| 758 |
"epoch": 1.09,
|
| 759 |
"learning_rate": 1.0951526032315979e-05,
|
| 760 |
-
"loss": 0.
|
| 761 |
"step": 1220
|
| 762 |
},
|
| 763 |
{
|
| 764 |
"epoch": 1.1,
|
| 765 |
"learning_rate": 1.1041292639138241e-05,
|
| 766 |
-
"loss": 0.
|
| 767 |
"step": 1230
|
| 768 |
},
|
| 769 |
{
|
| 770 |
"epoch": 1.11,
|
| 771 |
"learning_rate": 1.1131059245960503e-05,
|
| 772 |
-
"loss": 0.
|
| 773 |
"step": 1240
|
| 774 |
},
|
| 775 |
{
|
| 776 |
"epoch": 1.12,
|
| 777 |
"learning_rate": 1.1220825852782766e-05,
|
| 778 |
-
"loss": 0.
|
| 779 |
"step": 1250
|
| 780 |
},
|
| 781 |
{
|
| 782 |
"epoch": 1.13,
|
| 783 |
"learning_rate": 1.1310592459605028e-05,
|
| 784 |
-
"loss": 0.
|
| 785 |
"step": 1260
|
| 786 |
},
|
| 787 |
{
|
| 788 |
"epoch": 1.14,
|
| 789 |
"learning_rate": 1.140035906642729e-05,
|
| 790 |
-
"loss": 0.
|
| 791 |
"step": 1270
|
| 792 |
},
|
| 793 |
{
|
| 794 |
"epoch": 1.15,
|
| 795 |
"learning_rate": 1.1490125673249553e-05,
|
| 796 |
-
"loss": 0.
|
| 797 |
"step": 1280
|
| 798 |
},
|
| 799 |
{
|
| 800 |
"epoch": 1.16,
|
| 801 |
"learning_rate": 1.1579892280071814e-05,
|
| 802 |
-
"loss": 0.
|
| 803 |
"step": 1290
|
| 804 |
},
|
| 805 |
{
|
| 806 |
"epoch": 1.17,
|
| 807 |
"learning_rate": 1.1669658886894077e-05,
|
| 808 |
-
"loss": 0.
|
| 809 |
"step": 1300
|
| 810 |
},
|
| 811 |
{
|
| 812 |
"epoch": 1.18,
|
| 813 |
"learning_rate": 1.1759425493716338e-05,
|
| 814 |
-
"loss": 0.
|
| 815 |
"step": 1310
|
| 816 |
},
|
| 817 |
{
|
| 818 |
"epoch": 1.18,
|
| 819 |
"learning_rate": 1.1849192100538601e-05,
|
| 820 |
-
"loss": 0.
|
| 821 |
"step": 1320
|
| 822 |
},
|
| 823 |
{
|
| 824 |
"epoch": 1.19,
|
| 825 |
"learning_rate": 1.1938958707360862e-05,
|
| 826 |
-
"loss": 0.
|
| 827 |
"step": 1330
|
| 828 |
},
|
| 829 |
{
|
| 830 |
"epoch": 1.2,
|
| 831 |
"learning_rate": 1.2028725314183125e-05,
|
| 832 |
-
"loss": 0.
|
| 833 |
"step": 1340
|
| 834 |
},
|
| 835 |
{
|
| 836 |
"epoch": 1.21,
|
| 837 |
"learning_rate": 1.2118491921005388e-05,
|
| 838 |
-
"loss": 0.
|
| 839 |
"step": 1350
|
| 840 |
},
|
| 841 |
{
|
| 842 |
"epoch": 1.22,
|
| 843 |
"learning_rate": 1.2208258527827649e-05,
|
| 844 |
-
"loss": 0.
|
| 845 |
"step": 1360
|
| 846 |
},
|
| 847 |
{
|
| 848 |
"epoch": 1.23,
|
| 849 |
"learning_rate": 1.2298025134649912e-05,
|
| 850 |
-
"loss": 0.
|
| 851 |
"step": 1370
|
| 852 |
},
|
| 853 |
{
|
| 854 |
"epoch": 1.24,
|
| 855 |
"learning_rate": 1.2387791741472173e-05,
|
| 856 |
-
"loss": 0.
|
| 857 |
"step": 1380
|
| 858 |
},
|
| 859 |
{
|
| 860 |
"epoch": 1.25,
|
| 861 |
"learning_rate": 1.2477558348294434e-05,
|
| 862 |
-
"loss": 0.
|
| 863 |
"step": 1390
|
| 864 |
},
|
| 865 |
{
|
| 866 |
"epoch": 1.26,
|
| 867 |
"learning_rate": 1.2567324955116697e-05,
|
| 868 |
-
"loss": 0.
|
| 869 |
"step": 1400
|
| 870 |
},
|
| 871 |
{
|
| 872 |
"epoch": 1.26,
|
| 873 |
"learning_rate": 1.2657091561938959e-05,
|
| 874 |
-
"loss": 0.
|
| 875 |
"step": 1410
|
| 876 |
},
|
| 877 |
{
|
| 878 |
"epoch": 1.27,
|
| 879 |
"learning_rate": 1.2746858168761221e-05,
|
| 880 |
-
"loss": 0.
|
| 881 |
"step": 1420
|
| 882 |
},
|
| 883 |
{
|
| 884 |
"epoch": 1.28,
|
| 885 |
"learning_rate": 1.2836624775583483e-05,
|
| 886 |
-
"loss": 0.
|
| 887 |
"step": 1430
|
| 888 |
},
|
| 889 |
{
|
| 890 |
"epoch": 1.29,
|
| 891 |
"learning_rate": 1.2926391382405746e-05,
|
| 892 |
-
"loss": 0.
|
| 893 |
"step": 1440
|
| 894 |
},
|
| 895 |
{
|
| 896 |
"epoch": 1.3,
|
| 897 |
"learning_rate": 1.3016157989228009e-05,
|
| 898 |
-
"loss": 0.
|
| 899 |
"step": 1450
|
| 900 |
},
|
| 901 |
{
|
| 902 |
"epoch": 1.31,
|
| 903 |
"learning_rate": 1.310592459605027e-05,
|
| 904 |
-
"loss": 0.
|
| 905 |
"step": 1460
|
| 906 |
},
|
| 907 |
{
|
| 908 |
"epoch": 1.32,
|
| 909 |
"learning_rate": 1.3195691202872531e-05,
|
| 910 |
-
"loss": 0.
|
| 911 |
"step": 1470
|
| 912 |
},
|
| 913 |
{
|
| 914 |
"epoch": 1.33,
|
| 915 |
"learning_rate": 1.3285457809694792e-05,
|
| 916 |
-
"loss": 0.
|
| 917 |
"step": 1480
|
| 918 |
},
|
| 919 |
{
|
| 920 |
"epoch": 1.34,
|
| 921 |
"learning_rate": 1.3375224416517057e-05,
|
| 922 |
-
"loss": 0.
|
| 923 |
"step": 1490
|
| 924 |
},
|
| 925 |
{
|
| 926 |
"epoch": 1.35,
|
| 927 |
"learning_rate": 1.3464991023339318e-05,
|
| 928 |
-
"loss": 0.
|
| 929 |
"step": 1500
|
| 930 |
},
|
| 931 |
{
|
| 932 |
"epoch": 1.35,
|
| 933 |
"learning_rate": 1.355475763016158e-05,
|
| 934 |
-
"loss": 0.
|
| 935 |
"step": 1510
|
| 936 |
},
|
| 937 |
{
|
| 938 |
"epoch": 1.36,
|
| 939 |
"learning_rate": 1.3644524236983844e-05,
|
| 940 |
-
"loss": 0.
|
| 941 |
"step": 1520
|
| 942 |
},
|
| 943 |
{
|
| 944 |
"epoch": 1.37,
|
| 945 |
"learning_rate": 1.3734290843806105e-05,
|
| 946 |
-
"loss": 0.
|
| 947 |
"step": 1530
|
| 948 |
},
|
| 949 |
{
|
| 950 |
"epoch": 1.38,
|
| 951 |
"learning_rate": 1.3824057450628366e-05,
|
| 952 |
-
"loss": 0.
|
| 953 |
"step": 1540
|
| 954 |
},
|
| 955 |
{
|
| 956 |
"epoch": 1.39,
|
| 957 |
"learning_rate": 1.3913824057450627e-05,
|
| 958 |
-
"loss": 0.
|
| 959 |
"step": 1550
|
| 960 |
},
|
| 961 |
{
|
| 962 |
"epoch": 1.4,
|
| 963 |
"learning_rate": 1.4003590664272892e-05,
|
| 964 |
-
"loss": 0.
|
| 965 |
"step": 1560
|
| 966 |
},
|
| 967 |
{
|
| 968 |
"epoch": 1.41,
|
| 969 |
"learning_rate": 1.4093357271095153e-05,
|
| 970 |
-
"loss": 0.
|
| 971 |
"step": 1570
|
| 972 |
},
|
| 973 |
{
|
| 974 |
"epoch": 1.42,
|
| 975 |
"learning_rate": 1.4183123877917415e-05,
|
| 976 |
-
"loss": 0.
|
| 977 |
"step": 1580
|
| 978 |
},
|
| 979 |
{
|
| 980 |
"epoch": 1.43,
|
| 981 |
"learning_rate": 1.4272890484739679e-05,
|
| 982 |
-
"loss": 0.
|
| 983 |
"step": 1590
|
| 984 |
},
|
| 985 |
{
|
| 986 |
"epoch": 1.44,
|
| 987 |
"learning_rate": 1.436265709156194e-05,
|
| 988 |
-
"loss": 0.
|
| 989 |
"step": 1600
|
| 990 |
},
|
| 991 |
{
|
| 992 |
"epoch": 1.44,
|
| 993 |
"learning_rate": 1.4452423698384202e-05,
|
| 994 |
-
"loss": 0.
|
| 995 |
"step": 1610
|
| 996 |
},
|
| 997 |
{
|
| 998 |
"epoch": 1.45,
|
| 999 |
"learning_rate": 1.4542190305206463e-05,
|
| 1000 |
-
"loss": 0.
|
| 1001 |
"step": 1620
|
| 1002 |
},
|
| 1003 |
{
|
| 1004 |
"epoch": 1.46,
|
| 1005 |
"learning_rate": 1.4631956912028727e-05,
|
| 1006 |
-
"loss": 0.
|
| 1007 |
"step": 1630
|
| 1008 |
},
|
| 1009 |
{
|
| 1010 |
"epoch": 1.47,
|
| 1011 |
"learning_rate": 1.4721723518850989e-05,
|
| 1012 |
-
"loss": 0.
|
| 1013 |
"step": 1640
|
| 1014 |
},
|
| 1015 |
{
|
| 1016 |
"epoch": 1.48,
|
| 1017 |
"learning_rate": 1.481149012567325e-05,
|
| 1018 |
-
"loss": 0.
|
| 1019 |
"step": 1650
|
| 1020 |
},
|
| 1021 |
{
|
| 1022 |
"epoch": 1.49,
|
| 1023 |
"learning_rate": 1.4901256732495511e-05,
|
| 1024 |
-
"loss": 0.
|
| 1025 |
"step": 1660
|
| 1026 |
},
|
| 1027 |
{
|
| 1028 |
"epoch": 1.5,
|
| 1029 |
"learning_rate": 1.4991023339317776e-05,
|
| 1030 |
-
"loss": 0.
|
| 1031 |
"step": 1670
|
| 1032 |
},
|
| 1033 |
{
|
| 1034 |
"epoch": 1.51,
|
| 1035 |
"learning_rate": 1.5080789946140037e-05,
|
| 1036 |
-
"loss": 0.
|
| 1037 |
"step": 1680
|
| 1038 |
},
|
| 1039 |
{
|
| 1040 |
"epoch": 1.52,
|
| 1041 |
"learning_rate": 1.5170556552962298e-05,
|
| 1042 |
-
"loss": 0.
|
| 1043 |
"step": 1690
|
| 1044 |
},
|
| 1045 |
{
|
| 1046 |
"epoch": 1.53,
|
| 1047 |
"learning_rate": 1.5260323159784563e-05,
|
| 1048 |
-
"loss": 0.
|
| 1049 |
"step": 1700
|
| 1050 |
},
|
| 1051 |
{
|
| 1052 |
"epoch": 1.53,
|
| 1053 |
"learning_rate": 1.5350089766606824e-05,
|
| 1054 |
-
"loss": 0.
|
| 1055 |
"step": 1710
|
| 1056 |
},
|
| 1057 |
{
|
| 1058 |
"epoch": 1.54,
|
| 1059 |
"learning_rate": 1.5439856373429085e-05,
|
| 1060 |
-
"loss": 0.
|
| 1061 |
"step": 1720
|
| 1062 |
},
|
| 1063 |
{
|
| 1064 |
"epoch": 1.55,
|
| 1065 |
"learning_rate": 1.5529622980251346e-05,
|
| 1066 |
-
"loss": 0.
|
| 1067 |
"step": 1730
|
| 1068 |
},
|
| 1069 |
{
|
| 1070 |
"epoch": 1.56,
|
| 1071 |
"learning_rate": 1.561938958707361e-05,
|
| 1072 |
-
"loss": 0.
|
| 1073 |
"step": 1740
|
| 1074 |
},
|
| 1075 |
{
|
| 1076 |
"epoch": 1.57,
|
| 1077 |
"learning_rate": 1.5709156193895872e-05,
|
| 1078 |
-
"loss": 0.
|
| 1079 |
"step": 1750
|
| 1080 |
},
|
| 1081 |
{
|
| 1082 |
"epoch": 1.58,
|
| 1083 |
"learning_rate": 1.5798922800718133e-05,
|
| 1084 |
-
"loss": 0.
|
| 1085 |
"step": 1760
|
| 1086 |
},
|
| 1087 |
{
|
| 1088 |
"epoch": 1.59,
|
| 1089 |
"learning_rate": 1.5888689407540398e-05,
|
| 1090 |
-
"loss": 0.
|
| 1091 |
"step": 1770
|
| 1092 |
},
|
| 1093 |
{
|
| 1094 |
"epoch": 1.6,
|
| 1095 |
"learning_rate": 1.597845601436266e-05,
|
| 1096 |
-
"loss": 0.
|
| 1097 |
"step": 1780
|
| 1098 |
},
|
| 1099 |
{
|
| 1100 |
"epoch": 1.61,
|
| 1101 |
"learning_rate": 1.606822262118492e-05,
|
| 1102 |
-
"loss": 0.
|
| 1103 |
"step": 1790
|
| 1104 |
},
|
| 1105 |
{
|
| 1106 |
"epoch": 1.61,
|
| 1107 |
"learning_rate": 1.615798922800718e-05,
|
| 1108 |
-
"loss": 0.
|
| 1109 |
"step": 1800
|
| 1110 |
},
|
| 1111 |
{
|
| 1112 |
"epoch": 1.62,
|
| 1113 |
"learning_rate": 1.6247755834829446e-05,
|
| 1114 |
-
"loss": 0.
|
| 1115 |
"step": 1810
|
| 1116 |
},
|
| 1117 |
{
|
| 1118 |
"epoch": 1.63,
|
| 1119 |
"learning_rate": 1.6337522441651707e-05,
|
| 1120 |
-
"loss": 0.
|
| 1121 |
"step": 1820
|
| 1122 |
},
|
| 1123 |
{
|
| 1124 |
"epoch": 1.64,
|
| 1125 |
"learning_rate": 1.642728904847397e-05,
|
| 1126 |
-
"loss": 0.
|
| 1127 |
"step": 1830
|
| 1128 |
},
|
| 1129 |
{
|
| 1130 |
"epoch": 1.65,
|
| 1131 |
"learning_rate": 1.6517055655296233e-05,
|
| 1132 |
-
"loss": 0.
|
| 1133 |
"step": 1840
|
| 1134 |
},
|
| 1135 |
{
|
| 1136 |
"epoch": 1.66,
|
| 1137 |
"learning_rate": 1.6606822262118494e-05,
|
| 1138 |
-
"loss": 0.
|
| 1139 |
"step": 1850
|
| 1140 |
},
|
| 1141 |
{
|
| 1142 |
"epoch": 1.67,
|
| 1143 |
"learning_rate": 1.6696588868940756e-05,
|
| 1144 |
-
"loss": 0.
|
| 1145 |
"step": 1860
|
| 1146 |
},
|
| 1147 |
{
|
| 1148 |
"epoch": 1.68,
|
| 1149 |
"learning_rate": 1.6786355475763017e-05,
|
| 1150 |
-
"loss": 0.
|
| 1151 |
"step": 1870
|
| 1152 |
},
|
| 1153 |
{
|
| 1154 |
"epoch": 1.69,
|
| 1155 |
"learning_rate": 1.687612208258528e-05,
|
| 1156 |
-
"loss": 0.
|
| 1157 |
"step": 1880
|
| 1158 |
},
|
| 1159 |
{
|
| 1160 |
"epoch": 1.7,
|
| 1161 |
"learning_rate": 1.6965888689407543e-05,
|
| 1162 |
-
"loss": 0.
|
| 1163 |
"step": 1890
|
| 1164 |
},
|
| 1165 |
{
|
| 1166 |
"epoch": 1.7,
|
| 1167 |
"learning_rate": 1.7055655296229804e-05,
|
| 1168 |
-
"loss": 0.
|
| 1169 |
"step": 1900
|
| 1170 |
},
|
| 1171 |
{
|
| 1172 |
"epoch": 1.71,
|
| 1173 |
"learning_rate": 1.7145421903052065e-05,
|
| 1174 |
-
"loss": 0.
|
| 1175 |
"step": 1910
|
| 1176 |
},
|
| 1177 |
{
|
| 1178 |
"epoch": 1.72,
|
| 1179 |
"learning_rate": 1.723518850987433e-05,
|
| 1180 |
-
"loss": 0.
|
| 1181 |
"step": 1920
|
| 1182 |
},
|
| 1183 |
{
|
| 1184 |
"epoch": 1.73,
|
| 1185 |
"learning_rate": 1.732495511669659e-05,
|
| 1186 |
-
"loss": 0.
|
| 1187 |
"step": 1930
|
| 1188 |
},
|
| 1189 |
{
|
| 1190 |
"epoch": 1.74,
|
| 1191 |
"learning_rate": 1.7414721723518852e-05,
|
| 1192 |
-
"loss": 0.
|
| 1193 |
"step": 1940
|
| 1194 |
},
|
| 1195 |
{
|
| 1196 |
"epoch": 1.75,
|
| 1197 |
"learning_rate": 1.7504488330341113e-05,
|
| 1198 |
-
"loss": 0.
|
| 1199 |
"step": 1950
|
| 1200 |
},
|
| 1201 |
{
|
| 1202 |
"epoch": 1.76,
|
| 1203 |
"learning_rate": 1.7594254937163378e-05,
|
| 1204 |
-
"loss": 0.
|
| 1205 |
"step": 1960
|
| 1206 |
},
|
| 1207 |
{
|
| 1208 |
"epoch": 1.77,
|
| 1209 |
"learning_rate": 1.768402154398564e-05,
|
| 1210 |
-
"loss": 0.
|
| 1211 |
"step": 1970
|
| 1212 |
},
|
| 1213 |
{
|
| 1214 |
"epoch": 1.78,
|
| 1215 |
"learning_rate": 1.77737881508079e-05,
|
| 1216 |
-
"loss": 0.
|
| 1217 |
"step": 1980
|
| 1218 |
},
|
| 1219 |
{
|
| 1220 |
"epoch": 1.79,
|
| 1221 |
"learning_rate": 1.786355475763016e-05,
|
| 1222 |
-
"loss": 0.
|
| 1223 |
"step": 1990
|
| 1224 |
},
|
| 1225 |
{
|
| 1226 |
"epoch": 1.79,
|
| 1227 |
"learning_rate": 1.7953321364452423e-05,
|
| 1228 |
-
"loss": 0.
|
| 1229 |
"step": 2000
|
| 1230 |
},
|
| 1231 |
{
|
| 1232 |
"epoch": 1.8,
|
| 1233 |
"learning_rate": 1.8043087971274687e-05,
|
| 1234 |
-
"loss": 0.
|
| 1235 |
"step": 2010
|
| 1236 |
},
|
| 1237 |
{
|
| 1238 |
"epoch": 1.81,
|
| 1239 |
"learning_rate": 1.813285457809695e-05,
|
| 1240 |
-
"loss": 0.
|
| 1241 |
"step": 2020
|
| 1242 |
},
|
| 1243 |
{
|
| 1244 |
"epoch": 1.82,
|
| 1245 |
"learning_rate": 1.822262118491921e-05,
|
| 1246 |
-
"loss": 0.
|
| 1247 |
"step": 2030
|
| 1248 |
},
|
| 1249 |
{
|
| 1250 |
"epoch": 1.83,
|
| 1251 |
"learning_rate": 1.831238779174147e-05,
|
| 1252 |
-
"loss": 0.
|
| 1253 |
"step": 2040
|
| 1254 |
},
|
| 1255 |
{
|
| 1256 |
"epoch": 1.84,
|
| 1257 |
"learning_rate": 1.8402154398563732e-05,
|
| 1258 |
-
"loss": 0.
|
| 1259 |
"step": 2050
|
| 1260 |
},
|
| 1261 |
{
|
| 1262 |
"epoch": 1.85,
|
| 1263 |
"learning_rate": 1.8491921005385997e-05,
|
| 1264 |
-
"loss": 0.
|
| 1265 |
"step": 2060
|
| 1266 |
},
|
| 1267 |
{
|
| 1268 |
"epoch": 1.86,
|
| 1269 |
"learning_rate": 1.8581687612208258e-05,
|
| 1270 |
-
"loss": 0.
|
| 1271 |
"step": 2070
|
| 1272 |
},
|
| 1273 |
{
|
| 1274 |
"epoch": 1.87,
|
| 1275 |
"learning_rate": 1.867145421903052e-05,
|
| 1276 |
-
"loss": 0.
|
| 1277 |
"step": 2080
|
| 1278 |
},
|
| 1279 |
{
|
| 1280 |
"epoch": 1.87,
|
| 1281 |
"learning_rate": 1.8761220825852784e-05,
|
| 1282 |
-
"loss": 0.
|
| 1283 |
"step": 2090
|
| 1284 |
},
|
| 1285 |
{
|
| 1286 |
"epoch": 1.88,
|
| 1287 |
"learning_rate": 1.8850987432675045e-05,
|
| 1288 |
-
"loss": 0.
|
| 1289 |
"step": 2100
|
| 1290 |
},
|
| 1291 |
{
|
| 1292 |
"epoch": 1.89,
|
| 1293 |
"learning_rate": 1.8940754039497306e-05,
|
| 1294 |
-
"loss": 0.
|
| 1295 |
"step": 2110
|
| 1296 |
},
|
| 1297 |
{
|
| 1298 |
"epoch": 1.9,
|
| 1299 |
"learning_rate": 1.9030520646319568e-05,
|
| 1300 |
-
"loss": 0.
|
| 1301 |
"step": 2120
|
| 1302 |
},
|
| 1303 |
{
|
| 1304 |
"epoch": 1.91,
|
| 1305 |
"learning_rate": 1.9120287253141832e-05,
|
| 1306 |
-
"loss": 0.
|
| 1307 |
"step": 2130
|
| 1308 |
},
|
| 1309 |
{
|
| 1310 |
"epoch": 1.92,
|
| 1311 |
"learning_rate": 1.9210053859964093e-05,
|
| 1312 |
-
"loss": 0.
|
| 1313 |
"step": 2140
|
| 1314 |
},
|
| 1315 |
{
|
| 1316 |
"epoch": 1.93,
|
| 1317 |
"learning_rate": 1.9299820466786355e-05,
|
| 1318 |
-
"loss": 0.
|
| 1319 |
"step": 2150
|
| 1320 |
},
|
| 1321 |
{
|
| 1322 |
"epoch": 1.94,
|
| 1323 |
"learning_rate": 1.9389587073608616e-05,
|
| 1324 |
-
"loss": 0.
|
| 1325 |
"step": 2160
|
| 1326 |
},
|
| 1327 |
{
|
| 1328 |
"epoch": 1.95,
|
| 1329 |
"learning_rate": 1.947935368043088e-05,
|
| 1330 |
-
"loss": 0.
|
| 1331 |
"step": 2170
|
| 1332 |
},
|
| 1333 |
{
|
| 1334 |
"epoch": 1.96,
|
| 1335 |
"learning_rate": 1.9569120287253142e-05,
|
| 1336 |
-
"loss": 0.
|
| 1337 |
"step": 2180
|
| 1338 |
},
|
| 1339 |
{
|
| 1340 |
"epoch": 1.96,
|
| 1341 |
"learning_rate": 1.9658886894075403e-05,
|
| 1342 |
-
"loss": 0.
|
| 1343 |
"step": 2190
|
| 1344 |
},
|
| 1345 |
{
|
| 1346 |
"epoch": 1.97,
|
| 1347 |
"learning_rate": 1.9748653500897668e-05,
|
| 1348 |
-
"loss": 0.
|
| 1349 |
"step": 2200
|
| 1350 |
},
|
| 1351 |
{
|
| 1352 |
"epoch": 1.98,
|
| 1353 |
"learning_rate": 1.983842010771993e-05,
|
| 1354 |
-
"loss": 0.
|
| 1355 |
"step": 2210
|
| 1356 |
},
|
| 1357 |
{
|
| 1358 |
"epoch": 1.99,
|
| 1359 |
"learning_rate": 1.992818671454219e-05,
|
| 1360 |
-
"loss": 0.
|
| 1361 |
"step": 2220
|
| 1362 |
},
|
| 1363 |
{
|
| 1364 |
"epoch": 2.0,
|
| 1365 |
"eval_accuracy": {
|
| 1366 |
-
"accuracy": 0.
|
| 1367 |
},
|
| 1368 |
"eval_f1": {
|
| 1369 |
-
"f1": 0.
|
| 1370 |
},
|
| 1371 |
-
"eval_loss": 0.
|
| 1372 |
"eval_precision": {
|
| 1373 |
-
"precision": 0.
|
| 1374 |
},
|
| 1375 |
"eval_recall": {
|
| 1376 |
-
"recall": 0.
|
| 1377 |
},
|
| 1378 |
-
"eval_runtime":
|
| 1379 |
-
"eval_samples_per_second":
|
| 1380 |
-
"eval_steps_per_second": 6.
|
| 1381 |
"step": 2229
|
| 1382 |
},
|
| 1383 |
{
|
| 1384 |
"epoch": 2.0,
|
| 1385 |
"learning_rate": 2.001795332136445e-05,
|
| 1386 |
-
"loss": 0.
|
| 1387 |
"step": 2230
|
| 1388 |
},
|
| 1389 |
{
|
| 1390 |
"epoch": 2.01,
|
| 1391 |
"learning_rate": 2.0107719928186716e-05,
|
| 1392 |
-
"loss": 0.
|
| 1393 |
"step": 2240
|
| 1394 |
},
|
| 1395 |
{
|
| 1396 |
"epoch": 2.02,
|
| 1397 |
"learning_rate": 2.0197486535008977e-05,
|
| 1398 |
-
"loss": 0.
|
| 1399 |
"step": 2250
|
| 1400 |
},
|
| 1401 |
{
|
| 1402 |
"epoch": 2.03,
|
| 1403 |
"learning_rate": 2.0287253141831238e-05,
|
| 1404 |
-
"loss": 0.
|
| 1405 |
"step": 2260
|
| 1406 |
},
|
| 1407 |
{
|
| 1408 |
"epoch": 2.04,
|
| 1409 |
"learning_rate": 2.0377019748653503e-05,
|
| 1410 |
-
"loss": 0.
|
| 1411 |
"step": 2270
|
| 1412 |
},
|
| 1413 |
{
|
| 1414 |
"epoch": 2.05,
|
| 1415 |
"learning_rate": 2.0466786355475764e-05,
|
| 1416 |
-
"loss": 0.
|
| 1417 |
"step": 2280
|
| 1418 |
},
|
| 1419 |
{
|
| 1420 |
"epoch": 2.05,
|
| 1421 |
"learning_rate": 2.0556552962298025e-05,
|
| 1422 |
-
"loss": 0.
|
| 1423 |
"step": 2290
|
| 1424 |
},
|
| 1425 |
{
|
| 1426 |
"epoch": 2.06,
|
| 1427 |
"learning_rate": 2.0646319569120286e-05,
|
| 1428 |
-
"loss": 0.
|
| 1429 |
"step": 2300
|
| 1430 |
},
|
| 1431 |
{
|
| 1432 |
"epoch": 2.07,
|
| 1433 |
"learning_rate": 2.073608617594255e-05,
|
| 1434 |
-
"loss": 0.
|
| 1435 |
"step": 2310
|
| 1436 |
},
|
| 1437 |
{
|
| 1438 |
"epoch": 2.08,
|
| 1439 |
"learning_rate": 2.0825852782764812e-05,
|
| 1440 |
-
"loss": 0.
|
| 1441 |
"step": 2320
|
| 1442 |
},
|
| 1443 |
{
|
| 1444 |
"epoch": 2.09,
|
| 1445 |
"learning_rate": 2.0915619389587073e-05,
|
| 1446 |
-
"loss": 0.
|
| 1447 |
"step": 2330
|
| 1448 |
},
|
| 1449 |
{
|
| 1450 |
"epoch": 2.1,
|
| 1451 |
"learning_rate": 2.1005385996409338e-05,
|
| 1452 |
-
"loss": 0.
|
| 1453 |
"step": 2340
|
| 1454 |
},
|
| 1455 |
{
|
| 1456 |
"epoch": 2.11,
|
| 1457 |
"learning_rate": 2.10951526032316e-05,
|
| 1458 |
-
"loss": 0.
|
| 1459 |
"step": 2350
|
| 1460 |
},
|
| 1461 |
{
|
| 1462 |
"epoch": 2.12,
|
| 1463 |
"learning_rate": 2.118491921005386e-05,
|
| 1464 |
-
"loss": 0.
|
| 1465 |
"step": 2360
|
| 1466 |
},
|
| 1467 |
{
|
| 1468 |
"epoch": 2.13,
|
| 1469 |
"learning_rate": 2.1274685816876122e-05,
|
| 1470 |
-
"loss": 0.
|
| 1471 |
"step": 2370
|
| 1472 |
},
|
| 1473 |
{
|
| 1474 |
"epoch": 2.14,
|
| 1475 |
"learning_rate": 2.1364452423698386e-05,
|
| 1476 |
-
"loss": 0.
|
| 1477 |
"step": 2380
|
| 1478 |
},
|
| 1479 |
{
|
| 1480 |
"epoch": 2.14,
|
| 1481 |
"learning_rate": 2.1454219030520648e-05,
|
| 1482 |
-
"loss": 0.
|
| 1483 |
"step": 2390
|
| 1484 |
},
|
| 1485 |
{
|
| 1486 |
"epoch": 2.15,
|
| 1487 |
"learning_rate": 2.154398563734291e-05,
|
| 1488 |
-
"loss": 0.
|
| 1489 |
"step": 2400
|
| 1490 |
},
|
| 1491 |
{
|
| 1492 |
"epoch": 2.16,
|
| 1493 |
"learning_rate": 2.163375224416517e-05,
|
| 1494 |
-
"loss": 0.
|
| 1495 |
"step": 2410
|
| 1496 |
},
|
| 1497 |
{
|
| 1498 |
"epoch": 2.17,
|
| 1499 |
"learning_rate": 2.1723518850987435e-05,
|
| 1500 |
-
"loss": 0.
|
| 1501 |
"step": 2420
|
| 1502 |
},
|
| 1503 |
{
|
| 1504 |
"epoch": 2.18,
|
| 1505 |
"learning_rate": 2.1813285457809696e-05,
|
| 1506 |
-
"loss": 0.
|
| 1507 |
"step": 2430
|
| 1508 |
},
|
| 1509 |
{
|
| 1510 |
"epoch": 2.19,
|
| 1511 |
"learning_rate": 2.1903052064631957e-05,
|
| 1512 |
-
"loss": 0.
|
| 1513 |
"step": 2440
|
| 1514 |
},
|
| 1515 |
{
|
| 1516 |
"epoch": 2.2,
|
| 1517 |
"learning_rate": 2.199281867145422e-05,
|
| 1518 |
-
"loss": 0.
|
| 1519 |
"step": 2450
|
| 1520 |
},
|
| 1521 |
{
|
| 1522 |
"epoch": 2.21,
|
| 1523 |
"learning_rate": 2.2082585278276483e-05,
|
| 1524 |
-
"loss": 0.
|
| 1525 |
"step": 2460
|
| 1526 |
},
|
| 1527 |
{
|
| 1528 |
"epoch": 2.22,
|
| 1529 |
"learning_rate": 2.2172351885098744e-05,
|
| 1530 |
-
"loss": 0.
|
| 1531 |
"step": 2470
|
| 1532 |
},
|
| 1533 |
{
|
| 1534 |
"epoch": 2.22,
|
| 1535 |
"learning_rate": 2.2262118491921005e-05,
|
| 1536 |
-
"loss": 0.
|
| 1537 |
"step": 2480
|
| 1538 |
},
|
| 1539 |
{
|
| 1540 |
"epoch": 2.23,
|
| 1541 |
"learning_rate": 2.235188509874327e-05,
|
| 1542 |
-
"loss": 0.
|
| 1543 |
"step": 2490
|
| 1544 |
},
|
| 1545 |
{
|
| 1546 |
"epoch": 2.24,
|
| 1547 |
"learning_rate": 2.244165170556553e-05,
|
| 1548 |
-
"loss": 0.
|
| 1549 |
"step": 2500
|
| 1550 |
},
|
| 1551 |
{
|
| 1552 |
"epoch": 2.25,
|
| 1553 |
"learning_rate": 2.2531418312387792e-05,
|
| 1554 |
-
"loss": 0.
|
| 1555 |
"step": 2510
|
| 1556 |
},
|
| 1557 |
{
|
| 1558 |
"epoch": 2.26,
|
| 1559 |
"learning_rate": 2.2621184919210057e-05,
|
| 1560 |
-
"loss": 0.
|
| 1561 |
"step": 2520
|
| 1562 |
},
|
| 1563 |
{
|
| 1564 |
"epoch": 2.27,
|
| 1565 |
"learning_rate": 2.2710951526032318e-05,
|
| 1566 |
-
"loss": 0.
|
| 1567 |
"step": 2530
|
| 1568 |
},
|
| 1569 |
{
|
| 1570 |
"epoch": 2.28,
|
| 1571 |
"learning_rate": 2.280071813285458e-05,
|
| 1572 |
-
"loss": 0.
|
| 1573 |
"step": 2540
|
| 1574 |
},
|
| 1575 |
{
|
| 1576 |
"epoch": 2.29,
|
| 1577 |
"learning_rate": 2.289048473967684e-05,
|
| 1578 |
-
"loss": 0.
|
| 1579 |
"step": 2550
|
| 1580 |
},
|
| 1581 |
{
|
| 1582 |
"epoch": 2.3,
|
| 1583 |
"learning_rate": 2.2980251346499105e-05,
|
| 1584 |
-
"loss": 0.
|
| 1585 |
"step": 2560
|
| 1586 |
},
|
| 1587 |
{
|
| 1588 |
"epoch": 2.31,
|
| 1589 |
"learning_rate": 2.3070017953321366e-05,
|
| 1590 |
-
"loss": 0.
|
| 1591 |
"step": 2570
|
| 1592 |
},
|
| 1593 |
{
|
| 1594 |
"epoch": 2.31,
|
| 1595 |
"learning_rate": 2.3159784560143628e-05,
|
| 1596 |
-
"loss": 0.
|
| 1597 |
"step": 2580
|
| 1598 |
},
|
| 1599 |
{
|
| 1600 |
"epoch": 2.32,
|
| 1601 |
"learning_rate": 2.3249551166965892e-05,
|
| 1602 |
-
"loss": 0.
|
| 1603 |
"step": 2590
|
| 1604 |
},
|
| 1605 |
{
|
| 1606 |
"epoch": 2.33,
|
| 1607 |
"learning_rate": 2.3339317773788153e-05,
|
| 1608 |
-
"loss": 0.
|
| 1609 |
"step": 2600
|
| 1610 |
},
|
| 1611 |
{
|
| 1612 |
"epoch": 2.34,
|
| 1613 |
"learning_rate": 2.3429084380610415e-05,
|
| 1614 |
-
"loss": 0.
|
| 1615 |
"step": 2610
|
| 1616 |
},
|
| 1617 |
{
|
| 1618 |
"epoch": 2.35,
|
| 1619 |
"learning_rate": 2.3518850987432676e-05,
|
| 1620 |
-
"loss": 0.
|
| 1621 |
"step": 2620
|
| 1622 |
},
|
| 1623 |
{
|
| 1624 |
"epoch": 2.36,
|
| 1625 |
"learning_rate": 2.360861759425494e-05,
|
| 1626 |
-
"loss": 0.
|
| 1627 |
"step": 2630
|
| 1628 |
},
|
| 1629 |
{
|
| 1630 |
"epoch": 2.37,
|
| 1631 |
"learning_rate": 2.3698384201077202e-05,
|
| 1632 |
-
"loss": 0.
|
| 1633 |
"step": 2640
|
| 1634 |
},
|
| 1635 |
{
|
| 1636 |
"epoch": 2.38,
|
| 1637 |
"learning_rate": 2.3788150807899463e-05,
|
| 1638 |
-
"loss": 0.
|
| 1639 |
"step": 2650
|
| 1640 |
},
|
| 1641 |
{
|
| 1642 |
"epoch": 2.39,
|
| 1643 |
"learning_rate": 2.3877917414721724e-05,
|
| 1644 |
-
"loss": 0.
|
| 1645 |
"step": 2660
|
| 1646 |
},
|
| 1647 |
{
|
| 1648 |
"epoch": 2.4,
|
| 1649 |
"learning_rate": 2.396768402154399e-05,
|
| 1650 |
-
"loss": 0.
|
| 1651 |
"step": 2670
|
| 1652 |
},
|
| 1653 |
{
|
| 1654 |
"epoch": 2.4,
|
| 1655 |
"learning_rate": 2.405745062836625e-05,
|
| 1656 |
-
"loss": 0.
|
| 1657 |
"step": 2680
|
| 1658 |
},
|
| 1659 |
{
|
| 1660 |
"epoch": 2.41,
|
| 1661 |
"learning_rate": 2.414721723518851e-05,
|
| 1662 |
-
"loss": 0.
|
| 1663 |
"step": 2690
|
| 1664 |
},
|
| 1665 |
{
|
| 1666 |
"epoch": 2.42,
|
| 1667 |
"learning_rate": 2.4236983842010776e-05,
|
| 1668 |
-
"loss": 0.
|
| 1669 |
"step": 2700
|
| 1670 |
},
|
| 1671 |
{
|
| 1672 |
"epoch": 2.43,
|
| 1673 |
"learning_rate": 2.4326750448833037e-05,
|
| 1674 |
-
"loss": 0.
|
| 1675 |
"step": 2710
|
| 1676 |
},
|
| 1677 |
{
|
| 1678 |
"epoch": 2.44,
|
| 1679 |
"learning_rate": 2.4416517055655298e-05,
|
| 1680 |
-
"loss": 0.
|
| 1681 |
"step": 2720
|
| 1682 |
},
|
| 1683 |
{
|
| 1684 |
"epoch": 2.45,
|
| 1685 |
"learning_rate": 2.450628366247756e-05,
|
| 1686 |
-
"loss": 0.
|
| 1687 |
"step": 2730
|
| 1688 |
},
|
| 1689 |
{
|
| 1690 |
"epoch": 2.46,
|
| 1691 |
"learning_rate": 2.4596050269299824e-05,
|
| 1692 |
-
"loss": 0.
|
| 1693 |
"step": 2740
|
| 1694 |
},
|
| 1695 |
{
|
| 1696 |
"epoch": 2.47,
|
| 1697 |
"learning_rate": 2.4685816876122085e-05,
|
| 1698 |
-
"loss": 0.
|
| 1699 |
"step": 2750
|
| 1700 |
},
|
| 1701 |
{
|
| 1702 |
"epoch": 2.48,
|
| 1703 |
"learning_rate": 2.4775583482944346e-05,
|
| 1704 |
-
"loss": 0.
|
| 1705 |
"step": 2760
|
| 1706 |
},
|
| 1707 |
{
|
| 1708 |
"epoch": 2.48,
|
| 1709 |
"learning_rate": 2.4865350089766608e-05,
|
| 1710 |
-
"loss": 0.
|
| 1711 |
"step": 2770
|
| 1712 |
},
|
| 1713 |
{
|
| 1714 |
"epoch": 2.49,
|
| 1715 |
"learning_rate": 2.495511669658887e-05,
|
| 1716 |
-
"loss": 0.
|
| 1717 |
"step": 2780
|
| 1718 |
},
|
| 1719 |
{
|
| 1720 |
"epoch": 2.5,
|
| 1721 |
"learning_rate": 2.5044883303411134e-05,
|
| 1722 |
-
"loss": 0.
|
| 1723 |
"step": 2790
|
| 1724 |
},
|
| 1725 |
{
|
| 1726 |
"epoch": 2.51,
|
| 1727 |
"learning_rate": 2.5134649910233395e-05,
|
| 1728 |
-
"loss": 0.
|
| 1729 |
"step": 2800
|
| 1730 |
},
|
| 1731 |
{
|
| 1732 |
"epoch": 2.52,
|
| 1733 |
"learning_rate": 2.5224416517055656e-05,
|
| 1734 |
-
"loss": 0.
|
| 1735 |
"step": 2810
|
| 1736 |
},
|
| 1737 |
{
|
| 1738 |
"epoch": 2.53,
|
| 1739 |
"learning_rate": 2.5314183123877917e-05,
|
| 1740 |
-
"loss": 0.
|
| 1741 |
"step": 2820
|
| 1742 |
},
|
| 1743 |
{
|
| 1744 |
"epoch": 2.54,
|
| 1745 |
"learning_rate": 2.5403949730700182e-05,
|
| 1746 |
-
"loss": 0.
|
| 1747 |
"step": 2830
|
| 1748 |
},
|
| 1749 |
{
|
| 1750 |
"epoch": 2.55,
|
| 1751 |
"learning_rate": 2.5493716337522443e-05,
|
| 1752 |
-
"loss": 0.
|
| 1753 |
"step": 2840
|
| 1754 |
},
|
| 1755 |
{
|
| 1756 |
"epoch": 2.56,
|
| 1757 |
"learning_rate": 2.5583482944344704e-05,
|
| 1758 |
-
"loss": 0.
|
| 1759 |
"step": 2850
|
| 1760 |
},
|
| 1761 |
{
|
| 1762 |
"epoch": 2.57,
|
| 1763 |
"learning_rate": 2.5673249551166965e-05,
|
| 1764 |
-
"loss": 0.
|
| 1765 |
"step": 2860
|
| 1766 |
},
|
| 1767 |
{
|
| 1768 |
"epoch": 2.57,
|
| 1769 |
"learning_rate": 2.5763016157989227e-05,
|
| 1770 |
-
"loss": 0.
|
| 1771 |
"step": 2870
|
| 1772 |
},
|
| 1773 |
{
|
| 1774 |
"epoch": 2.58,
|
| 1775 |
"learning_rate": 2.585278276481149e-05,
|
| 1776 |
-
"loss": 0.
|
| 1777 |
"step": 2880
|
| 1778 |
},
|
| 1779 |
{
|
| 1780 |
"epoch": 2.59,
|
| 1781 |
"learning_rate": 2.5942549371633752e-05,
|
| 1782 |
-
"loss": 0.
|
| 1783 |
"step": 2890
|
| 1784 |
},
|
| 1785 |
{
|
| 1786 |
"epoch": 2.6,
|
| 1787 |
"learning_rate": 2.6032315978456017e-05,
|
| 1788 |
-
"loss": 0.
|
| 1789 |
"step": 2900
|
| 1790 |
},
|
| 1791 |
{
|
| 1792 |
"epoch": 2.61,
|
| 1793 |
"learning_rate": 2.6122082585278278e-05,
|
| 1794 |
-
"loss": 0.
|
| 1795 |
"step": 2910
|
| 1796 |
},
|
| 1797 |
{
|
| 1798 |
"epoch": 2.62,
|
| 1799 |
"learning_rate": 2.621184919210054e-05,
|
| 1800 |
-
"loss": 0.
|
| 1801 |
"step": 2920
|
| 1802 |
},
|
| 1803 |
{
|
| 1804 |
"epoch": 2.63,
|
| 1805 |
"learning_rate": 2.63016157989228e-05,
|
| 1806 |
-
"loss": 0.
|
| 1807 |
"step": 2930
|
| 1808 |
},
|
| 1809 |
{
|
| 1810 |
"epoch": 2.64,
|
| 1811 |
"learning_rate": 2.6391382405745062e-05,
|
| 1812 |
-
"loss": 0.
|
| 1813 |
"step": 2940
|
| 1814 |
},
|
| 1815 |
{
|
| 1816 |
"epoch": 2.65,
|
| 1817 |
"learning_rate": 2.6481149012567323e-05,
|
| 1818 |
-
"loss": 0.
|
| 1819 |
"step": 2950
|
| 1820 |
},
|
| 1821 |
{
|
| 1822 |
"epoch": 2.66,
|
| 1823 |
"learning_rate": 2.6570915619389584e-05,
|
| 1824 |
-
"loss": 0.
|
| 1825 |
"step": 2960
|
| 1826 |
},
|
| 1827 |
{
|
| 1828 |
"epoch": 2.66,
|
| 1829 |
"learning_rate": 2.6660682226211852e-05,
|
| 1830 |
-
"loss": 0.
|
| 1831 |
"step": 2970
|
| 1832 |
},
|
| 1833 |
{
|
| 1834 |
"epoch": 2.67,
|
| 1835 |
"learning_rate": 2.6750448833034114e-05,
|
| 1836 |
-
"loss": 0.
|
| 1837 |
"step": 2980
|
| 1838 |
},
|
| 1839 |
{
|
| 1840 |
"epoch": 2.68,
|
| 1841 |
"learning_rate": 2.6840215439856375e-05,
|
| 1842 |
-
"loss": 0.
|
| 1843 |
"step": 2990
|
| 1844 |
},
|
| 1845 |
{
|
| 1846 |
"epoch": 2.69,
|
| 1847 |
"learning_rate": 2.6929982046678636e-05,
|
| 1848 |
-
"loss": 0.
|
| 1849 |
"step": 3000
|
| 1850 |
},
|
| 1851 |
{
|
| 1852 |
"epoch": 2.7,
|
| 1853 |
"learning_rate": 2.7019748653500897e-05,
|
| 1854 |
-
"loss": 0.
|
| 1855 |
"step": 3010
|
| 1856 |
},
|
| 1857 |
{
|
| 1858 |
"epoch": 2.71,
|
| 1859 |
"learning_rate": 2.710951526032316e-05,
|
| 1860 |
-
"loss": 0.
|
| 1861 |
"step": 3020
|
| 1862 |
},
|
| 1863 |
{
|
| 1864 |
"epoch": 2.72,
|
| 1865 |
"learning_rate": 2.719928186714542e-05,
|
| 1866 |
-
"loss": 0.
|
| 1867 |
"step": 3030
|
| 1868 |
},
|
| 1869 |
{
|
| 1870 |
"epoch": 2.73,
|
| 1871 |
"learning_rate": 2.7289048473967688e-05,
|
| 1872 |
-
"loss": 0.
|
| 1873 |
"step": 3040
|
| 1874 |
},
|
| 1875 |
{
|
| 1876 |
"epoch": 2.74,
|
| 1877 |
"learning_rate": 2.737881508078995e-05,
|
| 1878 |
-
"loss": 0.
|
| 1879 |
"step": 3050
|
| 1880 |
},
|
| 1881 |
{
|
| 1882 |
"epoch": 2.75,
|
| 1883 |
"learning_rate": 2.746858168761221e-05,
|
| 1884 |
-
"loss": 0.
|
| 1885 |
"step": 3060
|
| 1886 |
},
|
| 1887 |
{
|
| 1888 |
"epoch": 2.75,
|
| 1889 |
"learning_rate": 2.755834829443447e-05,
|
| 1890 |
-
"loss": 0.
|
| 1891 |
"step": 3070
|
| 1892 |
},
|
| 1893 |
{
|
| 1894 |
"epoch": 2.76,
|
| 1895 |
"learning_rate": 2.7648114901256732e-05,
|
| 1896 |
-
"loss": 0.
|
| 1897 |
"step": 3080
|
| 1898 |
},
|
| 1899 |
{
|
| 1900 |
"epoch": 2.77,
|
| 1901 |
"learning_rate": 2.7737881508078994e-05,
|
| 1902 |
-
"loss": 0.
|
| 1903 |
"step": 3090
|
| 1904 |
},
|
| 1905 |
{
|
| 1906 |
"epoch": 2.78,
|
| 1907 |
"learning_rate": 2.7827648114901255e-05,
|
| 1908 |
-
"loss": 0.
|
| 1909 |
"step": 3100
|
| 1910 |
},
|
| 1911 |
{
|
| 1912 |
"epoch": 2.79,
|
| 1913 |
"learning_rate": 2.7917414721723523e-05,
|
| 1914 |
-
"loss": 0.
|
| 1915 |
"step": 3110
|
| 1916 |
},
|
| 1917 |
{
|
| 1918 |
"epoch": 2.8,
|
| 1919 |
"learning_rate": 2.8007181328545784e-05,
|
| 1920 |
-
"loss": 0.
|
| 1921 |
"step": 3120
|
| 1922 |
},
|
| 1923 |
{
|
| 1924 |
"epoch": 2.81,
|
| 1925 |
"learning_rate": 2.8096947935368045e-05,
|
| 1926 |
-
"loss": 0.
|
| 1927 |
"step": 3130
|
| 1928 |
},
|
| 1929 |
{
|
| 1930 |
"epoch": 2.82,
|
| 1931 |
"learning_rate": 2.8186714542190307e-05,
|
| 1932 |
-
"loss": 0.
|
| 1933 |
"step": 3140
|
| 1934 |
},
|
| 1935 |
{
|
| 1936 |
"epoch": 2.83,
|
| 1937 |
"learning_rate": 2.8276481149012568e-05,
|
| 1938 |
-
"loss": 0.
|
| 1939 |
"step": 3150
|
| 1940 |
},
|
| 1941 |
{
|
| 1942 |
"epoch": 2.83,
|
| 1943 |
"learning_rate": 2.836624775583483e-05,
|
| 1944 |
-
"loss": 0.
|
| 1945 |
"step": 3160
|
| 1946 |
},
|
| 1947 |
{
|
| 1948 |
"epoch": 2.84,
|
| 1949 |
"learning_rate": 2.845601436265709e-05,
|
| 1950 |
-
"loss": 0.
|
| 1951 |
"step": 3170
|
| 1952 |
},
|
| 1953 |
{
|
| 1954 |
"epoch": 2.85,
|
| 1955 |
"learning_rate": 2.8545780969479358e-05,
|
| 1956 |
-
"loss": 0.
|
| 1957 |
"step": 3180
|
| 1958 |
},
|
| 1959 |
{
|
| 1960 |
"epoch": 2.86,
|
| 1961 |
"learning_rate": 2.863554757630162e-05,
|
| 1962 |
-
"loss": 0.
|
| 1963 |
"step": 3190
|
| 1964 |
},
|
| 1965 |
{
|
| 1966 |
"epoch": 2.87,
|
| 1967 |
"learning_rate": 2.872531418312388e-05,
|
| 1968 |
-
"loss": 0.
|
| 1969 |
"step": 3200
|
| 1970 |
},
|
| 1971 |
{
|
| 1972 |
"epoch": 2.88,
|
| 1973 |
"learning_rate": 2.8815080789946142e-05,
|
| 1974 |
-
"loss": 0.
|
| 1975 |
"step": 3210
|
| 1976 |
},
|
| 1977 |
{
|
| 1978 |
"epoch": 2.89,
|
| 1979 |
"learning_rate": 2.8904847396768403e-05,
|
| 1980 |
-
"loss": 0.
|
| 1981 |
"step": 3220
|
| 1982 |
},
|
| 1983 |
{
|
| 1984 |
"epoch": 2.9,
|
| 1985 |
"learning_rate": 2.8994614003590664e-05,
|
| 1986 |
-
"loss": 0.
|
| 1987 |
"step": 3230
|
| 1988 |
},
|
| 1989 |
{
|
| 1990 |
"epoch": 2.91,
|
| 1991 |
"learning_rate": 2.9084380610412926e-05,
|
| 1992 |
-
"loss": 0.
|
| 1993 |
"step": 3240
|
| 1994 |
},
|
| 1995 |
{
|
| 1996 |
"epoch": 2.92,
|
| 1997 |
"learning_rate": 2.9174147217235194e-05,
|
| 1998 |
-
"loss": 0.
|
| 1999 |
"step": 3250
|
| 2000 |
},
|
| 2001 |
{
|
| 2002 |
"epoch": 2.92,
|
| 2003 |
"learning_rate": 2.9263913824057455e-05,
|
| 2004 |
-
"loss": 0.
|
| 2005 |
"step": 3260
|
| 2006 |
},
|
| 2007 |
{
|
| 2008 |
"epoch": 2.93,
|
| 2009 |
"learning_rate": 2.9353680430879716e-05,
|
| 2010 |
-
"loss": 0.
|
| 2011 |
"step": 3270
|
| 2012 |
},
|
| 2013 |
{
|
| 2014 |
"epoch": 2.94,
|
| 2015 |
"learning_rate": 2.9443447037701977e-05,
|
| 2016 |
-
"loss": 0.
|
| 2017 |
"step": 3280
|
| 2018 |
},
|
| 2019 |
{
|
| 2020 |
"epoch": 2.95,
|
| 2021 |
"learning_rate": 2.953321364452424e-05,
|
| 2022 |
-
"loss": 0.
|
| 2023 |
"step": 3290
|
| 2024 |
},
|
| 2025 |
{
|
| 2026 |
"epoch": 2.96,
|
| 2027 |
"learning_rate": 2.96229802513465e-05,
|
| 2028 |
-
"loss": 0.
|
| 2029 |
"step": 3300
|
| 2030 |
},
|
| 2031 |
{
|
| 2032 |
"epoch": 2.97,
|
| 2033 |
"learning_rate": 2.971274685816876e-05,
|
| 2034 |
-
"loss": 0.
|
| 2035 |
"step": 3310
|
| 2036 |
},
|
| 2037 |
{
|
| 2038 |
"epoch": 2.98,
|
| 2039 |
"learning_rate": 2.9802513464991022e-05,
|
| 2040 |
-
"loss": 0.
|
| 2041 |
"step": 3320
|
| 2042 |
},
|
| 2043 |
{
|
| 2044 |
"epoch": 2.99,
|
| 2045 |
"learning_rate": 2.989228007181329e-05,
|
| 2046 |
-
"loss": 0.
|
| 2047 |
"step": 3330
|
| 2048 |
},
|
| 2049 |
{
|
| 2050 |
"epoch": 3.0,
|
| 2051 |
"learning_rate": 2.998204667863555e-05,
|
| 2052 |
-
"loss": 0.
|
| 2053 |
"step": 3340
|
| 2054 |
},
|
| 2055 |
{
|
| 2056 |
"epoch": 3.0,
|
| 2057 |
"eval_accuracy": {
|
| 2058 |
-
"accuracy": 0.
|
| 2059 |
},
|
| 2060 |
"eval_f1": {
|
| 2061 |
-
"f1": 0.
|
| 2062 |
},
|
| 2063 |
-
"eval_loss": 0.
|
| 2064 |
"eval_precision": {
|
| 2065 |
-
"precision": 0.
|
| 2066 |
},
|
| 2067 |
"eval_recall": {
|
| 2068 |
-
"recall": 0.
|
| 2069 |
},
|
| 2070 |
-
"eval_runtime":
|
| 2071 |
-
"eval_samples_per_second":
|
| 2072 |
-
"eval_steps_per_second": 6.
|
| 2073 |
"step": 3344
|
| 2074 |
},
|
| 2075 |
{
|
| 2076 |
"epoch": 3.01,
|
| 2077 |
"learning_rate": 3.0071813285457812e-05,
|
| 2078 |
-
"loss": 0.
|
| 2079 |
"step": 3350
|
| 2080 |
},
|
| 2081 |
{
|
| 2082 |
"epoch": 3.01,
|
| 2083 |
"learning_rate": 3.0161579892280074e-05,
|
| 2084 |
-
"loss": 0.
|
| 2085 |
"step": 3360
|
| 2086 |
},
|
| 2087 |
{
|
| 2088 |
"epoch": 3.02,
|
| 2089 |
"learning_rate": 3.0251346499102335e-05,
|
| 2090 |
-
"loss": 0.
|
| 2091 |
"step": 3370
|
| 2092 |
},
|
| 2093 |
{
|
| 2094 |
"epoch": 3.03,
|
| 2095 |
"learning_rate": 3.0341113105924596e-05,
|
| 2096 |
-
"loss": 0.
|
| 2097 |
"step": 3380
|
| 2098 |
},
|
| 2099 |
{
|
| 2100 |
"epoch": 3.04,
|
| 2101 |
"learning_rate": 3.0430879712746857e-05,
|
| 2102 |
-
"loss": 0.
|
| 2103 |
"step": 3390
|
| 2104 |
},
|
| 2105 |
{
|
| 2106 |
"epoch": 3.05,
|
| 2107 |
"learning_rate": 3.0520646319569125e-05,
|
| 2108 |
-
"loss": 0.
|
| 2109 |
"step": 3400
|
| 2110 |
},
|
| 2111 |
{
|
| 2112 |
"epoch": 3.06,
|
| 2113 |
"learning_rate": 3.061041292639138e-05,
|
| 2114 |
-
"loss": 0.
|
| 2115 |
"step": 3410
|
| 2116 |
},
|
| 2117 |
{
|
| 2118 |
"epoch": 3.07,
|
| 2119 |
"learning_rate": 3.070017953321365e-05,
|
| 2120 |
-
"loss": 0.
|
| 2121 |
"step": 3420
|
| 2122 |
},
|
| 2123 |
{
|
| 2124 |
"epoch": 3.08,
|
| 2125 |
"learning_rate": 3.0789946140035906e-05,
|
| 2126 |
-
"loss": 0.
|
| 2127 |
"step": 3430
|
| 2128 |
},
|
| 2129 |
{
|
| 2130 |
"epoch": 3.09,
|
| 2131 |
"learning_rate": 3.087971274685817e-05,
|
| 2132 |
-
"loss": 0.
|
| 2133 |
"step": 3440
|
| 2134 |
},
|
| 2135 |
{
|
| 2136 |
"epoch": 3.09,
|
| 2137 |
"learning_rate": 3.096947935368043e-05,
|
| 2138 |
-
"loss": 0.
|
| 2139 |
"step": 3450
|
| 2140 |
},
|
| 2141 |
{
|
| 2142 |
"epoch": 3.1,
|
| 2143 |
"learning_rate": 3.105924596050269e-05,
|
| 2144 |
-
"loss": 0.
|
| 2145 |
"step": 3460
|
| 2146 |
},
|
| 2147 |
{
|
| 2148 |
"epoch": 3.11,
|
| 2149 |
"learning_rate": 3.114901256732496e-05,
|
| 2150 |
-
"loss": 0.
|
| 2151 |
"step": 3470
|
| 2152 |
},
|
| 2153 |
{
|
| 2154 |
"epoch": 3.12,
|
| 2155 |
"learning_rate": 3.123877917414722e-05,
|
| 2156 |
-
"loss": 0.
|
| 2157 |
"step": 3480
|
| 2158 |
},
|
| 2159 |
{
|
| 2160 |
"epoch": 3.13,
|
| 2161 |
"learning_rate": 3.132854578096948e-05,
|
| 2162 |
-
"loss": 0.
|
| 2163 |
"step": 3490
|
| 2164 |
},
|
| 2165 |
{
|
| 2166 |
"epoch": 3.14,
|
| 2167 |
"learning_rate": 3.1418312387791744e-05,
|
| 2168 |
-
"loss": 0.
|
| 2169 |
"step": 3500
|
| 2170 |
},
|
| 2171 |
{
|
| 2172 |
"epoch": 3.15,
|
| 2173 |
"learning_rate": 3.1508078994614e-05,
|
| 2174 |
-
"loss": 0.
|
| 2175 |
"step": 3510
|
| 2176 |
},
|
| 2177 |
{
|
| 2178 |
"epoch": 3.16,
|
| 2179 |
"learning_rate": 3.159784560143627e-05,
|
| 2180 |
-
"loss": 0.
|
| 2181 |
"step": 3520
|
| 2182 |
},
|
| 2183 |
{
|
| 2184 |
"epoch": 3.17,
|
| 2185 |
"learning_rate": 3.1687612208258525e-05,
|
| 2186 |
-
"loss": 0.
|
| 2187 |
"step": 3530
|
| 2188 |
},
|
| 2189 |
{
|
| 2190 |
"epoch": 3.18,
|
| 2191 |
"learning_rate": 3.1777378815080796e-05,
|
| 2192 |
-
"loss": 0.
|
| 2193 |
"step": 3540
|
| 2194 |
},
|
| 2195 |
{
|
| 2196 |
"epoch": 3.18,
|
| 2197 |
"learning_rate": 3.1867145421903054e-05,
|
| 2198 |
-
"loss": 0.
|
| 2199 |
"step": 3550
|
| 2200 |
},
|
| 2201 |
{
|
| 2202 |
"epoch": 3.19,
|
| 2203 |
"learning_rate": 3.195691202872532e-05,
|
| 2204 |
-
"loss": 0.
|
| 2205 |
"step": 3560
|
| 2206 |
},
|
| 2207 |
{
|
| 2208 |
"epoch": 3.2,
|
| 2209 |
"learning_rate": 3.2046678635547576e-05,
|
| 2210 |
-
"loss": 0.
|
| 2211 |
"step": 3570
|
| 2212 |
},
|
| 2213 |
{
|
| 2214 |
"epoch": 3.21,
|
| 2215 |
"learning_rate": 3.213644524236984e-05,
|
| 2216 |
-
"loss": 0.
|
| 2217 |
"step": 3580
|
| 2218 |
},
|
| 2219 |
{
|
| 2220 |
"epoch": 3.22,
|
| 2221 |
"learning_rate": 3.22262118491921e-05,
|
| 2222 |
-
"loss": 0.
|
| 2223 |
"step": 3590
|
| 2224 |
},
|
| 2225 |
{
|
| 2226 |
"epoch": 3.23,
|
| 2227 |
"learning_rate": 3.231597845601436e-05,
|
| 2228 |
-
"loss": 0.
|
| 2229 |
"step": 3600
|
| 2230 |
},
|
| 2231 |
{
|
| 2232 |
"epoch": 3.24,
|
| 2233 |
"learning_rate": 3.240574506283663e-05,
|
| 2234 |
-
"loss": 0.
|
| 2235 |
"step": 3610
|
| 2236 |
},
|
| 2237 |
{
|
| 2238 |
"epoch": 3.25,
|
| 2239 |
"learning_rate": 3.249551166965889e-05,
|
| 2240 |
-
"loss": 0.
|
| 2241 |
"step": 3620
|
| 2242 |
},
|
| 2243 |
{
|
| 2244 |
"epoch": 3.26,
|
| 2245 |
"learning_rate": 3.258527827648115e-05,
|
| 2246 |
-
"loss": 0.
|
| 2247 |
"step": 3630
|
| 2248 |
},
|
| 2249 |
{
|
| 2250 |
"epoch": 3.27,
|
| 2251 |
"learning_rate": 3.2675044883303415e-05,
|
| 2252 |
-
"loss": 0.
|
| 2253 |
"step": 3640
|
| 2254 |
},
|
| 2255 |
{
|
| 2256 |
"epoch": 3.27,
|
| 2257 |
"learning_rate": 3.276481149012567e-05,
|
| 2258 |
-
"loss": 0.
|
| 2259 |
"step": 3650
|
| 2260 |
},
|
| 2261 |
{
|
| 2262 |
"epoch": 3.28,
|
| 2263 |
"learning_rate": 3.285457809694794e-05,
|
| 2264 |
-
"loss": 0.
|
| 2265 |
"step": 3660
|
| 2266 |
},
|
| 2267 |
{
|
| 2268 |
"epoch": 3.29,
|
| 2269 |
"learning_rate": 3.2944344703770195e-05,
|
| 2270 |
-
"loss": 0.
|
| 2271 |
"step": 3670
|
| 2272 |
},
|
| 2273 |
{
|
| 2274 |
"epoch": 3.3,
|
| 2275 |
"learning_rate": 3.3034111310592466e-05,
|
| 2276 |
-
"loss": 0.
|
| 2277 |
"step": 3680
|
| 2278 |
},
|
| 2279 |
{
|
| 2280 |
"epoch": 3.31,
|
| 2281 |
"learning_rate": 3.3123877917414724e-05,
|
| 2282 |
-
"loss": 0.
|
| 2283 |
"step": 3690
|
| 2284 |
},
|
| 2285 |
{
|
| 2286 |
"epoch": 3.32,
|
| 2287 |
"learning_rate": 3.321364452423699e-05,
|
| 2288 |
-
"loss": 0.
|
| 2289 |
"step": 3700
|
| 2290 |
},
|
| 2291 |
{
|
| 2292 |
"epoch": 3.33,
|
| 2293 |
"learning_rate": 3.330341113105925e-05,
|
| 2294 |
-
"loss": 0.
|
| 2295 |
"step": 3710
|
| 2296 |
},
|
| 2297 |
{
|
| 2298 |
"epoch": 3.34,
|
| 2299 |
"learning_rate": 3.339317773788151e-05,
|
| 2300 |
-
"loss": 0.
|
| 2301 |
"step": 3720
|
| 2302 |
},
|
| 2303 |
{
|
| 2304 |
"epoch": 3.35,
|
| 2305 |
"learning_rate": 3.348294434470377e-05,
|
| 2306 |
-
"loss": 0.
|
| 2307 |
"step": 3730
|
| 2308 |
},
|
| 2309 |
{
|
| 2310 |
"epoch": 3.36,
|
| 2311 |
"learning_rate": 3.3572710951526034e-05,
|
| 2312 |
-
"loss": 0.
|
| 2313 |
"step": 3740
|
| 2314 |
},
|
| 2315 |
{
|
| 2316 |
"epoch": 3.36,
|
| 2317 |
"learning_rate": 3.36624775583483e-05,
|
| 2318 |
-
"loss": 0.
|
| 2319 |
"step": 3750
|
| 2320 |
},
|
| 2321 |
{
|
| 2322 |
"epoch": 3.37,
|
| 2323 |
"learning_rate": 3.375224416517056e-05,
|
| 2324 |
-
"loss": 0.
|
| 2325 |
"step": 3760
|
| 2326 |
},
|
| 2327 |
{
|
| 2328 |
"epoch": 3.38,
|
| 2329 |
"learning_rate": 3.384201077199282e-05,
|
| 2330 |
-
"loss": 0.
|
| 2331 |
"step": 3770
|
| 2332 |
},
|
| 2333 |
{
|
| 2334 |
"epoch": 3.39,
|
| 2335 |
"learning_rate": 3.3931777378815085e-05,
|
| 2336 |
-
"loss": 0.
|
| 2337 |
"step": 3780
|
| 2338 |
},
|
| 2339 |
{
|
| 2340 |
"epoch": 3.4,
|
| 2341 |
"learning_rate": 3.402154398563734e-05,
|
| 2342 |
-
"loss": 0.
|
| 2343 |
"step": 3790
|
| 2344 |
},
|
| 2345 |
{
|
| 2346 |
"epoch": 3.41,
|
| 2347 |
"learning_rate": 3.411131059245961e-05,
|
| 2348 |
-
"loss": 0.
|
| 2349 |
"step": 3800
|
| 2350 |
},
|
| 2351 |
{
|
| 2352 |
"epoch": 3.42,
|
| 2353 |
"learning_rate": 3.4201077199281866e-05,
|
| 2354 |
-
"loss": 0.
|
| 2355 |
"step": 3810
|
| 2356 |
},
|
| 2357 |
{
|
| 2358 |
"epoch": 3.43,
|
| 2359 |
"learning_rate": 3.429084380610413e-05,
|
| 2360 |
-
"loss": 0.
|
| 2361 |
"step": 3820
|
| 2362 |
},
|
| 2363 |
{
|
| 2364 |
"epoch": 3.44,
|
| 2365 |
"learning_rate": 3.4380610412926395e-05,
|
| 2366 |
-
"loss": 0.
|
| 2367 |
"step": 3830
|
| 2368 |
},
|
| 2369 |
{
|
| 2370 |
"epoch": 3.44,
|
| 2371 |
"learning_rate": 3.447037701974866e-05,
|
| 2372 |
-
"loss": 0.
|
| 2373 |
"step": 3840
|
| 2374 |
},
|
| 2375 |
{
|
| 2376 |
"epoch": 3.45,
|
| 2377 |
"learning_rate": 3.456014362657092e-05,
|
| 2378 |
-
"loss": 0.
|
| 2379 |
"step": 3850
|
| 2380 |
},
|
| 2381 |
{
|
| 2382 |
"epoch": 3.46,
|
| 2383 |
"learning_rate": 3.464991023339318e-05,
|
| 2384 |
-
"loss": 0.
|
| 2385 |
"step": 3860
|
| 2386 |
},
|
| 2387 |
{
|
| 2388 |
"epoch": 3.47,
|
| 2389 |
"learning_rate": 3.473967684021544e-05,
|
| 2390 |
-
"loss": 0.
|
| 2391 |
"step": 3870
|
| 2392 |
},
|
| 2393 |
{
|
| 2394 |
"epoch": 3.48,
|
| 2395 |
"learning_rate": 3.4829443447037704e-05,
|
| 2396 |
-
"loss": 0.
|
| 2397 |
"step": 3880
|
| 2398 |
},
|
| 2399 |
{
|
| 2400 |
"epoch": 3.49,
|
| 2401 |
"learning_rate": 3.491921005385996e-05,
|
| 2402 |
-
"loss": 0.
|
| 2403 |
"step": 3890
|
| 2404 |
},
|
| 2405 |
{
|
| 2406 |
"epoch": 3.5,
|
| 2407 |
"learning_rate": 3.500897666068223e-05,
|
| 2408 |
-
"loss": 0.
|
| 2409 |
"step": 3900
|
| 2410 |
},
|
| 2411 |
{
|
| 2412 |
"epoch": 3.51,
|
| 2413 |
"learning_rate": 3.509874326750449e-05,
|
| 2414 |
-
"loss": 0.
|
| 2415 |
"step": 3910
|
| 2416 |
},
|
| 2417 |
{
|
| 2418 |
"epoch": 3.52,
|
| 2419 |
"learning_rate": 3.5188509874326756e-05,
|
| 2420 |
-
"loss": 0.
|
| 2421 |
"step": 3920
|
| 2422 |
},
|
| 2423 |
{
|
| 2424 |
"epoch": 3.53,
|
| 2425 |
"learning_rate": 3.5278276481149014e-05,
|
| 2426 |
-
"loss": 0.
|
| 2427 |
"step": 3930
|
| 2428 |
},
|
| 2429 |
{
|
| 2430 |
"epoch": 3.53,
|
| 2431 |
"learning_rate": 3.536804308797128e-05,
|
| 2432 |
-
"loss": 0.
|
| 2433 |
"step": 3940
|
| 2434 |
},
|
| 2435 |
{
|
| 2436 |
"epoch": 3.54,
|
| 2437 |
"learning_rate": 3.5457809694793536e-05,
|
| 2438 |
-
"loss": 0.
|
| 2439 |
"step": 3950
|
| 2440 |
},
|
| 2441 |
{
|
| 2442 |
"epoch": 3.55,
|
| 2443 |
"learning_rate": 3.55475763016158e-05,
|
| 2444 |
-
"loss": 0.
|
| 2445 |
"step": 3960
|
| 2446 |
},
|
| 2447 |
{
|
| 2448 |
"epoch": 3.56,
|
| 2449 |
"learning_rate": 3.5637342908438065e-05,
|
| 2450 |
-
"loss": 0.
|
| 2451 |
"step": 3970
|
| 2452 |
},
|
| 2453 |
{
|
| 2454 |
"epoch": 3.57,
|
| 2455 |
"learning_rate": 3.572710951526032e-05,
|
| 2456 |
-
"loss": 0.
|
| 2457 |
"step": 3980
|
| 2458 |
},
|
| 2459 |
{
|
| 2460 |
"epoch": 3.58,
|
| 2461 |
"learning_rate": 3.581687612208259e-05,
|
| 2462 |
-
"loss": 0.
|
| 2463 |
"step": 3990
|
| 2464 |
},
|
| 2465 |
{
|
| 2466 |
"epoch": 3.59,
|
| 2467 |
"learning_rate": 3.5906642728904846e-05,
|
| 2468 |
-
"loss": 0.
|
| 2469 |
"step": 4000
|
| 2470 |
},
|
| 2471 |
{
|
| 2472 |
"epoch": 3.6,
|
| 2473 |
"learning_rate": 3.599640933572711e-05,
|
| 2474 |
-
"loss": 0.
|
| 2475 |
"step": 4010
|
| 2476 |
},
|
| 2477 |
{
|
| 2478 |
"epoch": 3.61,
|
| 2479 |
"learning_rate": 3.6086175942549375e-05,
|
| 2480 |
-
"loss": 0.
|
| 2481 |
"step": 4020
|
| 2482 |
},
|
| 2483 |
{
|
| 2484 |
"epoch": 3.62,
|
| 2485 |
"learning_rate": 3.617594254937163e-05,
|
| 2486 |
-
"loss": 0.
|
| 2487 |
"step": 4030
|
| 2488 |
},
|
| 2489 |
{
|
| 2490 |
"epoch": 3.62,
|
| 2491 |
"learning_rate": 3.62657091561939e-05,
|
| 2492 |
-
"loss": 0.
|
| 2493 |
"step": 4040
|
| 2494 |
},
|
| 2495 |
{
|
| 2496 |
"epoch": 3.63,
|
| 2497 |
"learning_rate": 3.635547576301616e-05,
|
| 2498 |
-
"loss": 0.
|
| 2499 |
"step": 4050
|
| 2500 |
},
|
| 2501 |
{
|
| 2502 |
"epoch": 3.64,
|
| 2503 |
"learning_rate": 3.644524236983842e-05,
|
| 2504 |
-
"loss": 0.
|
| 2505 |
"step": 4060
|
| 2506 |
},
|
| 2507 |
{
|
| 2508 |
"epoch": 3.65,
|
| 2509 |
"learning_rate": 3.6535008976660684e-05,
|
| 2510 |
-
"loss": 0.
|
| 2511 |
"step": 4070
|
| 2512 |
},
|
| 2513 |
{
|
| 2514 |
"epoch": 3.66,
|
| 2515 |
"learning_rate": 3.662477558348294e-05,
|
| 2516 |
-
"loss": 0.
|
| 2517 |
"step": 4080
|
| 2518 |
},
|
| 2519 |
{
|
| 2520 |
"epoch": 3.67,
|
| 2521 |
"learning_rate": 3.671454219030521e-05,
|
| 2522 |
-
"loss": 0.
|
| 2523 |
"step": 4090
|
| 2524 |
},
|
| 2525 |
{
|
| 2526 |
"epoch": 3.68,
|
| 2527 |
"learning_rate": 3.6804308797127465e-05,
|
| 2528 |
-
"loss": 0.
|
| 2529 |
"step": 4100
|
| 2530 |
},
|
| 2531 |
{
|
| 2532 |
"epoch": 3.69,
|
| 2533 |
"learning_rate": 3.6894075403949736e-05,
|
| 2534 |
-
"loss": 0.
|
| 2535 |
"step": 4110
|
| 2536 |
},
|
| 2537 |
{
|
| 2538 |
"epoch": 3.7,
|
| 2539 |
"learning_rate": 3.6983842010771994e-05,
|
| 2540 |
-
"loss": 0.
|
| 2541 |
"step": 4120
|
| 2542 |
},
|
| 2543 |
{
|
| 2544 |
"epoch": 3.7,
|
| 2545 |
"learning_rate": 3.707360861759426e-05,
|
| 2546 |
-
"loss": 0.
|
| 2547 |
"step": 4130
|
| 2548 |
},
|
| 2549 |
{
|
| 2550 |
"epoch": 3.71,
|
| 2551 |
"learning_rate": 3.7163375224416516e-05,
|
| 2552 |
-
"loss": 0.
|
| 2553 |
"step": 4140
|
| 2554 |
},
|
| 2555 |
{
|
| 2556 |
"epoch": 3.72,
|
| 2557 |
"learning_rate": 3.725314183123878e-05,
|
| 2558 |
-
"loss": 0.
|
| 2559 |
"step": 4150
|
| 2560 |
},
|
| 2561 |
{
|
| 2562 |
"epoch": 3.73,
|
| 2563 |
"learning_rate": 3.734290843806104e-05,
|
| 2564 |
-
"loss": 0.
|
| 2565 |
"step": 4160
|
| 2566 |
},
|
| 2567 |
{
|
| 2568 |
"epoch": 3.74,
|
| 2569 |
"learning_rate": 3.74326750448833e-05,
|
| 2570 |
-
"loss": 0.
|
| 2571 |
"step": 4170
|
| 2572 |
},
|
| 2573 |
{
|
| 2574 |
"epoch": 3.75,
|
| 2575 |
"learning_rate": 3.752244165170557e-05,
|
| 2576 |
-
"loss": 0.
|
| 2577 |
"step": 4180
|
| 2578 |
},
|
| 2579 |
{
|
| 2580 |
"epoch": 3.76,
|
| 2581 |
"learning_rate": 3.761220825852783e-05,
|
| 2582 |
-
"loss": 0.
|
| 2583 |
"step": 4190
|
| 2584 |
},
|
| 2585 |
{
|
| 2586 |
"epoch": 3.77,
|
| 2587 |
"learning_rate": 3.770197486535009e-05,
|
| 2588 |
-
"loss": 0.
|
| 2589 |
"step": 4200
|
| 2590 |
},
|
| 2591 |
{
|
| 2592 |
"epoch": 3.78,
|
| 2593 |
"learning_rate": 3.7791741472172355e-05,
|
| 2594 |
-
"loss": 0.
|
| 2595 |
"step": 4210
|
| 2596 |
},
|
| 2597 |
{
|
| 2598 |
"epoch": 3.79,
|
| 2599 |
"learning_rate": 3.788150807899461e-05,
|
| 2600 |
-
"loss": 0.
|
| 2601 |
"step": 4220
|
| 2602 |
},
|
| 2603 |
{
|
| 2604 |
"epoch": 3.79,
|
| 2605 |
"learning_rate": 3.797127468581688e-05,
|
| 2606 |
-
"loss": 0.
|
| 2607 |
"step": 4230
|
| 2608 |
},
|
| 2609 |
{
|
| 2610 |
"epoch": 3.8,
|
| 2611 |
"learning_rate": 3.8061041292639135e-05,
|
| 2612 |
-
"loss": 0.
|
| 2613 |
"step": 4240
|
| 2614 |
},
|
| 2615 |
{
|
| 2616 |
"epoch": 3.81,
|
| 2617 |
"learning_rate": 3.815080789946141e-05,
|
| 2618 |
-
"loss": 0.
|
| 2619 |
"step": 4250
|
| 2620 |
},
|
| 2621 |
{
|
| 2622 |
"epoch": 3.82,
|
| 2623 |
"learning_rate": 3.8240574506283664e-05,
|
| 2624 |
-
"loss": 0.
|
| 2625 |
"step": 4260
|
| 2626 |
},
|
| 2627 |
{
|
| 2628 |
"epoch": 3.83,
|
| 2629 |
"learning_rate": 3.833034111310593e-05,
|
| 2630 |
-
"loss": 0.
|
| 2631 |
"step": 4270
|
| 2632 |
},
|
| 2633 |
{
|
| 2634 |
"epoch": 3.84,
|
| 2635 |
"learning_rate": 3.842010771992819e-05,
|
| 2636 |
-
"loss": 0.
|
| 2637 |
"step": 4280
|
| 2638 |
},
|
| 2639 |
{
|
| 2640 |
"epoch": 3.85,
|
| 2641 |
"learning_rate": 3.850987432675045e-05,
|
| 2642 |
-
"loss": 0.
|
| 2643 |
"step": 4290
|
| 2644 |
},
|
| 2645 |
{
|
| 2646 |
"epoch": 3.86,
|
| 2647 |
"learning_rate": 3.859964093357271e-05,
|
| 2648 |
-
"loss": 0.
|
| 2649 |
"step": 4300
|
| 2650 |
},
|
| 2651 |
{
|
| 2652 |
"epoch": 3.87,
|
| 2653 |
"learning_rate": 3.8689407540394974e-05,
|
| 2654 |
-
"loss": 0.
|
| 2655 |
"step": 4310
|
| 2656 |
},
|
| 2657 |
{
|
| 2658 |
"epoch": 3.88,
|
| 2659 |
"learning_rate": 3.877917414721723e-05,
|
| 2660 |
-
"loss": 0.
|
| 2661 |
"step": 4320
|
| 2662 |
},
|
| 2663 |
{
|
| 2664 |
"epoch": 3.88,
|
| 2665 |
"learning_rate": 3.88689407540395e-05,
|
| 2666 |
-
"loss": 0.
|
| 2667 |
"step": 4330
|
| 2668 |
},
|
| 2669 |
{
|
| 2670 |
"epoch": 3.89,
|
| 2671 |
"learning_rate": 3.895870736086176e-05,
|
| 2672 |
-
"loss": 0.
|
| 2673 |
"step": 4340
|
| 2674 |
},
|
| 2675 |
{
|
| 2676 |
"epoch": 3.9,
|
| 2677 |
"learning_rate": 3.9048473967684026e-05,
|
| 2678 |
-
"loss": 0.
|
| 2679 |
"step": 4350
|
| 2680 |
},
|
| 2681 |
{
|
| 2682 |
"epoch": 3.91,
|
| 2683 |
"learning_rate": 3.9138240574506283e-05,
|
| 2684 |
-
"loss": 0.
|
| 2685 |
"step": 4360
|
| 2686 |
},
|
| 2687 |
{
|
| 2688 |
"epoch": 3.92,
|
| 2689 |
"learning_rate": 3.922800718132855e-05,
|
| 2690 |
-
"loss": 0.
|
| 2691 |
"step": 4370
|
| 2692 |
},
|
| 2693 |
{
|
| 2694 |
"epoch": 3.93,
|
| 2695 |
"learning_rate": 3.9317773788150806e-05,
|
| 2696 |
-
"loss": 0.
|
| 2697 |
"step": 4380
|
| 2698 |
},
|
| 2699 |
{
|
| 2700 |
"epoch": 3.94,
|
| 2701 |
"learning_rate": 3.940754039497307e-05,
|
| 2702 |
-
"loss": 0.
|
| 2703 |
"step": 4390
|
| 2704 |
},
|
| 2705 |
{
|
| 2706 |
"epoch": 3.95,
|
| 2707 |
"learning_rate": 3.9497307001795335e-05,
|
| 2708 |
-
"loss": 0.
|
| 2709 |
"step": 4400
|
| 2710 |
},
|
| 2711 |
{
|
| 2712 |
"epoch": 3.96,
|
| 2713 |
"learning_rate": 3.95870736086176e-05,
|
| 2714 |
-
"loss": 0.
|
| 2715 |
"step": 4410
|
| 2716 |
},
|
| 2717 |
{
|
| 2718 |
"epoch": 3.97,
|
| 2719 |
"learning_rate": 3.967684021543986e-05,
|
| 2720 |
-
"loss": 0.
|
| 2721 |
"step": 4420
|
| 2722 |
},
|
| 2723 |
{
|
| 2724 |
"epoch": 3.97,
|
| 2725 |
"learning_rate": 3.976660682226212e-05,
|
| 2726 |
-
"loss": 0.
|
| 2727 |
"step": 4430
|
| 2728 |
},
|
| 2729 |
{
|
| 2730 |
"epoch": 3.98,
|
| 2731 |
"learning_rate": 3.985637342908438e-05,
|
| 2732 |
-
"loss": 0.
|
| 2733 |
"step": 4440
|
| 2734 |
},
|
| 2735 |
{
|
| 2736 |
"epoch": 3.99,
|
| 2737 |
"learning_rate": 3.9946140035906645e-05,
|
| 2738 |
-
"loss": 0.
|
| 2739 |
"step": 4450
|
| 2740 |
},
|
| 2741 |
{
|
| 2742 |
"epoch": 4.0,
|
| 2743 |
"eval_accuracy": {
|
| 2744 |
-
"accuracy": 0.
|
| 2745 |
},
|
| 2746 |
"eval_f1": {
|
| 2747 |
-
"f1": 0.
|
| 2748 |
},
|
| 2749 |
-
"eval_loss": 0.
|
| 2750 |
"eval_precision": {
|
| 2751 |
-
"precision": 0.
|
| 2752 |
},
|
| 2753 |
"eval_recall": {
|
| 2754 |
-
"recall": 0.
|
| 2755 |
},
|
| 2756 |
-
"eval_runtime": 165.
|
| 2757 |
-
"eval_samples_per_second": 107.
|
| 2758 |
-
"eval_steps_per_second": 6.
|
| 2759 |
"step": 4459
|
| 2760 |
},
|
| 2761 |
{
|
| 2762 |
"epoch": 4.0,
|
| 2763 |
"learning_rate": 4.00359066427289e-05,
|
| 2764 |
-
"loss": 0.
|
| 2765 |
"step": 4460
|
| 2766 |
},
|
| 2767 |
{
|
| 2768 |
"epoch": 4.01,
|
| 2769 |
"learning_rate": 4.0125673249551174e-05,
|
| 2770 |
-
"loss": 0.
|
| 2771 |
"step": 4470
|
| 2772 |
},
|
| 2773 |
{
|
| 2774 |
"epoch": 4.02,
|
| 2775 |
"learning_rate": 4.021543985637343e-05,
|
| 2776 |
-
"loss": 0.
|
| 2777 |
"step": 4480
|
| 2778 |
},
|
| 2779 |
{
|
| 2780 |
"epoch": 4.03,
|
| 2781 |
"learning_rate": 4.0305206463195696e-05,
|
| 2782 |
-
"loss": 0.
|
| 2783 |
"step": 4490
|
| 2784 |
},
|
| 2785 |
{
|
| 2786 |
"epoch": 4.04,
|
| 2787 |
"learning_rate": 4.0394973070017954e-05,
|
| 2788 |
-
"loss": 0.
|
| 2789 |
"step": 4500
|
| 2790 |
},
|
| 2791 |
{
|
| 2792 |
"epoch": 4.05,
|
| 2793 |
"learning_rate": 4.048473967684022e-05,
|
| 2794 |
-
"loss": 0.
|
| 2795 |
"step": 4510
|
| 2796 |
},
|
| 2797 |
{
|
| 2798 |
"epoch": 4.05,
|
| 2799 |
"learning_rate": 4.0574506283662476e-05,
|
| 2800 |
-
"loss": 0.
|
| 2801 |
"step": 4520
|
| 2802 |
},
|
| 2803 |
{
|
| 2804 |
"epoch": 4.06,
|
| 2805 |
"learning_rate": 4.066427289048474e-05,
|
| 2806 |
-
"loss": 0.
|
| 2807 |
"step": 4530
|
| 2808 |
},
|
| 2809 |
{
|
| 2810 |
"epoch": 4.07,
|
| 2811 |
"learning_rate": 4.0754039497307006e-05,
|
| 2812 |
-
"loss": 0.
|
| 2813 |
"step": 4540
|
| 2814 |
},
|
| 2815 |
{
|
| 2816 |
"epoch": 4.08,
|
| 2817 |
"learning_rate": 4.084380610412927e-05,
|
| 2818 |
-
"loss": 0.
|
| 2819 |
"step": 4550
|
| 2820 |
},
|
| 2821 |
{
|
| 2822 |
"epoch": 4.09,
|
| 2823 |
"learning_rate": 4.093357271095153e-05,
|
| 2824 |
-
"loss": 0.
|
| 2825 |
"step": 4560
|
| 2826 |
},
|
| 2827 |
{
|
| 2828 |
"epoch": 4.1,
|
| 2829 |
"learning_rate": 4.102333931777379e-05,
|
| 2830 |
-
"loss": 0.
|
| 2831 |
"step": 4570
|
| 2832 |
},
|
| 2833 |
{
|
| 2834 |
"epoch": 4.11,
|
| 2835 |
"learning_rate": 4.111310592459605e-05,
|
| 2836 |
-
"loss": 0.
|
| 2837 |
"step": 4580
|
| 2838 |
},
|
| 2839 |
{
|
| 2840 |
"epoch": 4.12,
|
| 2841 |
"learning_rate": 4.1202872531418315e-05,
|
| 2842 |
-
"loss": 0.
|
| 2843 |
"step": 4590
|
| 2844 |
},
|
| 2845 |
{
|
| 2846 |
"epoch": 4.13,
|
| 2847 |
"learning_rate": 4.129263913824057e-05,
|
| 2848 |
-
"loss": 0.
|
| 2849 |
"step": 4600
|
| 2850 |
},
|
| 2851 |
{
|
| 2852 |
"epoch": 4.14,
|
| 2853 |
"learning_rate": 4.1382405745062844e-05,
|
| 2854 |
-
"loss": 0.
|
| 2855 |
"step": 4610
|
| 2856 |
},
|
| 2857 |
{
|
| 2858 |
"epoch": 4.14,
|
| 2859 |
"learning_rate": 4.14721723518851e-05,
|
| 2860 |
-
"loss": 0.
|
| 2861 |
"step": 4620
|
| 2862 |
},
|
| 2863 |
{
|
| 2864 |
"epoch": 4.15,
|
| 2865 |
"learning_rate": 4.156193895870737e-05,
|
| 2866 |
-
"loss": 0.
|
| 2867 |
"step": 4630
|
| 2868 |
},
|
| 2869 |
{
|
| 2870 |
"epoch": 4.16,
|
| 2871 |
"learning_rate": 4.1651705565529625e-05,
|
| 2872 |
-
"loss": 0.
|
| 2873 |
"step": 4640
|
| 2874 |
},
|
| 2875 |
{
|
| 2876 |
"epoch": 4.17,
|
| 2877 |
"learning_rate": 4.174147217235189e-05,
|
| 2878 |
-
"loss": 0.
|
| 2879 |
"step": 4650
|
| 2880 |
},
|
| 2881 |
{
|
| 2882 |
"epoch": 4.18,
|
| 2883 |
"learning_rate": 4.183123877917415e-05,
|
| 2884 |
-
"loss": 0.
|
| 2885 |
"step": 4660
|
| 2886 |
},
|
| 2887 |
{
|
| 2888 |
"epoch": 4.19,
|
| 2889 |
"learning_rate": 4.192100538599641e-05,
|
| 2890 |
-
"loss": 0.
|
| 2891 |
"step": 4670
|
| 2892 |
},
|
| 2893 |
{
|
| 2894 |
"epoch": 4.2,
|
| 2895 |
"learning_rate": 4.2010771992818676e-05,
|
| 2896 |
-
"loss": 0.
|
| 2897 |
"step": 4680
|
| 2898 |
},
|
| 2899 |
{
|
| 2900 |
"epoch": 4.21,
|
| 2901 |
"learning_rate": 4.2100538599640934e-05,
|
| 2902 |
-
"loss": 0.
|
| 2903 |
"step": 4690
|
| 2904 |
},
|
| 2905 |
{
|
| 2906 |
"epoch": 4.22,
|
| 2907 |
"learning_rate": 4.21903052064632e-05,
|
| 2908 |
-
"loss": 0.
|
| 2909 |
"step": 4700
|
| 2910 |
},
|
| 2911 |
{
|
| 2912 |
"epoch": 4.23,
|
| 2913 |
"learning_rate": 4.228007181328546e-05,
|
| 2914 |
-
"loss": 0.
|
| 2915 |
"step": 4710
|
| 2916 |
},
|
| 2917 |
{
|
| 2918 |
"epoch": 4.23,
|
| 2919 |
"learning_rate": 4.236983842010772e-05,
|
| 2920 |
-
"loss": 0.
|
| 2921 |
"step": 4720
|
| 2922 |
},
|
| 2923 |
{
|
| 2924 |
"epoch": 4.24,
|
| 2925 |
"learning_rate": 4.2459605026929986e-05,
|
| 2926 |
-
"loss": 0.
|
| 2927 |
"step": 4730
|
| 2928 |
},
|
| 2929 |
{
|
| 2930 |
"epoch": 4.25,
|
| 2931 |
"learning_rate": 4.2549371633752244e-05,
|
| 2932 |
-
"loss": 0.
|
| 2933 |
"step": 4740
|
| 2934 |
},
|
| 2935 |
{
|
| 2936 |
"epoch": 4.26,
|
| 2937 |
"learning_rate": 4.263913824057451e-05,
|
| 2938 |
-
"loss": 0.
|
| 2939 |
"step": 4750
|
| 2940 |
},
|
| 2941 |
{
|
| 2942 |
"epoch": 4.27,
|
| 2943 |
"learning_rate": 4.272890484739677e-05,
|
| 2944 |
-
"loss": 0.
|
| 2945 |
"step": 4760
|
| 2946 |
},
|
| 2947 |
{
|
| 2948 |
"epoch": 4.28,
|
| 2949 |
"learning_rate": 4.281867145421903e-05,
|
| 2950 |
-
"loss": 0.
|
| 2951 |
"step": 4770
|
| 2952 |
},
|
| 2953 |
{
|
| 2954 |
"epoch": 4.29,
|
| 2955 |
"learning_rate": 4.2908438061041295e-05,
|
| 2956 |
-
"loss": 0.
|
| 2957 |
"step": 4780
|
| 2958 |
},
|
| 2959 |
{
|
| 2960 |
"epoch": 4.3,
|
| 2961 |
"learning_rate": 4.299820466786356e-05,
|
| 2962 |
-
"loss": 0.
|
| 2963 |
"step": 4790
|
| 2964 |
},
|
| 2965 |
{
|
| 2966 |
"epoch": 4.31,
|
| 2967 |
"learning_rate": 4.308797127468582e-05,
|
| 2968 |
-
"loss": 0.
|
| 2969 |
"step": 4800
|
| 2970 |
},
|
| 2971 |
{
|
| 2972 |
"epoch": 4.31,
|
| 2973 |
"learning_rate": 4.317773788150808e-05,
|
| 2974 |
-
"loss": 0.
|
| 2975 |
"step": 4810
|
| 2976 |
},
|
| 2977 |
{
|
| 2978 |
"epoch": 4.32,
|
| 2979 |
"learning_rate": 4.326750448833034e-05,
|
| 2980 |
-
"loss": 0.
|
| 2981 |
"step": 4820
|
| 2982 |
},
|
| 2983 |
{
|
| 2984 |
"epoch": 4.33,
|
| 2985 |
"learning_rate": 4.3357271095152605e-05,
|
| 2986 |
-
"loss": 0.
|
| 2987 |
"step": 4830
|
| 2988 |
},
|
| 2989 |
{
|
| 2990 |
"epoch": 4.34,
|
| 2991 |
"learning_rate": 4.344703770197487e-05,
|
| 2992 |
-
"loss": 0.
|
| 2993 |
"step": 4840
|
| 2994 |
},
|
| 2995 |
{
|
| 2996 |
"epoch": 4.35,
|
| 2997 |
"learning_rate": 4.353680430879713e-05,
|
| 2998 |
-
"loss": 0.
|
| 2999 |
"step": 4850
|
| 3000 |
},
|
| 3001 |
{
|
| 3002 |
"epoch": 4.36,
|
| 3003 |
"learning_rate": 4.362657091561939e-05,
|
| 3004 |
-
"loss": 0.
|
| 3005 |
"step": 4860
|
| 3006 |
},
|
| 3007 |
{
|
| 3008 |
"epoch": 4.37,
|
| 3009 |
"learning_rate": 4.371633752244165e-05,
|
| 3010 |
-
"loss": 0.
|
| 3011 |
"step": 4870
|
| 3012 |
},
|
| 3013 |
{
|
| 3014 |
"epoch": 4.38,
|
| 3015 |
"learning_rate": 4.3806104129263914e-05,
|
| 3016 |
-
"loss": 0.
|
| 3017 |
"step": 4880
|
| 3018 |
},
|
| 3019 |
{
|
| 3020 |
"epoch": 4.39,
|
| 3021 |
"learning_rate": 4.389587073608618e-05,
|
| 3022 |
-
"loss": 0.
|
| 3023 |
"step": 4890
|
| 3024 |
},
|
| 3025 |
{
|
| 3026 |
"epoch": 4.4,
|
| 3027 |
"learning_rate": 4.398563734290844e-05,
|
| 3028 |
-
"loss": 0.
|
| 3029 |
"step": 4900
|
| 3030 |
},
|
| 3031 |
{
|
| 3032 |
"epoch": 4.4,
|
| 3033 |
"learning_rate": 4.40754039497307e-05,
|
| 3034 |
-
"loss": 0.
|
| 3035 |
"step": 4910
|
| 3036 |
},
|
| 3037 |
{
|
| 3038 |
"epoch": 4.41,
|
| 3039 |
"learning_rate": 4.4165170556552966e-05,
|
| 3040 |
-
"loss": 0.
|
| 3041 |
"step": 4920
|
| 3042 |
},
|
| 3043 |
{
|
| 3044 |
"epoch": 4.42,
|
| 3045 |
"learning_rate": 4.4254937163375224e-05,
|
| 3046 |
-
"loss": 0.
|
| 3047 |
"step": 4930
|
| 3048 |
},
|
| 3049 |
{
|
| 3050 |
"epoch": 4.43,
|
| 3051 |
"learning_rate": 4.434470377019749e-05,
|
| 3052 |
-
"loss": 0.
|
| 3053 |
"step": 4940
|
| 3054 |
},
|
| 3055 |
{
|
| 3056 |
"epoch": 4.44,
|
| 3057 |
"learning_rate": 4.4434470377019746e-05,
|
| 3058 |
-
"loss": 0.
|
| 3059 |
"step": 4950
|
| 3060 |
},
|
| 3061 |
{
|
| 3062 |
"epoch": 4.45,
|
| 3063 |
"learning_rate": 4.452423698384201e-05,
|
| 3064 |
-
"loss": 0.
|
| 3065 |
"step": 4960
|
| 3066 |
},
|
| 3067 |
{
|
| 3068 |
"epoch": 4.46,
|
| 3069 |
"learning_rate": 4.4614003590664275e-05,
|
| 3070 |
-
"loss": 0.
|
| 3071 |
"step": 4970
|
| 3072 |
},
|
| 3073 |
{
|
| 3074 |
"epoch": 4.47,
|
| 3075 |
"learning_rate": 4.470377019748654e-05,
|
| 3076 |
-
"loss": 0.
|
| 3077 |
"step": 4980
|
| 3078 |
},
|
| 3079 |
{
|
| 3080 |
"epoch": 4.48,
|
| 3081 |
"learning_rate": 4.47935368043088e-05,
|
| 3082 |
-
"loss": 0.
|
| 3083 |
"step": 4990
|
| 3084 |
},
|
| 3085 |
{
|
| 3086 |
"epoch": 4.49,
|
| 3087 |
"learning_rate": 4.488330341113106e-05,
|
| 3088 |
-
"loss": 0.
|
| 3089 |
"step": 5000
|
| 3090 |
},
|
| 3091 |
{
|
| 3092 |
"epoch": 4.49,
|
| 3093 |
"learning_rate": 4.497307001795332e-05,
|
| 3094 |
-
"loss": 0.
|
| 3095 |
"step": 5010
|
| 3096 |
},
|
| 3097 |
{
|
| 3098 |
"epoch": 4.5,
|
| 3099 |
"learning_rate": 4.5062836624775585e-05,
|
| 3100 |
-
"loss": 0.
|
| 3101 |
"step": 5020
|
| 3102 |
},
|
| 3103 |
{
|
| 3104 |
"epoch": 4.51,
|
| 3105 |
"learning_rate": 4.515260323159784e-05,
|
| 3106 |
-
"loss": 0.
|
| 3107 |
"step": 5030
|
| 3108 |
},
|
| 3109 |
{
|
| 3110 |
"epoch": 4.52,
|
| 3111 |
"learning_rate": 4.5242369838420114e-05,
|
| 3112 |
-
"loss": 0.
|
| 3113 |
"step": 5040
|
| 3114 |
},
|
| 3115 |
{
|
| 3116 |
"epoch": 4.53,
|
| 3117 |
"learning_rate": 4.533213644524237e-05,
|
| 3118 |
-
"loss": 0.
|
| 3119 |
"step": 5050
|
| 3120 |
},
|
| 3121 |
{
|
| 3122 |
"epoch": 4.54,
|
| 3123 |
"learning_rate": 4.5421903052064636e-05,
|
| 3124 |
-
"loss": 0.
|
| 3125 |
"step": 5060
|
| 3126 |
},
|
| 3127 |
{
|
| 3128 |
"epoch": 4.55,
|
| 3129 |
"learning_rate": 4.5511669658886894e-05,
|
| 3130 |
-
"loss": 0.
|
| 3131 |
"step": 5070
|
| 3132 |
},
|
| 3133 |
{
|
| 3134 |
"epoch": 4.56,
|
| 3135 |
"learning_rate": 4.560143626570916e-05,
|
| 3136 |
-
"loss": 0.
|
| 3137 |
"step": 5080
|
| 3138 |
},
|
| 3139 |
{
|
| 3140 |
"epoch": 4.57,
|
| 3141 |
"learning_rate": 4.5691202872531417e-05,
|
| 3142 |
-
"loss": 0.
|
| 3143 |
"step": 5090
|
| 3144 |
},
|
| 3145 |
{
|
| 3146 |
"epoch": 4.58,
|
| 3147 |
"learning_rate": 4.578096947935368e-05,
|
| 3148 |
-
"loss": 0.
|
| 3149 |
"step": 5100
|
| 3150 |
},
|
| 3151 |
{
|
| 3152 |
"epoch": 4.58,
|
| 3153 |
"learning_rate": 4.5870736086175946e-05,
|
| 3154 |
-
"loss": 0.
|
| 3155 |
"step": 5110
|
| 3156 |
},
|
| 3157 |
{
|
| 3158 |
"epoch": 4.59,
|
| 3159 |
"learning_rate": 4.596050269299821e-05,
|
| 3160 |
-
"loss": 0.
|
| 3161 |
"step": 5120
|
| 3162 |
},
|
| 3163 |
{
|
| 3164 |
"epoch": 4.6,
|
| 3165 |
"learning_rate": 4.605026929982047e-05,
|
| 3166 |
-
"loss": 0.
|
| 3167 |
"step": 5130
|
| 3168 |
},
|
| 3169 |
{
|
| 3170 |
"epoch": 4.61,
|
| 3171 |
"learning_rate": 4.614003590664273e-05,
|
| 3172 |
-
"loss": 0.
|
| 3173 |
"step": 5140
|
| 3174 |
},
|
| 3175 |
{
|
| 3176 |
"epoch": 4.62,
|
| 3177 |
"learning_rate": 4.622980251346499e-05,
|
| 3178 |
-
"loss": 0.
|
| 3179 |
"step": 5150
|
| 3180 |
},
|
| 3181 |
{
|
| 3182 |
"epoch": 4.63,
|
| 3183 |
"learning_rate": 4.6319569120287255e-05,
|
| 3184 |
-
"loss": 0.
|
| 3185 |
"step": 5160
|
| 3186 |
},
|
| 3187 |
{
|
| 3188 |
"epoch": 4.64,
|
| 3189 |
"learning_rate": 4.640933572710951e-05,
|
| 3190 |
-
"loss": 0.
|
| 3191 |
"step": 5170
|
| 3192 |
},
|
| 3193 |
{
|
| 3194 |
"epoch": 4.65,
|
| 3195 |
"learning_rate": 4.6499102333931784e-05,
|
| 3196 |
-
"loss": 0.
|
| 3197 |
"step": 5180
|
| 3198 |
},
|
| 3199 |
{
|
| 3200 |
"epoch": 4.66,
|
| 3201 |
"learning_rate": 4.658886894075404e-05,
|
| 3202 |
-
"loss": 0.
|
| 3203 |
"step": 5190
|
| 3204 |
},
|
| 3205 |
{
|
| 3206 |
"epoch": 4.66,
|
| 3207 |
"learning_rate": 4.667863554757631e-05,
|
| 3208 |
-
"loss": 0.
|
| 3209 |
"step": 5200
|
| 3210 |
},
|
| 3211 |
{
|
| 3212 |
"epoch": 4.67,
|
| 3213 |
"learning_rate": 4.6768402154398565e-05,
|
| 3214 |
-
"loss": 0.
|
| 3215 |
"step": 5210
|
| 3216 |
},
|
| 3217 |
{
|
| 3218 |
"epoch": 4.68,
|
| 3219 |
"learning_rate": 4.685816876122083e-05,
|
| 3220 |
-
"loss": 0.
|
| 3221 |
"step": 5220
|
| 3222 |
},
|
| 3223 |
{
|
| 3224 |
"epoch": 4.69,
|
| 3225 |
"learning_rate": 4.694793536804309e-05,
|
| 3226 |
-
"loss": 0.
|
| 3227 |
"step": 5230
|
| 3228 |
},
|
| 3229 |
{
|
| 3230 |
"epoch": 4.7,
|
| 3231 |
"learning_rate": 4.703770197486535e-05,
|
| 3232 |
-
"loss": 0.
|
| 3233 |
"step": 5240
|
| 3234 |
},
|
| 3235 |
{
|
| 3236 |
"epoch": 4.71,
|
| 3237 |
"learning_rate": 4.7127468581687616e-05,
|
| 3238 |
-
"loss": 0.
|
| 3239 |
"step": 5250
|
| 3240 |
},
|
| 3241 |
{
|
| 3242 |
"epoch": 4.72,
|
| 3243 |
"learning_rate": 4.721723518850988e-05,
|
| 3244 |
-
"loss": 0.
|
| 3245 |
"step": 5260
|
| 3246 |
},
|
| 3247 |
{
|
| 3248 |
"epoch": 4.73,
|
| 3249 |
"learning_rate": 4.730700179533214e-05,
|
| 3250 |
-
"loss": 0.
|
| 3251 |
"step": 5270
|
| 3252 |
},
|
| 3253 |
{
|
| 3254 |
"epoch": 4.74,
|
| 3255 |
"learning_rate": 4.7396768402154403e-05,
|
| 3256 |
-
"loss": 0.
|
| 3257 |
"step": 5280
|
| 3258 |
},
|
| 3259 |
{
|
| 3260 |
"epoch": 4.75,
|
| 3261 |
"learning_rate": 4.748653500897666e-05,
|
| 3262 |
-
"loss": 0.
|
| 3263 |
"step": 5290
|
| 3264 |
},
|
| 3265 |
{
|
| 3266 |
"epoch": 4.75,
|
| 3267 |
"learning_rate": 4.7576301615798926e-05,
|
| 3268 |
-
"loss": 0.
|
| 3269 |
"step": 5300
|
| 3270 |
},
|
| 3271 |
{
|
| 3272 |
"epoch": 4.76,
|
| 3273 |
"learning_rate": 4.7666068222621184e-05,
|
| 3274 |
-
"loss": 0.
|
| 3275 |
"step": 5310
|
| 3276 |
},
|
| 3277 |
{
|
| 3278 |
"epoch": 4.77,
|
| 3279 |
"learning_rate": 4.775583482944345e-05,
|
| 3280 |
-
"loss": 0.
|
| 3281 |
"step": 5320
|
| 3282 |
},
|
| 3283 |
{
|
| 3284 |
"epoch": 4.78,
|
| 3285 |
"learning_rate": 4.784560143626571e-05,
|
| 3286 |
-
"loss": 0.
|
| 3287 |
"step": 5330
|
| 3288 |
},
|
| 3289 |
{
|
| 3290 |
"epoch": 4.79,
|
| 3291 |
"learning_rate": 4.793536804308798e-05,
|
| 3292 |
-
"loss": 0.
|
| 3293 |
"step": 5340
|
| 3294 |
},
|
| 3295 |
{
|
| 3296 |
"epoch": 4.8,
|
| 3297 |
"learning_rate": 4.8025134649910235e-05,
|
| 3298 |
-
"loss": 0.
|
| 3299 |
"step": 5350
|
| 3300 |
},
|
| 3301 |
{
|
| 3302 |
"epoch": 4.81,
|
| 3303 |
"learning_rate": 4.81149012567325e-05,
|
| 3304 |
-
"loss": 0.
|
| 3305 |
"step": 5360
|
| 3306 |
},
|
| 3307 |
{
|
| 3308 |
"epoch": 4.82,
|
| 3309 |
"learning_rate": 4.820466786355476e-05,
|
| 3310 |
-
"loss": 0.
|
| 3311 |
"step": 5370
|
| 3312 |
},
|
| 3313 |
{
|
| 3314 |
"epoch": 4.83,
|
| 3315 |
"learning_rate": 4.829443447037702e-05,
|
| 3316 |
-
"loss": 0.
|
| 3317 |
"step": 5380
|
| 3318 |
},
|
| 3319 |
{
|
| 3320 |
"epoch": 4.84,
|
| 3321 |
"learning_rate": 4.838420107719928e-05,
|
| 3322 |
-
"loss": 0.
|
| 3323 |
"step": 5390
|
| 3324 |
},
|
| 3325 |
{
|
| 3326 |
"epoch": 4.84,
|
| 3327 |
"learning_rate": 4.847396768402155e-05,
|
| 3328 |
-
"loss": 0.
|
| 3329 |
"step": 5400
|
| 3330 |
},
|
| 3331 |
{
|
| 3332 |
"epoch": 4.85,
|
| 3333 |
"learning_rate": 4.856373429084381e-05,
|
| 3334 |
-
"loss": 0.
|
| 3335 |
"step": 5410
|
| 3336 |
},
|
| 3337 |
{
|
| 3338 |
"epoch": 4.86,
|
| 3339 |
"learning_rate": 4.8653500897666074e-05,
|
| 3340 |
-
"loss": 0.
|
| 3341 |
"step": 5420
|
| 3342 |
},
|
| 3343 |
{
|
| 3344 |
"epoch": 4.87,
|
| 3345 |
"learning_rate": 4.874326750448833e-05,
|
| 3346 |
-
"loss": 0.
|
| 3347 |
"step": 5430
|
| 3348 |
},
|
| 3349 |
{
|
| 3350 |
"epoch": 4.88,
|
| 3351 |
"learning_rate": 4.8833034111310596e-05,
|
| 3352 |
-
"loss": 0.
|
| 3353 |
"step": 5440
|
| 3354 |
},
|
| 3355 |
{
|
| 3356 |
"epoch": 4.89,
|
| 3357 |
"learning_rate": 4.8922800718132854e-05,
|
| 3358 |
-
"loss": 0.
|
| 3359 |
"step": 5450
|
| 3360 |
},
|
| 3361 |
{
|
| 3362 |
"epoch": 4.9,
|
| 3363 |
"learning_rate": 4.901256732495512e-05,
|
| 3364 |
-
"loss": 0.
|
| 3365 |
"step": 5460
|
| 3366 |
},
|
| 3367 |
{
|
| 3368 |
"epoch": 4.91,
|
| 3369 |
"learning_rate": 4.9102333931777383e-05,
|
| 3370 |
-
"loss": 0.
|
| 3371 |
"step": 5470
|
| 3372 |
},
|
| 3373 |
{
|
| 3374 |
"epoch": 4.92,
|
| 3375 |
"learning_rate": 4.919210053859965e-05,
|
| 3376 |
-
"loss": 0.
|
| 3377 |
"step": 5480
|
| 3378 |
},
|
| 3379 |
{
|
| 3380 |
"epoch": 4.92,
|
| 3381 |
"learning_rate": 4.9281867145421906e-05,
|
| 3382 |
-
"loss": 0.
|
| 3383 |
"step": 5490
|
| 3384 |
},
|
| 3385 |
{
|
| 3386 |
"epoch": 4.93,
|
| 3387 |
"learning_rate": 4.937163375224417e-05,
|
| 3388 |
-
"loss": 0.
|
| 3389 |
"step": 5500
|
| 3390 |
},
|
| 3391 |
{
|
| 3392 |
"epoch": 4.94,
|
| 3393 |
"learning_rate": 4.946140035906643e-05,
|
| 3394 |
-
"loss": 0.
|
| 3395 |
"step": 5510
|
| 3396 |
},
|
| 3397 |
{
|
| 3398 |
"epoch": 4.95,
|
| 3399 |
"learning_rate": 4.955116696588869e-05,
|
| 3400 |
-
"loss": 0.
|
| 3401 |
"step": 5520
|
| 3402 |
},
|
| 3403 |
{
|
| 3404 |
"epoch": 4.96,
|
| 3405 |
"learning_rate": 4.964093357271095e-05,
|
| 3406 |
-
"loss": 0.
|
| 3407 |
"step": 5530
|
| 3408 |
},
|
| 3409 |
{
|
| 3410 |
"epoch": 4.97,
|
| 3411 |
"learning_rate": 4.9730700179533215e-05,
|
| 3412 |
-
"loss": 0.
|
| 3413 |
"step": 5540
|
| 3414 |
},
|
| 3415 |
{
|
| 3416 |
"epoch": 4.98,
|
| 3417 |
"learning_rate": 4.982046678635548e-05,
|
| 3418 |
-
"loss": 0.
|
| 3419 |
"step": 5550
|
| 3420 |
},
|
| 3421 |
{
|
| 3422 |
"epoch": 4.99,
|
| 3423 |
"learning_rate": 4.991023339317774e-05,
|
| 3424 |
-
"loss": 0.
|
| 3425 |
"step": 5560
|
| 3426 |
},
|
| 3427 |
{
|
| 3428 |
"epoch": 5.0,
|
| 3429 |
"learning_rate": 5e-05,
|
| 3430 |
-
"loss": 0.
|
| 3431 |
"step": 5570
|
| 3432 |
},
|
| 3433 |
{
|
| 3434 |
"epoch": 5.0,
|
| 3435 |
"eval_accuracy": {
|
| 3436 |
-
"accuracy": 0.
|
| 3437 |
},
|
| 3438 |
"eval_f1": {
|
| 3439 |
-
"f1": 0.
|
| 3440 |
},
|
| 3441 |
-
"eval_loss": 0.
|
| 3442 |
"eval_precision": {
|
| 3443 |
-
"precision": 0.
|
| 3444 |
},
|
| 3445 |
"eval_recall": {
|
| 3446 |
-
"recall": 0.
|
| 3447 |
},
|
| 3448 |
-
"eval_runtime":
|
| 3449 |
-
"eval_samples_per_second":
|
| 3450 |
-
"eval_steps_per_second": 6.
|
| 3451 |
"step": 5573
|
| 3452 |
}
|
| 3453 |
],
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.1283472627401352,
|
| 3 |
"best_model_checkpoint": "vit-base-patch16-224-in21k-crack-detectorVITmain50epochs\\checkpoint-5573",
|
| 4 |
"epoch": 4.999327203408836,
|
| 5 |
"eval_steps": 500,
|
|
|
|
| 11 |
{
|
| 12 |
"epoch": 0.01,
|
| 13 |
"learning_rate": 8.976660682226213e-08,
|
| 14 |
+
"loss": 1.3961,
|
| 15 |
"step": 10
|
| 16 |
},
|
| 17 |
{
|
| 18 |
"epoch": 0.02,
|
| 19 |
"learning_rate": 1.7953321364452426e-07,
|
| 20 |
+
"loss": 1.395,
|
| 21 |
"step": 20
|
| 22 |
},
|
| 23 |
{
|
| 24 |
"epoch": 0.03,
|
| 25 |
"learning_rate": 2.692998204667864e-07,
|
| 26 |
+
"loss": 1.3978,
|
| 27 |
"step": 30
|
| 28 |
},
|
| 29 |
{
|
| 30 |
"epoch": 0.04,
|
| 31 |
"learning_rate": 3.590664272890485e-07,
|
| 32 |
+
"loss": 1.3963,
|
| 33 |
"step": 40
|
| 34 |
},
|
| 35 |
{
|
| 36 |
"epoch": 0.04,
|
| 37 |
"learning_rate": 4.4883303411131064e-07,
|
| 38 |
+
"loss": 1.3907,
|
| 39 |
"step": 50
|
| 40 |
},
|
| 41 |
{
|
| 42 |
"epoch": 0.05,
|
| 43 |
"learning_rate": 5.385996409335728e-07,
|
| 44 |
+
"loss": 1.396,
|
| 45 |
"step": 60
|
| 46 |
},
|
| 47 |
{
|
| 48 |
"epoch": 0.06,
|
| 49 |
"learning_rate": 6.283662477558349e-07,
|
| 50 |
+
"loss": 1.3885,
|
| 51 |
"step": 70
|
| 52 |
},
|
| 53 |
{
|
| 54 |
"epoch": 0.07,
|
| 55 |
"learning_rate": 7.18132854578097e-07,
|
| 56 |
+
"loss": 1.387,
|
| 57 |
"step": 80
|
| 58 |
},
|
| 59 |
{
|
| 60 |
"epoch": 0.08,
|
| 61 |
"learning_rate": 8.078994614003591e-07,
|
| 62 |
+
"loss": 1.3894,
|
| 63 |
"step": 90
|
| 64 |
},
|
| 65 |
{
|
| 66 |
"epoch": 0.09,
|
| 67 |
"learning_rate": 8.976660682226213e-07,
|
| 68 |
+
"loss": 1.3806,
|
| 69 |
"step": 100
|
| 70 |
},
|
| 71 |
{
|
| 72 |
"epoch": 0.1,
|
| 73 |
"learning_rate": 9.874326750448833e-07,
|
| 74 |
+
"loss": 1.383,
|
| 75 |
"step": 110
|
| 76 |
},
|
| 77 |
{
|
| 78 |
"epoch": 0.11,
|
| 79 |
"learning_rate": 1.0771992818671456e-06,
|
| 80 |
+
"loss": 1.3764,
|
| 81 |
"step": 120
|
| 82 |
},
|
| 83 |
{
|
| 84 |
"epoch": 0.12,
|
| 85 |
"learning_rate": 1.1669658886894075e-06,
|
| 86 |
+
"loss": 1.3688,
|
| 87 |
"step": 130
|
| 88 |
},
|
| 89 |
{
|
| 90 |
"epoch": 0.13,
|
| 91 |
"learning_rate": 1.2567324955116697e-06,
|
| 92 |
+
"loss": 1.3708,
|
| 93 |
"step": 140
|
| 94 |
},
|
| 95 |
{
|
| 96 |
"epoch": 0.13,
|
| 97 |
"learning_rate": 1.3464991023339318e-06,
|
| 98 |
+
"loss": 1.366,
|
| 99 |
"step": 150
|
| 100 |
},
|
| 101 |
{
|
| 102 |
"epoch": 0.14,
|
| 103 |
"learning_rate": 1.436265709156194e-06,
|
| 104 |
+
"loss": 1.371,
|
| 105 |
"step": 160
|
| 106 |
},
|
| 107 |
{
|
| 108 |
"epoch": 0.15,
|
| 109 |
"learning_rate": 1.5260323159784561e-06,
|
| 110 |
+
"loss": 1.3558,
|
| 111 |
"step": 170
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"epoch": 0.16,
|
| 115 |
"learning_rate": 1.6157989228007182e-06,
|
| 116 |
+
"loss": 1.3584,
|
| 117 |
"step": 180
|
| 118 |
},
|
| 119 |
{
|
| 120 |
"epoch": 0.17,
|
| 121 |
"learning_rate": 1.7055655296229805e-06,
|
| 122 |
+
"loss": 1.3409,
|
| 123 |
"step": 190
|
| 124 |
},
|
| 125 |
{
|
| 126 |
"epoch": 0.18,
|
| 127 |
"learning_rate": 1.7953321364452425e-06,
|
| 128 |
+
"loss": 1.346,
|
| 129 |
"step": 200
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.19,
|
| 133 |
"learning_rate": 1.8850987432675046e-06,
|
| 134 |
+
"loss": 1.3355,
|
| 135 |
"step": 210
|
| 136 |
},
|
| 137 |
{
|
| 138 |
"epoch": 0.2,
|
| 139 |
"learning_rate": 1.9748653500897667e-06,
|
| 140 |
+
"loss": 1.3323,
|
| 141 |
"step": 220
|
| 142 |
},
|
| 143 |
{
|
| 144 |
"epoch": 0.21,
|
| 145 |
"learning_rate": 2.064631956912029e-06,
|
| 146 |
+
"loss": 1.3231,
|
| 147 |
"step": 230
|
| 148 |
},
|
| 149 |
{
|
| 150 |
"epoch": 0.22,
|
| 151 |
"learning_rate": 2.1543985637342912e-06,
|
| 152 |
+
"loss": 1.3129,
|
| 153 |
"step": 240
|
| 154 |
},
|
| 155 |
{
|
| 156 |
"epoch": 0.22,
|
| 157 |
"learning_rate": 2.244165170556553e-06,
|
| 158 |
+
"loss": 1.3178,
|
| 159 |
"step": 250
|
| 160 |
},
|
| 161 |
{
|
| 162 |
"epoch": 0.23,
|
| 163 |
"learning_rate": 2.333931777378815e-06,
|
| 164 |
+
"loss": 1.2953,
|
| 165 |
"step": 260
|
| 166 |
},
|
| 167 |
{
|
| 168 |
"epoch": 0.24,
|
| 169 |
"learning_rate": 2.423698384201077e-06,
|
| 170 |
+
"loss": 1.2938,
|
| 171 |
"step": 270
|
| 172 |
},
|
| 173 |
{
|
| 174 |
"epoch": 0.25,
|
| 175 |
"learning_rate": 2.5134649910233395e-06,
|
| 176 |
+
"loss": 1.2849,
|
| 177 |
"step": 280
|
| 178 |
},
|
| 179 |
{
|
| 180 |
"epoch": 0.26,
|
| 181 |
"learning_rate": 2.6032315978456015e-06,
|
| 182 |
+
"loss": 1.2813,
|
| 183 |
"step": 290
|
| 184 |
},
|
| 185 |
{
|
| 186 |
"epoch": 0.27,
|
| 187 |
"learning_rate": 2.6929982046678636e-06,
|
| 188 |
+
"loss": 1.2811,
|
| 189 |
"step": 300
|
| 190 |
},
|
| 191 |
{
|
| 192 |
"epoch": 0.28,
|
| 193 |
"learning_rate": 2.7827648114901257e-06,
|
| 194 |
+
"loss": 1.2526,
|
| 195 |
"step": 310
|
| 196 |
},
|
| 197 |
{
|
| 198 |
"epoch": 0.29,
|
| 199 |
"learning_rate": 2.872531418312388e-06,
|
| 200 |
+
"loss": 1.2438,
|
| 201 |
"step": 320
|
| 202 |
},
|
| 203 |
{
|
| 204 |
"epoch": 0.3,
|
| 205 |
"learning_rate": 2.9622980251346502e-06,
|
| 206 |
+
"loss": 1.2145,
|
| 207 |
"step": 330
|
| 208 |
},
|
| 209 |
{
|
| 210 |
"epoch": 0.31,
|
| 211 |
"learning_rate": 3.0520646319569123e-06,
|
| 212 |
+
"loss": 1.2146,
|
| 213 |
"step": 340
|
| 214 |
},
|
| 215 |
{
|
| 216 |
"epoch": 0.31,
|
| 217 |
"learning_rate": 3.1418312387791743e-06,
|
| 218 |
+
"loss": 1.1883,
|
| 219 |
"step": 350
|
| 220 |
},
|
| 221 |
{
|
| 222 |
"epoch": 0.32,
|
| 223 |
"learning_rate": 3.2315978456014364e-06,
|
| 224 |
+
"loss": 1.1891,
|
| 225 |
"step": 360
|
| 226 |
},
|
| 227 |
{
|
| 228 |
"epoch": 0.33,
|
| 229 |
"learning_rate": 3.321364452423698e-06,
|
| 230 |
+
"loss": 1.1723,
|
| 231 |
"step": 370
|
| 232 |
},
|
| 233 |
{
|
| 234 |
"epoch": 0.34,
|
| 235 |
"learning_rate": 3.411131059245961e-06,
|
| 236 |
+
"loss": 1.1417,
|
| 237 |
"step": 380
|
| 238 |
},
|
| 239 |
{
|
| 240 |
"epoch": 0.35,
|
| 241 |
"learning_rate": 3.500897666068223e-06,
|
| 242 |
+
"loss": 1.1338,
|
| 243 |
"step": 390
|
| 244 |
},
|
| 245 |
{
|
| 246 |
"epoch": 0.36,
|
| 247 |
"learning_rate": 3.590664272890485e-06,
|
| 248 |
+
"loss": 1.1004,
|
| 249 |
"step": 400
|
| 250 |
},
|
| 251 |
{
|
| 252 |
"epoch": 0.37,
|
| 253 |
"learning_rate": 3.680430879712747e-06,
|
| 254 |
+
"loss": 1.0873,
|
| 255 |
"step": 410
|
| 256 |
},
|
| 257 |
{
|
| 258 |
"epoch": 0.38,
|
| 259 |
"learning_rate": 3.770197486535009e-06,
|
| 260 |
+
"loss": 1.0358,
|
| 261 |
"step": 420
|
| 262 |
},
|
| 263 |
{
|
| 264 |
"epoch": 0.39,
|
| 265 |
"learning_rate": 3.859964093357271e-06,
|
| 266 |
+
"loss": 1.0517,
|
| 267 |
"step": 430
|
| 268 |
},
|
| 269 |
{
|
| 270 |
"epoch": 0.39,
|
| 271 |
"learning_rate": 3.949730700179533e-06,
|
| 272 |
+
"loss": 1.0139,
|
| 273 |
"step": 440
|
| 274 |
},
|
| 275 |
{
|
| 276 |
"epoch": 0.4,
|
| 277 |
"learning_rate": 4.039497307001795e-06,
|
| 278 |
+
"loss": 1.0294,
|
| 279 |
"step": 450
|
| 280 |
},
|
| 281 |
{
|
| 282 |
"epoch": 0.41,
|
| 283 |
"learning_rate": 4.129263913824058e-06,
|
| 284 |
+
"loss": 1.0171,
|
| 285 |
"step": 460
|
| 286 |
},
|
| 287 |
{
|
| 288 |
"epoch": 0.42,
|
| 289 |
"learning_rate": 4.21903052064632e-06,
|
| 290 |
+
"loss": 0.9827,
|
| 291 |
"step": 470
|
| 292 |
},
|
| 293 |
{
|
| 294 |
"epoch": 0.43,
|
| 295 |
"learning_rate": 4.3087971274685824e-06,
|
| 296 |
+
"loss": 0.9382,
|
| 297 |
"step": 480
|
| 298 |
},
|
| 299 |
{
|
| 300 |
"epoch": 0.44,
|
| 301 |
"learning_rate": 4.3985637342908445e-06,
|
| 302 |
+
"loss": 0.9145,
|
| 303 |
"step": 490
|
| 304 |
},
|
| 305 |
{
|
| 306 |
"epoch": 0.45,
|
| 307 |
"learning_rate": 4.488330341113106e-06,
|
| 308 |
+
"loss": 0.9276,
|
| 309 |
"step": 500
|
| 310 |
},
|
| 311 |
{
|
| 312 |
"epoch": 0.46,
|
| 313 |
"learning_rate": 4.578096947935368e-06,
|
| 314 |
+
"loss": 0.8801,
|
| 315 |
"step": 510
|
| 316 |
},
|
| 317 |
{
|
| 318 |
"epoch": 0.47,
|
| 319 |
"learning_rate": 4.66786355475763e-06,
|
| 320 |
+
"loss": 0.8906,
|
| 321 |
"step": 520
|
| 322 |
},
|
| 323 |
{
|
| 324 |
"epoch": 0.48,
|
| 325 |
"learning_rate": 4.757630161579892e-06,
|
| 326 |
+
"loss": 0.9025,
|
| 327 |
"step": 530
|
| 328 |
},
|
| 329 |
{
|
| 330 |
"epoch": 0.48,
|
| 331 |
"learning_rate": 4.847396768402154e-06,
|
| 332 |
+
"loss": 0.8621,
|
| 333 |
"step": 540
|
| 334 |
},
|
| 335 |
{
|
| 336 |
"epoch": 0.49,
|
| 337 |
"learning_rate": 4.937163375224417e-06,
|
| 338 |
+
"loss": 0.8653,
|
| 339 |
"step": 550
|
| 340 |
},
|
| 341 |
{
|
| 342 |
"epoch": 0.5,
|
| 343 |
"learning_rate": 5.026929982046679e-06,
|
| 344 |
+
"loss": 0.8292,
|
| 345 |
"step": 560
|
| 346 |
},
|
| 347 |
{
|
| 348 |
"epoch": 0.51,
|
| 349 |
"learning_rate": 5.116696588868941e-06,
|
| 350 |
+
"loss": 0.8195,
|
| 351 |
"step": 570
|
| 352 |
},
|
| 353 |
{
|
| 354 |
"epoch": 0.52,
|
| 355 |
"learning_rate": 5.206463195691203e-06,
|
| 356 |
+
"loss": 0.8026,
|
| 357 |
"step": 580
|
| 358 |
},
|
| 359 |
{
|
| 360 |
"epoch": 0.53,
|
| 361 |
"learning_rate": 5.296229802513465e-06,
|
| 362 |
+
"loss": 0.841,
|
| 363 |
"step": 590
|
| 364 |
},
|
| 365 |
{
|
| 366 |
"epoch": 0.54,
|
| 367 |
"learning_rate": 5.385996409335727e-06,
|
| 368 |
+
"loss": 0.778,
|
| 369 |
"step": 600
|
| 370 |
},
|
| 371 |
{
|
| 372 |
"epoch": 0.55,
|
| 373 |
"learning_rate": 5.475763016157989e-06,
|
| 374 |
+
"loss": 0.7706,
|
| 375 |
"step": 610
|
| 376 |
},
|
| 377 |
{
|
| 378 |
"epoch": 0.56,
|
| 379 |
"learning_rate": 5.565529622980251e-06,
|
| 380 |
+
"loss": 0.733,
|
| 381 |
"step": 620
|
| 382 |
},
|
| 383 |
{
|
| 384 |
"epoch": 0.57,
|
| 385 |
"learning_rate": 5.655296229802514e-06,
|
| 386 |
+
"loss": 0.7437,
|
| 387 |
"step": 630
|
| 388 |
},
|
| 389 |
{
|
| 390 |
"epoch": 0.57,
|
| 391 |
"learning_rate": 5.745062836624776e-06,
|
| 392 |
+
"loss": 0.7644,
|
| 393 |
"step": 640
|
| 394 |
},
|
| 395 |
{
|
| 396 |
"epoch": 0.58,
|
| 397 |
"learning_rate": 5.834829443447038e-06,
|
| 398 |
+
"loss": 0.7125,
|
| 399 |
"step": 650
|
| 400 |
},
|
| 401 |
{
|
| 402 |
"epoch": 0.59,
|
| 403 |
"learning_rate": 5.9245960502693004e-06,
|
| 404 |
+
"loss": 0.7146,
|
| 405 |
"step": 660
|
| 406 |
},
|
| 407 |
{
|
| 408 |
"epoch": 0.6,
|
| 409 |
"learning_rate": 6.0143626570915625e-06,
|
| 410 |
+
"loss": 0.7201,
|
| 411 |
"step": 670
|
| 412 |
},
|
| 413 |
{
|
| 414 |
"epoch": 0.61,
|
| 415 |
"learning_rate": 6.1041292639138246e-06,
|
| 416 |
+
"loss": 0.6851,
|
| 417 |
"step": 680
|
| 418 |
},
|
| 419 |
{
|
| 420 |
"epoch": 0.62,
|
| 421 |
"learning_rate": 6.193895870736087e-06,
|
| 422 |
+
"loss": 0.6417,
|
| 423 |
"step": 690
|
| 424 |
},
|
| 425 |
{
|
| 426 |
"epoch": 0.63,
|
| 427 |
"learning_rate": 6.283662477558349e-06,
|
| 428 |
+
"loss": 0.6656,
|
| 429 |
"step": 700
|
| 430 |
},
|
| 431 |
{
|
| 432 |
"epoch": 0.64,
|
| 433 |
"learning_rate": 6.373429084380611e-06,
|
| 434 |
+
"loss": 0.6122,
|
| 435 |
"step": 710
|
| 436 |
},
|
| 437 |
{
|
| 438 |
"epoch": 0.65,
|
| 439 |
"learning_rate": 6.463195691202873e-06,
|
| 440 |
+
"loss": 0.6536,
|
| 441 |
"step": 720
|
| 442 |
},
|
| 443 |
{
|
| 444 |
"epoch": 0.65,
|
| 445 |
"learning_rate": 6.552962298025135e-06,
|
| 446 |
+
"loss": 0.6315,
|
| 447 |
"step": 730
|
| 448 |
},
|
| 449 |
{
|
| 450 |
"epoch": 0.66,
|
| 451 |
"learning_rate": 6.642728904847396e-06,
|
| 452 |
+
"loss": 0.6618,
|
| 453 |
"step": 740
|
| 454 |
},
|
| 455 |
{
|
| 456 |
"epoch": 0.67,
|
| 457 |
"learning_rate": 6.732495511669659e-06,
|
| 458 |
+
"loss": 0.6346,
|
| 459 |
"step": 750
|
| 460 |
},
|
| 461 |
{
|
| 462 |
"epoch": 0.68,
|
| 463 |
"learning_rate": 6.822262118491922e-06,
|
| 464 |
+
"loss": 0.6866,
|
| 465 |
"step": 760
|
| 466 |
},
|
| 467 |
{
|
| 468 |
"epoch": 0.69,
|
| 469 |
"learning_rate": 6.912028725314183e-06,
|
| 470 |
+
"loss": 0.6093,
|
| 471 |
"step": 770
|
| 472 |
},
|
| 473 |
{
|
| 474 |
"epoch": 0.7,
|
| 475 |
"learning_rate": 7.001795332136446e-06,
|
| 476 |
+
"loss": 0.6116,
|
| 477 |
"step": 780
|
| 478 |
},
|
| 479 |
{
|
| 480 |
"epoch": 0.71,
|
| 481 |
"learning_rate": 7.091561938958707e-06,
|
| 482 |
+
"loss": 0.602,
|
| 483 |
"step": 790
|
| 484 |
},
|
| 485 |
{
|
| 486 |
"epoch": 0.72,
|
| 487 |
"learning_rate": 7.18132854578097e-06,
|
| 488 |
+
"loss": 0.6266,
|
| 489 |
"step": 800
|
| 490 |
},
|
| 491 |
{
|
| 492 |
"epoch": 0.73,
|
| 493 |
"learning_rate": 7.271095152603231e-06,
|
| 494 |
+
"loss": 0.6082,
|
| 495 |
"step": 810
|
| 496 |
},
|
| 497 |
{
|
| 498 |
"epoch": 0.74,
|
| 499 |
"learning_rate": 7.360861759425494e-06,
|
| 500 |
+
"loss": 0.6194,
|
| 501 |
"step": 820
|
| 502 |
},
|
| 503 |
{
|
| 504 |
"epoch": 0.74,
|
| 505 |
"learning_rate": 7.4506283662477555e-06,
|
| 506 |
+
"loss": 0.6277,
|
| 507 |
"step": 830
|
| 508 |
},
|
| 509 |
{
|
| 510 |
"epoch": 0.75,
|
| 511 |
"learning_rate": 7.540394973070018e-06,
|
| 512 |
+
"loss": 0.609,
|
| 513 |
"step": 840
|
| 514 |
},
|
| 515 |
{
|
| 516 |
"epoch": 0.76,
|
| 517 |
"learning_rate": 7.630161579892281e-06,
|
| 518 |
+
"loss": 0.6252,
|
| 519 |
"step": 850
|
| 520 |
},
|
| 521 |
{
|
| 522 |
"epoch": 0.77,
|
| 523 |
"learning_rate": 7.719928186714543e-06,
|
| 524 |
+
"loss": 0.5813,
|
| 525 |
"step": 860
|
| 526 |
},
|
| 527 |
{
|
| 528 |
"epoch": 0.78,
|
| 529 |
"learning_rate": 7.809694793536805e-06,
|
| 530 |
+
"loss": 0.5799,
|
| 531 |
"step": 870
|
| 532 |
},
|
| 533 |
{
|
| 534 |
"epoch": 0.79,
|
| 535 |
"learning_rate": 7.899461400359067e-06,
|
| 536 |
+
"loss": 0.6398,
|
| 537 |
"step": 880
|
| 538 |
},
|
| 539 |
{
|
| 540 |
"epoch": 0.8,
|
| 541 |
"learning_rate": 7.98922800718133e-06,
|
| 542 |
+
"loss": 0.5549,
|
| 543 |
"step": 890
|
| 544 |
},
|
| 545 |
{
|
| 546 |
"epoch": 0.81,
|
| 547 |
"learning_rate": 8.07899461400359e-06,
|
| 548 |
+
"loss": 0.6023,
|
| 549 |
"step": 900
|
| 550 |
},
|
| 551 |
{
|
| 552 |
"epoch": 0.82,
|
| 553 |
"learning_rate": 8.168761220825854e-06,
|
| 554 |
+
"loss": 0.582,
|
| 555 |
"step": 910
|
| 556 |
},
|
| 557 |
{
|
| 558 |
"epoch": 0.83,
|
| 559 |
"learning_rate": 8.258527827648117e-06,
|
| 560 |
+
"loss": 0.544,
|
| 561 |
"step": 920
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.83,
|
| 565 |
"learning_rate": 8.348294434470378e-06,
|
| 566 |
+
"loss": 0.5912,
|
| 567 |
"step": 930
|
| 568 |
},
|
| 569 |
{
|
| 570 |
"epoch": 0.84,
|
| 571 |
"learning_rate": 8.43806104129264e-06,
|
| 572 |
+
"loss": 0.5461,
|
| 573 |
"step": 940
|
| 574 |
},
|
| 575 |
{
|
| 576 |
"epoch": 0.85,
|
| 577 |
"learning_rate": 8.527827648114902e-06,
|
| 578 |
+
"loss": 0.5238,
|
| 579 |
"step": 950
|
| 580 |
},
|
| 581 |
{
|
| 582 |
"epoch": 0.86,
|
| 583 |
"learning_rate": 8.617594254937165e-06,
|
| 584 |
+
"loss": 0.5558,
|
| 585 |
"step": 960
|
| 586 |
},
|
| 587 |
{
|
| 588 |
"epoch": 0.87,
|
| 589 |
"learning_rate": 8.707360861759426e-06,
|
| 590 |
+
"loss": 0.5371,
|
| 591 |
"step": 970
|
| 592 |
},
|
| 593 |
{
|
| 594 |
"epoch": 0.88,
|
| 595 |
"learning_rate": 8.797127468581689e-06,
|
| 596 |
+
"loss": 0.5529,
|
| 597 |
"step": 980
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.89,
|
| 601 |
"learning_rate": 8.88689407540395e-06,
|
| 602 |
+
"loss": 0.5691,
|
| 603 |
"step": 990
|
| 604 |
},
|
| 605 |
{
|
| 606 |
"epoch": 0.9,
|
| 607 |
"learning_rate": 8.976660682226211e-06,
|
| 608 |
+
"loss": 0.5567,
|
| 609 |
"step": 1000
|
| 610 |
},
|
| 611 |
{
|
| 612 |
"epoch": 0.91,
|
| 613 |
"learning_rate": 9.066427289048474e-06,
|
| 614 |
+
"loss": 0.5135,
|
| 615 |
"step": 1010
|
| 616 |
},
|
| 617 |
{
|
| 618 |
"epoch": 0.92,
|
| 619 |
"learning_rate": 9.156193895870736e-06,
|
| 620 |
+
"loss": 0.5099,
|
| 621 |
"step": 1020
|
| 622 |
},
|
| 623 |
{
|
| 624 |
"epoch": 0.92,
|
| 625 |
"learning_rate": 9.245960502692998e-06,
|
| 626 |
+
"loss": 0.5422,
|
| 627 |
"step": 1030
|
| 628 |
},
|
| 629 |
{
|
| 630 |
"epoch": 0.93,
|
| 631 |
"learning_rate": 9.33572710951526e-06,
|
| 632 |
+
"loss": 0.5106,
|
| 633 |
"step": 1040
|
| 634 |
},
|
| 635 |
{
|
| 636 |
"epoch": 0.94,
|
| 637 |
"learning_rate": 9.425493716337523e-06,
|
| 638 |
+
"loss": 0.4892,
|
| 639 |
"step": 1050
|
| 640 |
},
|
| 641 |
{
|
| 642 |
"epoch": 0.95,
|
| 643 |
"learning_rate": 9.515260323159784e-06,
|
| 644 |
+
"loss": 0.4541,
|
| 645 |
"step": 1060
|
| 646 |
},
|
| 647 |
{
|
| 648 |
"epoch": 0.96,
|
| 649 |
"learning_rate": 9.605026929982047e-06,
|
| 650 |
+
"loss": 0.4675,
|
| 651 |
"step": 1070
|
| 652 |
},
|
| 653 |
{
|
| 654 |
"epoch": 0.97,
|
| 655 |
"learning_rate": 9.694793536804308e-06,
|
| 656 |
+
"loss": 0.5123,
|
| 657 |
"step": 1080
|
| 658 |
},
|
| 659 |
{
|
| 660 |
"epoch": 0.98,
|
| 661 |
"learning_rate": 9.784560143626571e-06,
|
| 662 |
+
"loss": 0.5089,
|
| 663 |
"step": 1090
|
| 664 |
},
|
| 665 |
{
|
| 666 |
"epoch": 0.99,
|
| 667 |
"learning_rate": 9.874326750448834e-06,
|
| 668 |
+
"loss": 0.5233,
|
| 669 |
"step": 1100
|
| 670 |
},
|
| 671 |
{
|
| 672 |
"epoch": 1.0,
|
| 673 |
"learning_rate": 9.964093357271095e-06,
|
| 674 |
+
"loss": 0.5622,
|
| 675 |
"step": 1110
|
| 676 |
},
|
| 677 |
{
|
| 678 |
"epoch": 1.0,
|
| 679 |
"eval_accuracy": {
|
| 680 |
+
"accuracy": 0.8736263736263736
|
| 681 |
},
|
| 682 |
"eval_f1": {
|
| 683 |
+
"f1": 0.8653360894930927
|
| 684 |
},
|
| 685 |
+
"eval_loss": 0.4108576774597168,
|
| 686 |
"eval_precision": {
|
| 687 |
+
"precision": 0.8669392826942903
|
| 688 |
},
|
| 689 |
"eval_recall": {
|
| 690 |
+
"recall": 0.8641606609285752
|
| 691 |
},
|
| 692 |
+
"eval_runtime": 236.9507,
|
| 693 |
+
"eval_samples_per_second": 75.273,
|
| 694 |
+
"eval_steps_per_second": 4.706,
|
| 695 |
"step": 1114
|
| 696 |
},
|
| 697 |
{
|
| 698 |
"epoch": 1.0,
|
| 699 |
"learning_rate": 1.0053859964093358e-05,
|
| 700 |
+
"loss": 0.4957,
|
| 701 |
"step": 1120
|
| 702 |
},
|
| 703 |
{
|
| 704 |
"epoch": 1.01,
|
| 705 |
"learning_rate": 1.0143626570915619e-05,
|
| 706 |
+
"loss": 0.5729,
|
| 707 |
"step": 1130
|
| 708 |
},
|
| 709 |
{
|
| 710 |
"epoch": 1.02,
|
| 711 |
"learning_rate": 1.0233393177737882e-05,
|
| 712 |
+
"loss": 0.4884,
|
| 713 |
"step": 1140
|
| 714 |
},
|
| 715 |
{
|
| 716 |
"epoch": 1.03,
|
| 717 |
"learning_rate": 1.0323159784560143e-05,
|
| 718 |
+
"loss": 0.4744,
|
| 719 |
"step": 1150
|
| 720 |
},
|
| 721 |
{
|
| 722 |
"epoch": 1.04,
|
| 723 |
"learning_rate": 1.0412926391382406e-05,
|
| 724 |
+
"loss": 0.4683,
|
| 725 |
"step": 1160
|
| 726 |
},
|
| 727 |
{
|
| 728 |
"epoch": 1.05,
|
| 729 |
"learning_rate": 1.0502692998204669e-05,
|
| 730 |
+
"loss": 0.5,
|
| 731 |
"step": 1170
|
| 732 |
},
|
| 733 |
{
|
| 734 |
"epoch": 1.06,
|
| 735 |
"learning_rate": 1.059245960502693e-05,
|
| 736 |
+
"loss": 0.4649,
|
| 737 |
"step": 1180
|
| 738 |
},
|
| 739 |
{
|
| 740 |
"epoch": 1.07,
|
| 741 |
"learning_rate": 1.0682226211849193e-05,
|
| 742 |
+
"loss": 0.508,
|
| 743 |
"step": 1190
|
| 744 |
},
|
| 745 |
{
|
| 746 |
"epoch": 1.08,
|
| 747 |
"learning_rate": 1.0771992818671454e-05,
|
| 748 |
+
"loss": 0.4898,
|
| 749 |
"step": 1200
|
| 750 |
},
|
| 751 |
{
|
| 752 |
"epoch": 1.09,
|
| 753 |
"learning_rate": 1.0861759425493717e-05,
|
| 754 |
+
"loss": 0.4722,
|
| 755 |
"step": 1210
|
| 756 |
},
|
| 757 |
{
|
| 758 |
"epoch": 1.09,
|
| 759 |
"learning_rate": 1.0951526032315979e-05,
|
| 760 |
+
"loss": 0.4814,
|
| 761 |
"step": 1220
|
| 762 |
},
|
| 763 |
{
|
| 764 |
"epoch": 1.1,
|
| 765 |
"learning_rate": 1.1041292639138241e-05,
|
| 766 |
+
"loss": 0.4505,
|
| 767 |
"step": 1230
|
| 768 |
},
|
| 769 |
{
|
| 770 |
"epoch": 1.11,
|
| 771 |
"learning_rate": 1.1131059245960503e-05,
|
| 772 |
+
"loss": 0.4505,
|
| 773 |
"step": 1240
|
| 774 |
},
|
| 775 |
{
|
| 776 |
"epoch": 1.12,
|
| 777 |
"learning_rate": 1.1220825852782766e-05,
|
| 778 |
+
"loss": 0.4187,
|
| 779 |
"step": 1250
|
| 780 |
},
|
| 781 |
{
|
| 782 |
"epoch": 1.13,
|
| 783 |
"learning_rate": 1.1310592459605028e-05,
|
| 784 |
+
"loss": 0.4603,
|
| 785 |
"step": 1260
|
| 786 |
},
|
| 787 |
{
|
| 788 |
"epoch": 1.14,
|
| 789 |
"learning_rate": 1.140035906642729e-05,
|
| 790 |
+
"loss": 0.4721,
|
| 791 |
"step": 1270
|
| 792 |
},
|
| 793 |
{
|
| 794 |
"epoch": 1.15,
|
| 795 |
"learning_rate": 1.1490125673249553e-05,
|
| 796 |
+
"loss": 0.4368,
|
| 797 |
"step": 1280
|
| 798 |
},
|
| 799 |
{
|
| 800 |
"epoch": 1.16,
|
| 801 |
"learning_rate": 1.1579892280071814e-05,
|
| 802 |
+
"loss": 0.4722,
|
| 803 |
"step": 1290
|
| 804 |
},
|
| 805 |
{
|
| 806 |
"epoch": 1.17,
|
| 807 |
"learning_rate": 1.1669658886894077e-05,
|
| 808 |
+
"loss": 0.4718,
|
| 809 |
"step": 1300
|
| 810 |
},
|
| 811 |
{
|
| 812 |
"epoch": 1.18,
|
| 813 |
"learning_rate": 1.1759425493716338e-05,
|
| 814 |
+
"loss": 0.4899,
|
| 815 |
"step": 1310
|
| 816 |
},
|
| 817 |
{
|
| 818 |
"epoch": 1.18,
|
| 819 |
"learning_rate": 1.1849192100538601e-05,
|
| 820 |
+
"loss": 0.4758,
|
| 821 |
"step": 1320
|
| 822 |
},
|
| 823 |
{
|
| 824 |
"epoch": 1.19,
|
| 825 |
"learning_rate": 1.1938958707360862e-05,
|
| 826 |
+
"loss": 0.4523,
|
| 827 |
"step": 1330
|
| 828 |
},
|
| 829 |
{
|
| 830 |
"epoch": 1.2,
|
| 831 |
"learning_rate": 1.2028725314183125e-05,
|
| 832 |
+
"loss": 0.485,
|
| 833 |
"step": 1340
|
| 834 |
},
|
| 835 |
{
|
| 836 |
"epoch": 1.21,
|
| 837 |
"learning_rate": 1.2118491921005388e-05,
|
| 838 |
+
"loss": 0.4977,
|
| 839 |
"step": 1350
|
| 840 |
},
|
| 841 |
{
|
| 842 |
"epoch": 1.22,
|
| 843 |
"learning_rate": 1.2208258527827649e-05,
|
| 844 |
+
"loss": 0.4305,
|
| 845 |
"step": 1360
|
| 846 |
},
|
| 847 |
{
|
| 848 |
"epoch": 1.23,
|
| 849 |
"learning_rate": 1.2298025134649912e-05,
|
| 850 |
+
"loss": 0.4221,
|
| 851 |
"step": 1370
|
| 852 |
},
|
| 853 |
{
|
| 854 |
"epoch": 1.24,
|
| 855 |
"learning_rate": 1.2387791741472173e-05,
|
| 856 |
+
"loss": 0.4975,
|
| 857 |
"step": 1380
|
| 858 |
},
|
| 859 |
{
|
| 860 |
"epoch": 1.25,
|
| 861 |
"learning_rate": 1.2477558348294434e-05,
|
| 862 |
+
"loss": 0.4506,
|
| 863 |
"step": 1390
|
| 864 |
},
|
| 865 |
{
|
| 866 |
"epoch": 1.26,
|
| 867 |
"learning_rate": 1.2567324955116697e-05,
|
| 868 |
+
"loss": 0.4326,
|
| 869 |
"step": 1400
|
| 870 |
},
|
| 871 |
{
|
| 872 |
"epoch": 1.26,
|
| 873 |
"learning_rate": 1.2657091561938959e-05,
|
| 874 |
+
"loss": 0.4479,
|
| 875 |
"step": 1410
|
| 876 |
},
|
| 877 |
{
|
| 878 |
"epoch": 1.27,
|
| 879 |
"learning_rate": 1.2746858168761221e-05,
|
| 880 |
+
"loss": 0.4468,
|
| 881 |
"step": 1420
|
| 882 |
},
|
| 883 |
{
|
| 884 |
"epoch": 1.28,
|
| 885 |
"learning_rate": 1.2836624775583483e-05,
|
| 886 |
+
"loss": 0.402,
|
| 887 |
"step": 1430
|
| 888 |
},
|
| 889 |
{
|
| 890 |
"epoch": 1.29,
|
| 891 |
"learning_rate": 1.2926391382405746e-05,
|
| 892 |
+
"loss": 0.4711,
|
| 893 |
"step": 1440
|
| 894 |
},
|
| 895 |
{
|
| 896 |
"epoch": 1.3,
|
| 897 |
"learning_rate": 1.3016157989228009e-05,
|
| 898 |
+
"loss": 0.4686,
|
| 899 |
"step": 1450
|
| 900 |
},
|
| 901 |
{
|
| 902 |
"epoch": 1.31,
|
| 903 |
"learning_rate": 1.310592459605027e-05,
|
| 904 |
+
"loss": 0.4639,
|
| 905 |
"step": 1460
|
| 906 |
},
|
| 907 |
{
|
| 908 |
"epoch": 1.32,
|
| 909 |
"learning_rate": 1.3195691202872531e-05,
|
| 910 |
+
"loss": 0.435,
|
| 911 |
"step": 1470
|
| 912 |
},
|
| 913 |
{
|
| 914 |
"epoch": 1.33,
|
| 915 |
"learning_rate": 1.3285457809694792e-05,
|
| 916 |
+
"loss": 0.3868,
|
| 917 |
"step": 1480
|
| 918 |
},
|
| 919 |
{
|
| 920 |
"epoch": 1.34,
|
| 921 |
"learning_rate": 1.3375224416517057e-05,
|
| 922 |
+
"loss": 0.4303,
|
| 923 |
"step": 1490
|
| 924 |
},
|
| 925 |
{
|
| 926 |
"epoch": 1.35,
|
| 927 |
"learning_rate": 1.3464991023339318e-05,
|
| 928 |
+
"loss": 0.432,
|
| 929 |
"step": 1500
|
| 930 |
},
|
| 931 |
{
|
| 932 |
"epoch": 1.35,
|
| 933 |
"learning_rate": 1.355475763016158e-05,
|
| 934 |
+
"loss": 0.4206,
|
| 935 |
"step": 1510
|
| 936 |
},
|
| 937 |
{
|
| 938 |
"epoch": 1.36,
|
| 939 |
"learning_rate": 1.3644524236983844e-05,
|
| 940 |
+
"loss": 0.4282,
|
| 941 |
"step": 1520
|
| 942 |
},
|
| 943 |
{
|
| 944 |
"epoch": 1.37,
|
| 945 |
"learning_rate": 1.3734290843806105e-05,
|
| 946 |
+
"loss": 0.4284,
|
| 947 |
"step": 1530
|
| 948 |
},
|
| 949 |
{
|
| 950 |
"epoch": 1.38,
|
| 951 |
"learning_rate": 1.3824057450628366e-05,
|
| 952 |
+
"loss": 0.4096,
|
| 953 |
"step": 1540
|
| 954 |
},
|
| 955 |
{
|
| 956 |
"epoch": 1.39,
|
| 957 |
"learning_rate": 1.3913824057450627e-05,
|
| 958 |
+
"loss": 0.4259,
|
| 959 |
"step": 1550
|
| 960 |
},
|
| 961 |
{
|
| 962 |
"epoch": 1.4,
|
| 963 |
"learning_rate": 1.4003590664272892e-05,
|
| 964 |
+
"loss": 0.4436,
|
| 965 |
"step": 1560
|
| 966 |
},
|
| 967 |
{
|
| 968 |
"epoch": 1.41,
|
| 969 |
"learning_rate": 1.4093357271095153e-05,
|
| 970 |
+
"loss": 0.4453,
|
| 971 |
"step": 1570
|
| 972 |
},
|
| 973 |
{
|
| 974 |
"epoch": 1.42,
|
| 975 |
"learning_rate": 1.4183123877917415e-05,
|
| 976 |
+
"loss": 0.4065,
|
| 977 |
"step": 1580
|
| 978 |
},
|
| 979 |
{
|
| 980 |
"epoch": 1.43,
|
| 981 |
"learning_rate": 1.4272890484739679e-05,
|
| 982 |
+
"loss": 0.4011,
|
| 983 |
"step": 1590
|
| 984 |
},
|
| 985 |
{
|
| 986 |
"epoch": 1.44,
|
| 987 |
"learning_rate": 1.436265709156194e-05,
|
| 988 |
+
"loss": 0.4488,
|
| 989 |
"step": 1600
|
| 990 |
},
|
| 991 |
{
|
| 992 |
"epoch": 1.44,
|
| 993 |
"learning_rate": 1.4452423698384202e-05,
|
| 994 |
+
"loss": 0.4122,
|
| 995 |
"step": 1610
|
| 996 |
},
|
| 997 |
{
|
| 998 |
"epoch": 1.45,
|
| 999 |
"learning_rate": 1.4542190305206463e-05,
|
| 1000 |
+
"loss": 0.429,
|
| 1001 |
"step": 1620
|
| 1002 |
},
|
| 1003 |
{
|
| 1004 |
"epoch": 1.46,
|
| 1005 |
"learning_rate": 1.4631956912028727e-05,
|
| 1006 |
+
"loss": 0.4377,
|
| 1007 |
"step": 1630
|
| 1008 |
},
|
| 1009 |
{
|
| 1010 |
"epoch": 1.47,
|
| 1011 |
"learning_rate": 1.4721723518850989e-05,
|
| 1012 |
+
"loss": 0.4186,
|
| 1013 |
"step": 1640
|
| 1014 |
},
|
| 1015 |
{
|
| 1016 |
"epoch": 1.48,
|
| 1017 |
"learning_rate": 1.481149012567325e-05,
|
| 1018 |
+
"loss": 0.4063,
|
| 1019 |
"step": 1650
|
| 1020 |
},
|
| 1021 |
{
|
| 1022 |
"epoch": 1.49,
|
| 1023 |
"learning_rate": 1.4901256732495511e-05,
|
| 1024 |
+
"loss": 0.4055,
|
| 1025 |
"step": 1660
|
| 1026 |
},
|
| 1027 |
{
|
| 1028 |
"epoch": 1.5,
|
| 1029 |
"learning_rate": 1.4991023339317776e-05,
|
| 1030 |
+
"loss": 0.4173,
|
| 1031 |
"step": 1670
|
| 1032 |
},
|
| 1033 |
{
|
| 1034 |
"epoch": 1.51,
|
| 1035 |
"learning_rate": 1.5080789946140037e-05,
|
| 1036 |
+
"loss": 0.4361,
|
| 1037 |
"step": 1680
|
| 1038 |
},
|
| 1039 |
{
|
| 1040 |
"epoch": 1.52,
|
| 1041 |
"learning_rate": 1.5170556552962298e-05,
|
| 1042 |
+
"loss": 0.4528,
|
| 1043 |
"step": 1690
|
| 1044 |
},
|
| 1045 |
{
|
| 1046 |
"epoch": 1.53,
|
| 1047 |
"learning_rate": 1.5260323159784563e-05,
|
| 1048 |
+
"loss": 0.3665,
|
| 1049 |
"step": 1700
|
| 1050 |
},
|
| 1051 |
{
|
| 1052 |
"epoch": 1.53,
|
| 1053 |
"learning_rate": 1.5350089766606824e-05,
|
| 1054 |
+
"loss": 0.3913,
|
| 1055 |
"step": 1710
|
| 1056 |
},
|
| 1057 |
{
|
| 1058 |
"epoch": 1.54,
|
| 1059 |
"learning_rate": 1.5439856373429085e-05,
|
| 1060 |
+
"loss": 0.4002,
|
| 1061 |
"step": 1720
|
| 1062 |
},
|
| 1063 |
{
|
| 1064 |
"epoch": 1.55,
|
| 1065 |
"learning_rate": 1.5529622980251346e-05,
|
| 1066 |
+
"loss": 0.4004,
|
| 1067 |
"step": 1730
|
| 1068 |
},
|
| 1069 |
{
|
| 1070 |
"epoch": 1.56,
|
| 1071 |
"learning_rate": 1.561938958707361e-05,
|
| 1072 |
+
"loss": 0.4432,
|
| 1073 |
"step": 1740
|
| 1074 |
},
|
| 1075 |
{
|
| 1076 |
"epoch": 1.57,
|
| 1077 |
"learning_rate": 1.5709156193895872e-05,
|
| 1078 |
+
"loss": 0.3743,
|
| 1079 |
"step": 1750
|
| 1080 |
},
|
| 1081 |
{
|
| 1082 |
"epoch": 1.58,
|
| 1083 |
"learning_rate": 1.5798922800718133e-05,
|
| 1084 |
+
"loss": 0.3972,
|
| 1085 |
"step": 1760
|
| 1086 |
},
|
| 1087 |
{
|
| 1088 |
"epoch": 1.59,
|
| 1089 |
"learning_rate": 1.5888689407540398e-05,
|
| 1090 |
+
"loss": 0.368,
|
| 1091 |
"step": 1770
|
| 1092 |
},
|
| 1093 |
{
|
| 1094 |
"epoch": 1.6,
|
| 1095 |
"learning_rate": 1.597845601436266e-05,
|
| 1096 |
+
"loss": 0.4525,
|
| 1097 |
"step": 1780
|
| 1098 |
},
|
| 1099 |
{
|
| 1100 |
"epoch": 1.61,
|
| 1101 |
"learning_rate": 1.606822262118492e-05,
|
| 1102 |
+
"loss": 0.3962,
|
| 1103 |
"step": 1790
|
| 1104 |
},
|
| 1105 |
{
|
| 1106 |
"epoch": 1.61,
|
| 1107 |
"learning_rate": 1.615798922800718e-05,
|
| 1108 |
+
"loss": 0.3888,
|
| 1109 |
"step": 1800
|
| 1110 |
},
|
| 1111 |
{
|
| 1112 |
"epoch": 1.62,
|
| 1113 |
"learning_rate": 1.6247755834829446e-05,
|
| 1114 |
+
"loss": 0.4111,
|
| 1115 |
"step": 1810
|
| 1116 |
},
|
| 1117 |
{
|
| 1118 |
"epoch": 1.63,
|
| 1119 |
"learning_rate": 1.6337522441651707e-05,
|
| 1120 |
+
"loss": 0.4084,
|
| 1121 |
"step": 1820
|
| 1122 |
},
|
| 1123 |
{
|
| 1124 |
"epoch": 1.64,
|
| 1125 |
"learning_rate": 1.642728904847397e-05,
|
| 1126 |
+
"loss": 0.4181,
|
| 1127 |
"step": 1830
|
| 1128 |
},
|
| 1129 |
{
|
| 1130 |
"epoch": 1.65,
|
| 1131 |
"learning_rate": 1.6517055655296233e-05,
|
| 1132 |
+
"loss": 0.4022,
|
| 1133 |
"step": 1840
|
| 1134 |
},
|
| 1135 |
{
|
| 1136 |
"epoch": 1.66,
|
| 1137 |
"learning_rate": 1.6606822262118494e-05,
|
| 1138 |
+
"loss": 0.3886,
|
| 1139 |
"step": 1850
|
| 1140 |
},
|
| 1141 |
{
|
| 1142 |
"epoch": 1.67,
|
| 1143 |
"learning_rate": 1.6696588868940756e-05,
|
| 1144 |
+
"loss": 0.4092,
|
| 1145 |
"step": 1860
|
| 1146 |
},
|
| 1147 |
{
|
| 1148 |
"epoch": 1.68,
|
| 1149 |
"learning_rate": 1.6786355475763017e-05,
|
| 1150 |
+
"loss": 0.3843,
|
| 1151 |
"step": 1870
|
| 1152 |
},
|
| 1153 |
{
|
| 1154 |
"epoch": 1.69,
|
| 1155 |
"learning_rate": 1.687612208258528e-05,
|
| 1156 |
+
"loss": 0.4237,
|
| 1157 |
"step": 1880
|
| 1158 |
},
|
| 1159 |
{
|
| 1160 |
"epoch": 1.7,
|
| 1161 |
"learning_rate": 1.6965888689407543e-05,
|
| 1162 |
+
"loss": 0.3801,
|
| 1163 |
"step": 1890
|
| 1164 |
},
|
| 1165 |
{
|
| 1166 |
"epoch": 1.7,
|
| 1167 |
"learning_rate": 1.7055655296229804e-05,
|
| 1168 |
+
"loss": 0.385,
|
| 1169 |
"step": 1900
|
| 1170 |
},
|
| 1171 |
{
|
| 1172 |
"epoch": 1.71,
|
| 1173 |
"learning_rate": 1.7145421903052065e-05,
|
| 1174 |
+
"loss": 0.4452,
|
| 1175 |
"step": 1910
|
| 1176 |
},
|
| 1177 |
{
|
| 1178 |
"epoch": 1.72,
|
| 1179 |
"learning_rate": 1.723518850987433e-05,
|
| 1180 |
+
"loss": 0.5147,
|
| 1181 |
"step": 1920
|
| 1182 |
},
|
| 1183 |
{
|
| 1184 |
"epoch": 1.73,
|
| 1185 |
"learning_rate": 1.732495511669659e-05,
|
| 1186 |
+
"loss": 0.353,
|
| 1187 |
"step": 1930
|
| 1188 |
},
|
| 1189 |
{
|
| 1190 |
"epoch": 1.74,
|
| 1191 |
"learning_rate": 1.7414721723518852e-05,
|
| 1192 |
+
"loss": 0.3816,
|
| 1193 |
"step": 1940
|
| 1194 |
},
|
| 1195 |
{
|
| 1196 |
"epoch": 1.75,
|
| 1197 |
"learning_rate": 1.7504488330341113e-05,
|
| 1198 |
+
"loss": 0.361,
|
| 1199 |
"step": 1950
|
| 1200 |
},
|
| 1201 |
{
|
| 1202 |
"epoch": 1.76,
|
| 1203 |
"learning_rate": 1.7594254937163378e-05,
|
| 1204 |
+
"loss": 0.3884,
|
| 1205 |
"step": 1960
|
| 1206 |
},
|
| 1207 |
{
|
| 1208 |
"epoch": 1.77,
|
| 1209 |
"learning_rate": 1.768402154398564e-05,
|
| 1210 |
+
"loss": 0.3391,
|
| 1211 |
"step": 1970
|
| 1212 |
},
|
| 1213 |
{
|
| 1214 |
"epoch": 1.78,
|
| 1215 |
"learning_rate": 1.77737881508079e-05,
|
| 1216 |
+
"loss": 0.4211,
|
| 1217 |
"step": 1980
|
| 1218 |
},
|
| 1219 |
{
|
| 1220 |
"epoch": 1.79,
|
| 1221 |
"learning_rate": 1.786355475763016e-05,
|
| 1222 |
+
"loss": 0.3796,
|
| 1223 |
"step": 1990
|
| 1224 |
},
|
| 1225 |
{
|
| 1226 |
"epoch": 1.79,
|
| 1227 |
"learning_rate": 1.7953321364452423e-05,
|
| 1228 |
+
"loss": 0.3881,
|
| 1229 |
"step": 2000
|
| 1230 |
},
|
| 1231 |
{
|
| 1232 |
"epoch": 1.8,
|
| 1233 |
"learning_rate": 1.8043087971274687e-05,
|
| 1234 |
+
"loss": 0.3904,
|
| 1235 |
"step": 2010
|
| 1236 |
},
|
| 1237 |
{
|
| 1238 |
"epoch": 1.81,
|
| 1239 |
"learning_rate": 1.813285457809695e-05,
|
| 1240 |
+
"loss": 0.3845,
|
| 1241 |
"step": 2020
|
| 1242 |
},
|
| 1243 |
{
|
| 1244 |
"epoch": 1.82,
|
| 1245 |
"learning_rate": 1.822262118491921e-05,
|
| 1246 |
+
"loss": 0.3721,
|
| 1247 |
"step": 2030
|
| 1248 |
},
|
| 1249 |
{
|
| 1250 |
"epoch": 1.83,
|
| 1251 |
"learning_rate": 1.831238779174147e-05,
|
| 1252 |
+
"loss": 0.4141,
|
| 1253 |
"step": 2040
|
| 1254 |
},
|
| 1255 |
{
|
| 1256 |
"epoch": 1.84,
|
| 1257 |
"learning_rate": 1.8402154398563732e-05,
|
| 1258 |
+
"loss": 0.3682,
|
| 1259 |
"step": 2050
|
| 1260 |
},
|
| 1261 |
{
|
| 1262 |
"epoch": 1.85,
|
| 1263 |
"learning_rate": 1.8491921005385997e-05,
|
| 1264 |
+
"loss": 0.3826,
|
| 1265 |
"step": 2060
|
| 1266 |
},
|
| 1267 |
{
|
| 1268 |
"epoch": 1.86,
|
| 1269 |
"learning_rate": 1.8581687612208258e-05,
|
| 1270 |
+
"loss": 0.3588,
|
| 1271 |
"step": 2070
|
| 1272 |
},
|
| 1273 |
{
|
| 1274 |
"epoch": 1.87,
|
| 1275 |
"learning_rate": 1.867145421903052e-05,
|
| 1276 |
+
"loss": 0.3663,
|
| 1277 |
"step": 2080
|
| 1278 |
},
|
| 1279 |
{
|
| 1280 |
"epoch": 1.87,
|
| 1281 |
"learning_rate": 1.8761220825852784e-05,
|
| 1282 |
+
"loss": 0.4186,
|
| 1283 |
"step": 2090
|
| 1284 |
},
|
| 1285 |
{
|
| 1286 |
"epoch": 1.88,
|
| 1287 |
"learning_rate": 1.8850987432675045e-05,
|
| 1288 |
+
"loss": 0.3756,
|
| 1289 |
"step": 2100
|
| 1290 |
},
|
| 1291 |
{
|
| 1292 |
"epoch": 1.89,
|
| 1293 |
"learning_rate": 1.8940754039497306e-05,
|
| 1294 |
+
"loss": 0.3655,
|
| 1295 |
"step": 2110
|
| 1296 |
},
|
| 1297 |
{
|
| 1298 |
"epoch": 1.9,
|
| 1299 |
"learning_rate": 1.9030520646319568e-05,
|
| 1300 |
+
"loss": 0.4158,
|
| 1301 |
"step": 2120
|
| 1302 |
},
|
| 1303 |
{
|
| 1304 |
"epoch": 1.91,
|
| 1305 |
"learning_rate": 1.9120287253141832e-05,
|
| 1306 |
+
"loss": 0.3985,
|
| 1307 |
"step": 2130
|
| 1308 |
},
|
| 1309 |
{
|
| 1310 |
"epoch": 1.92,
|
| 1311 |
"learning_rate": 1.9210053859964093e-05,
|
| 1312 |
+
"loss": 0.3955,
|
| 1313 |
"step": 2140
|
| 1314 |
},
|
| 1315 |
{
|
| 1316 |
"epoch": 1.93,
|
| 1317 |
"learning_rate": 1.9299820466786355e-05,
|
| 1318 |
+
"loss": 0.3757,
|
| 1319 |
"step": 2150
|
| 1320 |
},
|
| 1321 |
{
|
| 1322 |
"epoch": 1.94,
|
| 1323 |
"learning_rate": 1.9389587073608616e-05,
|
| 1324 |
+
"loss": 0.3712,
|
| 1325 |
"step": 2160
|
| 1326 |
},
|
| 1327 |
{
|
| 1328 |
"epoch": 1.95,
|
| 1329 |
"learning_rate": 1.947935368043088e-05,
|
| 1330 |
+
"loss": 0.3723,
|
| 1331 |
"step": 2170
|
| 1332 |
},
|
| 1333 |
{
|
| 1334 |
"epoch": 1.96,
|
| 1335 |
"learning_rate": 1.9569120287253142e-05,
|
| 1336 |
+
"loss": 0.367,
|
| 1337 |
"step": 2180
|
| 1338 |
},
|
| 1339 |
{
|
| 1340 |
"epoch": 1.96,
|
| 1341 |
"learning_rate": 1.9658886894075403e-05,
|
| 1342 |
+
"loss": 0.3659,
|
| 1343 |
"step": 2190
|
| 1344 |
},
|
| 1345 |
{
|
| 1346 |
"epoch": 1.97,
|
| 1347 |
"learning_rate": 1.9748653500897668e-05,
|
| 1348 |
+
"loss": 0.3752,
|
| 1349 |
"step": 2200
|
| 1350 |
},
|
| 1351 |
{
|
| 1352 |
"epoch": 1.98,
|
| 1353 |
"learning_rate": 1.983842010771993e-05,
|
| 1354 |
+
"loss": 0.318,
|
| 1355 |
"step": 2210
|
| 1356 |
},
|
| 1357 |
{
|
| 1358 |
"epoch": 1.99,
|
| 1359 |
"learning_rate": 1.992818671454219e-05,
|
| 1360 |
+
"loss": 0.3696,
|
| 1361 |
"step": 2220
|
| 1362 |
},
|
| 1363 |
{
|
| 1364 |
"epoch": 2.0,
|
| 1365 |
"eval_accuracy": {
|
| 1366 |
+
"accuracy": 0.9249271137026239
|
| 1367 |
},
|
| 1368 |
"eval_f1": {
|
| 1369 |
+
"f1": 0.9207602119769538
|
| 1370 |
},
|
| 1371 |
+
"eval_loss": 0.2492757886648178,
|
| 1372 |
"eval_precision": {
|
| 1373 |
+
"precision": 0.9199949527238288
|
| 1374 |
},
|
| 1375 |
"eval_recall": {
|
| 1376 |
+
"recall": 0.9219224853720074
|
| 1377 |
},
|
| 1378 |
+
"eval_runtime": 166.1808,
|
| 1379 |
+
"eval_samples_per_second": 107.329,
|
| 1380 |
+
"eval_steps_per_second": 6.71,
|
| 1381 |
"step": 2229
|
| 1382 |
},
|
| 1383 |
{
|
| 1384 |
"epoch": 2.0,
|
| 1385 |
"learning_rate": 2.001795332136445e-05,
|
| 1386 |
+
"loss": 0.3681,
|
| 1387 |
"step": 2230
|
| 1388 |
},
|
| 1389 |
{
|
| 1390 |
"epoch": 2.01,
|
| 1391 |
"learning_rate": 2.0107719928186716e-05,
|
| 1392 |
+
"loss": 0.321,
|
| 1393 |
"step": 2240
|
| 1394 |
},
|
| 1395 |
{
|
| 1396 |
"epoch": 2.02,
|
| 1397 |
"learning_rate": 2.0197486535008977e-05,
|
| 1398 |
+
"loss": 0.332,
|
| 1399 |
"step": 2250
|
| 1400 |
},
|
| 1401 |
{
|
| 1402 |
"epoch": 2.03,
|
| 1403 |
"learning_rate": 2.0287253141831238e-05,
|
| 1404 |
+
"loss": 0.3343,
|
| 1405 |
"step": 2260
|
| 1406 |
},
|
| 1407 |
{
|
| 1408 |
"epoch": 2.04,
|
| 1409 |
"learning_rate": 2.0377019748653503e-05,
|
| 1410 |
+
"loss": 0.3341,
|
| 1411 |
"step": 2270
|
| 1412 |
},
|
| 1413 |
{
|
| 1414 |
"epoch": 2.05,
|
| 1415 |
"learning_rate": 2.0466786355475764e-05,
|
| 1416 |
+
"loss": 0.3473,
|
| 1417 |
"step": 2280
|
| 1418 |
},
|
| 1419 |
{
|
| 1420 |
"epoch": 2.05,
|
| 1421 |
"learning_rate": 2.0556552962298025e-05,
|
| 1422 |
+
"loss": 0.3479,
|
| 1423 |
"step": 2290
|
| 1424 |
},
|
| 1425 |
{
|
| 1426 |
"epoch": 2.06,
|
| 1427 |
"learning_rate": 2.0646319569120286e-05,
|
| 1428 |
+
"loss": 0.3235,
|
| 1429 |
"step": 2300
|
| 1430 |
},
|
| 1431 |
{
|
| 1432 |
"epoch": 2.07,
|
| 1433 |
"learning_rate": 2.073608617594255e-05,
|
| 1434 |
+
"loss": 0.3455,
|
| 1435 |
"step": 2310
|
| 1436 |
},
|
| 1437 |
{
|
| 1438 |
"epoch": 2.08,
|
| 1439 |
"learning_rate": 2.0825852782764812e-05,
|
| 1440 |
+
"loss": 0.3675,
|
| 1441 |
"step": 2320
|
| 1442 |
},
|
| 1443 |
{
|
| 1444 |
"epoch": 2.09,
|
| 1445 |
"learning_rate": 2.0915619389587073e-05,
|
| 1446 |
+
"loss": 0.3267,
|
| 1447 |
"step": 2330
|
| 1448 |
},
|
| 1449 |
{
|
| 1450 |
"epoch": 2.1,
|
| 1451 |
"learning_rate": 2.1005385996409338e-05,
|
| 1452 |
+
"loss": 0.3588,
|
| 1453 |
"step": 2340
|
| 1454 |
},
|
| 1455 |
{
|
| 1456 |
"epoch": 2.11,
|
| 1457 |
"learning_rate": 2.10951526032316e-05,
|
| 1458 |
+
"loss": 0.3762,
|
| 1459 |
"step": 2350
|
| 1460 |
},
|
| 1461 |
{
|
| 1462 |
"epoch": 2.12,
|
| 1463 |
"learning_rate": 2.118491921005386e-05,
|
| 1464 |
+
"loss": 0.3474,
|
| 1465 |
"step": 2360
|
| 1466 |
},
|
| 1467 |
{
|
| 1468 |
"epoch": 2.13,
|
| 1469 |
"learning_rate": 2.1274685816876122e-05,
|
| 1470 |
+
"loss": 0.3614,
|
| 1471 |
"step": 2370
|
| 1472 |
},
|
| 1473 |
{
|
| 1474 |
"epoch": 2.14,
|
| 1475 |
"learning_rate": 2.1364452423698386e-05,
|
| 1476 |
+
"loss": 0.3611,
|
| 1477 |
"step": 2380
|
| 1478 |
},
|
| 1479 |
{
|
| 1480 |
"epoch": 2.14,
|
| 1481 |
"learning_rate": 2.1454219030520648e-05,
|
| 1482 |
+
"loss": 0.4051,
|
| 1483 |
"step": 2390
|
| 1484 |
},
|
| 1485 |
{
|
| 1486 |
"epoch": 2.15,
|
| 1487 |
"learning_rate": 2.154398563734291e-05,
|
| 1488 |
+
"loss": 0.3314,
|
| 1489 |
"step": 2400
|
| 1490 |
},
|
| 1491 |
{
|
| 1492 |
"epoch": 2.16,
|
| 1493 |
"learning_rate": 2.163375224416517e-05,
|
| 1494 |
+
"loss": 0.3427,
|
| 1495 |
"step": 2410
|
| 1496 |
},
|
| 1497 |
{
|
| 1498 |
"epoch": 2.17,
|
| 1499 |
"learning_rate": 2.1723518850987435e-05,
|
| 1500 |
+
"loss": 0.2886,
|
| 1501 |
"step": 2420
|
| 1502 |
},
|
| 1503 |
{
|
| 1504 |
"epoch": 2.18,
|
| 1505 |
"learning_rate": 2.1813285457809696e-05,
|
| 1506 |
+
"loss": 0.3262,
|
| 1507 |
"step": 2430
|
| 1508 |
},
|
| 1509 |
{
|
| 1510 |
"epoch": 2.19,
|
| 1511 |
"learning_rate": 2.1903052064631957e-05,
|
| 1512 |
+
"loss": 0.3516,
|
| 1513 |
"step": 2440
|
| 1514 |
},
|
| 1515 |
{
|
| 1516 |
"epoch": 2.2,
|
| 1517 |
"learning_rate": 2.199281867145422e-05,
|
| 1518 |
+
"loss": 0.3744,
|
| 1519 |
"step": 2450
|
| 1520 |
},
|
| 1521 |
{
|
| 1522 |
"epoch": 2.21,
|
| 1523 |
"learning_rate": 2.2082585278276483e-05,
|
| 1524 |
+
"loss": 0.3545,
|
| 1525 |
"step": 2460
|
| 1526 |
},
|
| 1527 |
{
|
| 1528 |
"epoch": 2.22,
|
| 1529 |
"learning_rate": 2.2172351885098744e-05,
|
| 1530 |
+
"loss": 0.2993,
|
| 1531 |
"step": 2470
|
| 1532 |
},
|
| 1533 |
{
|
| 1534 |
"epoch": 2.22,
|
| 1535 |
"learning_rate": 2.2262118491921005e-05,
|
| 1536 |
+
"loss": 0.3412,
|
| 1537 |
"step": 2480
|
| 1538 |
},
|
| 1539 |
{
|
| 1540 |
"epoch": 2.23,
|
| 1541 |
"learning_rate": 2.235188509874327e-05,
|
| 1542 |
+
"loss": 0.3823,
|
| 1543 |
"step": 2490
|
| 1544 |
},
|
| 1545 |
{
|
| 1546 |
"epoch": 2.24,
|
| 1547 |
"learning_rate": 2.244165170556553e-05,
|
| 1548 |
+
"loss": 0.3206,
|
| 1549 |
"step": 2500
|
| 1550 |
},
|
| 1551 |
{
|
| 1552 |
"epoch": 2.25,
|
| 1553 |
"learning_rate": 2.2531418312387792e-05,
|
| 1554 |
+
"loss": 0.2969,
|
| 1555 |
"step": 2510
|
| 1556 |
},
|
| 1557 |
{
|
| 1558 |
"epoch": 2.26,
|
| 1559 |
"learning_rate": 2.2621184919210057e-05,
|
| 1560 |
+
"loss": 0.3306,
|
| 1561 |
"step": 2520
|
| 1562 |
},
|
| 1563 |
{
|
| 1564 |
"epoch": 2.27,
|
| 1565 |
"learning_rate": 2.2710951526032318e-05,
|
| 1566 |
+
"loss": 0.3641,
|
| 1567 |
"step": 2530
|
| 1568 |
},
|
| 1569 |
{
|
| 1570 |
"epoch": 2.28,
|
| 1571 |
"learning_rate": 2.280071813285458e-05,
|
| 1572 |
+
"loss": 0.3588,
|
| 1573 |
"step": 2540
|
| 1574 |
},
|
| 1575 |
{
|
| 1576 |
"epoch": 2.29,
|
| 1577 |
"learning_rate": 2.289048473967684e-05,
|
| 1578 |
+
"loss": 0.3632,
|
| 1579 |
"step": 2550
|
| 1580 |
},
|
| 1581 |
{
|
| 1582 |
"epoch": 2.3,
|
| 1583 |
"learning_rate": 2.2980251346499105e-05,
|
| 1584 |
+
"loss": 0.3072,
|
| 1585 |
"step": 2560
|
| 1586 |
},
|
| 1587 |
{
|
| 1588 |
"epoch": 2.31,
|
| 1589 |
"learning_rate": 2.3070017953321366e-05,
|
| 1590 |
+
"loss": 0.3606,
|
| 1591 |
"step": 2570
|
| 1592 |
},
|
| 1593 |
{
|
| 1594 |
"epoch": 2.31,
|
| 1595 |
"learning_rate": 2.3159784560143628e-05,
|
| 1596 |
+
"loss": 0.4114,
|
| 1597 |
"step": 2580
|
| 1598 |
},
|
| 1599 |
{
|
| 1600 |
"epoch": 2.32,
|
| 1601 |
"learning_rate": 2.3249551166965892e-05,
|
| 1602 |
+
"loss": 0.282,
|
| 1603 |
"step": 2590
|
| 1604 |
},
|
| 1605 |
{
|
| 1606 |
"epoch": 2.33,
|
| 1607 |
"learning_rate": 2.3339317773788153e-05,
|
| 1608 |
+
"loss": 0.3336,
|
| 1609 |
"step": 2600
|
| 1610 |
},
|
| 1611 |
{
|
| 1612 |
"epoch": 2.34,
|
| 1613 |
"learning_rate": 2.3429084380610415e-05,
|
| 1614 |
+
"loss": 0.3406,
|
| 1615 |
"step": 2610
|
| 1616 |
},
|
| 1617 |
{
|
| 1618 |
"epoch": 2.35,
|
| 1619 |
"learning_rate": 2.3518850987432676e-05,
|
| 1620 |
+
"loss": 0.3433,
|
| 1621 |
"step": 2620
|
| 1622 |
},
|
| 1623 |
{
|
| 1624 |
"epoch": 2.36,
|
| 1625 |
"learning_rate": 2.360861759425494e-05,
|
| 1626 |
+
"loss": 0.3169,
|
| 1627 |
"step": 2630
|
| 1628 |
},
|
| 1629 |
{
|
| 1630 |
"epoch": 2.37,
|
| 1631 |
"learning_rate": 2.3698384201077202e-05,
|
| 1632 |
+
"loss": 0.3855,
|
| 1633 |
"step": 2640
|
| 1634 |
},
|
| 1635 |
{
|
| 1636 |
"epoch": 2.38,
|
| 1637 |
"learning_rate": 2.3788150807899463e-05,
|
| 1638 |
+
"loss": 0.3036,
|
| 1639 |
"step": 2650
|
| 1640 |
},
|
| 1641 |
{
|
| 1642 |
"epoch": 2.39,
|
| 1643 |
"learning_rate": 2.3877917414721724e-05,
|
| 1644 |
+
"loss": 0.3076,
|
| 1645 |
"step": 2660
|
| 1646 |
},
|
| 1647 |
{
|
| 1648 |
"epoch": 2.4,
|
| 1649 |
"learning_rate": 2.396768402154399e-05,
|
| 1650 |
+
"loss": 0.3052,
|
| 1651 |
"step": 2670
|
| 1652 |
},
|
| 1653 |
{
|
| 1654 |
"epoch": 2.4,
|
| 1655 |
"learning_rate": 2.405745062836625e-05,
|
| 1656 |
+
"loss": 0.3253,
|
| 1657 |
"step": 2680
|
| 1658 |
},
|
| 1659 |
{
|
| 1660 |
"epoch": 2.41,
|
| 1661 |
"learning_rate": 2.414721723518851e-05,
|
| 1662 |
+
"loss": 0.3235,
|
| 1663 |
"step": 2690
|
| 1664 |
},
|
| 1665 |
{
|
| 1666 |
"epoch": 2.42,
|
| 1667 |
"learning_rate": 2.4236983842010776e-05,
|
| 1668 |
+
"loss": 0.2835,
|
| 1669 |
"step": 2700
|
| 1670 |
},
|
| 1671 |
{
|
| 1672 |
"epoch": 2.43,
|
| 1673 |
"learning_rate": 2.4326750448833037e-05,
|
| 1674 |
+
"loss": 0.2824,
|
| 1675 |
"step": 2710
|
| 1676 |
},
|
| 1677 |
{
|
| 1678 |
"epoch": 2.44,
|
| 1679 |
"learning_rate": 2.4416517055655298e-05,
|
| 1680 |
+
"loss": 0.3183,
|
| 1681 |
"step": 2720
|
| 1682 |
},
|
| 1683 |
{
|
| 1684 |
"epoch": 2.45,
|
| 1685 |
"learning_rate": 2.450628366247756e-05,
|
| 1686 |
+
"loss": 0.3281,
|
| 1687 |
"step": 2730
|
| 1688 |
},
|
| 1689 |
{
|
| 1690 |
"epoch": 2.46,
|
| 1691 |
"learning_rate": 2.4596050269299824e-05,
|
| 1692 |
+
"loss": 0.2939,
|
| 1693 |
"step": 2740
|
| 1694 |
},
|
| 1695 |
{
|
| 1696 |
"epoch": 2.47,
|
| 1697 |
"learning_rate": 2.4685816876122085e-05,
|
| 1698 |
+
"loss": 0.3679,
|
| 1699 |
"step": 2750
|
| 1700 |
},
|
| 1701 |
{
|
| 1702 |
"epoch": 2.48,
|
| 1703 |
"learning_rate": 2.4775583482944346e-05,
|
| 1704 |
+
"loss": 0.3165,
|
| 1705 |
"step": 2760
|
| 1706 |
},
|
| 1707 |
{
|
| 1708 |
"epoch": 2.48,
|
| 1709 |
"learning_rate": 2.4865350089766608e-05,
|
| 1710 |
+
"loss": 0.3146,
|
| 1711 |
"step": 2770
|
| 1712 |
},
|
| 1713 |
{
|
| 1714 |
"epoch": 2.49,
|
| 1715 |
"learning_rate": 2.495511669658887e-05,
|
| 1716 |
+
"loss": 0.2981,
|
| 1717 |
"step": 2780
|
| 1718 |
},
|
| 1719 |
{
|
| 1720 |
"epoch": 2.5,
|
| 1721 |
"learning_rate": 2.5044883303411134e-05,
|
| 1722 |
+
"loss": 0.3099,
|
| 1723 |
"step": 2790
|
| 1724 |
},
|
| 1725 |
{
|
| 1726 |
"epoch": 2.51,
|
| 1727 |
"learning_rate": 2.5134649910233395e-05,
|
| 1728 |
+
"loss": 0.267,
|
| 1729 |
"step": 2800
|
| 1730 |
},
|
| 1731 |
{
|
| 1732 |
"epoch": 2.52,
|
| 1733 |
"learning_rate": 2.5224416517055656e-05,
|
| 1734 |
+
"loss": 0.3317,
|
| 1735 |
"step": 2810
|
| 1736 |
},
|
| 1737 |
{
|
| 1738 |
"epoch": 2.53,
|
| 1739 |
"learning_rate": 2.5314183123877917e-05,
|
| 1740 |
+
"loss": 0.3241,
|
| 1741 |
"step": 2820
|
| 1742 |
},
|
| 1743 |
{
|
| 1744 |
"epoch": 2.54,
|
| 1745 |
"learning_rate": 2.5403949730700182e-05,
|
| 1746 |
+
"loss": 0.371,
|
| 1747 |
"step": 2830
|
| 1748 |
},
|
| 1749 |
{
|
| 1750 |
"epoch": 2.55,
|
| 1751 |
"learning_rate": 2.5493716337522443e-05,
|
| 1752 |
+
"loss": 0.3431,
|
| 1753 |
"step": 2840
|
| 1754 |
},
|
| 1755 |
{
|
| 1756 |
"epoch": 2.56,
|
| 1757 |
"learning_rate": 2.5583482944344704e-05,
|
| 1758 |
+
"loss": 0.3218,
|
| 1759 |
"step": 2850
|
| 1760 |
},
|
| 1761 |
{
|
| 1762 |
"epoch": 2.57,
|
| 1763 |
"learning_rate": 2.5673249551166965e-05,
|
| 1764 |
+
"loss": 0.3222,
|
| 1765 |
"step": 2860
|
| 1766 |
},
|
| 1767 |
{
|
| 1768 |
"epoch": 2.57,
|
| 1769 |
"learning_rate": 2.5763016157989227e-05,
|
| 1770 |
+
"loss": 0.3176,
|
| 1771 |
"step": 2870
|
| 1772 |
},
|
| 1773 |
{
|
| 1774 |
"epoch": 2.58,
|
| 1775 |
"learning_rate": 2.585278276481149e-05,
|
| 1776 |
+
"loss": 0.3256,
|
| 1777 |
"step": 2880
|
| 1778 |
},
|
| 1779 |
{
|
| 1780 |
"epoch": 2.59,
|
| 1781 |
"learning_rate": 2.5942549371633752e-05,
|
| 1782 |
+
"loss": 0.2971,
|
| 1783 |
"step": 2890
|
| 1784 |
},
|
| 1785 |
{
|
| 1786 |
"epoch": 2.6,
|
| 1787 |
"learning_rate": 2.6032315978456017e-05,
|
| 1788 |
+
"loss": 0.253,
|
| 1789 |
"step": 2900
|
| 1790 |
},
|
| 1791 |
{
|
| 1792 |
"epoch": 2.61,
|
| 1793 |
"learning_rate": 2.6122082585278278e-05,
|
| 1794 |
+
"loss": 0.3261,
|
| 1795 |
"step": 2910
|
| 1796 |
},
|
| 1797 |
{
|
| 1798 |
"epoch": 2.62,
|
| 1799 |
"learning_rate": 2.621184919210054e-05,
|
| 1800 |
+
"loss": 0.3325,
|
| 1801 |
"step": 2920
|
| 1802 |
},
|
| 1803 |
{
|
| 1804 |
"epoch": 2.63,
|
| 1805 |
"learning_rate": 2.63016157989228e-05,
|
| 1806 |
+
"loss": 0.3005,
|
| 1807 |
"step": 2930
|
| 1808 |
},
|
| 1809 |
{
|
| 1810 |
"epoch": 2.64,
|
| 1811 |
"learning_rate": 2.6391382405745062e-05,
|
| 1812 |
+
"loss": 0.2897,
|
| 1813 |
"step": 2940
|
| 1814 |
},
|
| 1815 |
{
|
| 1816 |
"epoch": 2.65,
|
| 1817 |
"learning_rate": 2.6481149012567323e-05,
|
| 1818 |
+
"loss": 0.3249,
|
| 1819 |
"step": 2950
|
| 1820 |
},
|
| 1821 |
{
|
| 1822 |
"epoch": 2.66,
|
| 1823 |
"learning_rate": 2.6570915619389584e-05,
|
| 1824 |
+
"loss": 0.3647,
|
| 1825 |
"step": 2960
|
| 1826 |
},
|
| 1827 |
{
|
| 1828 |
"epoch": 2.66,
|
| 1829 |
"learning_rate": 2.6660682226211852e-05,
|
| 1830 |
+
"loss": 0.3034,
|
| 1831 |
"step": 2970
|
| 1832 |
},
|
| 1833 |
{
|
| 1834 |
"epoch": 2.67,
|
| 1835 |
"learning_rate": 2.6750448833034114e-05,
|
| 1836 |
+
"loss": 0.2735,
|
| 1837 |
"step": 2980
|
| 1838 |
},
|
| 1839 |
{
|
| 1840 |
"epoch": 2.68,
|
| 1841 |
"learning_rate": 2.6840215439856375e-05,
|
| 1842 |
+
"loss": 0.3122,
|
| 1843 |
"step": 2990
|
| 1844 |
},
|
| 1845 |
{
|
| 1846 |
"epoch": 2.69,
|
| 1847 |
"learning_rate": 2.6929982046678636e-05,
|
| 1848 |
+
"loss": 0.2622,
|
| 1849 |
"step": 3000
|
| 1850 |
},
|
| 1851 |
{
|
| 1852 |
"epoch": 2.7,
|
| 1853 |
"learning_rate": 2.7019748653500897e-05,
|
| 1854 |
+
"loss": 0.322,
|
| 1855 |
"step": 3010
|
| 1856 |
},
|
| 1857 |
{
|
| 1858 |
"epoch": 2.71,
|
| 1859 |
"learning_rate": 2.710951526032316e-05,
|
| 1860 |
+
"loss": 0.2671,
|
| 1861 |
"step": 3020
|
| 1862 |
},
|
| 1863 |
{
|
| 1864 |
"epoch": 2.72,
|
| 1865 |
"learning_rate": 2.719928186714542e-05,
|
| 1866 |
+
"loss": 0.3392,
|
| 1867 |
"step": 3030
|
| 1868 |
},
|
| 1869 |
{
|
| 1870 |
"epoch": 2.73,
|
| 1871 |
"learning_rate": 2.7289048473967688e-05,
|
| 1872 |
+
"loss": 0.2507,
|
| 1873 |
"step": 3040
|
| 1874 |
},
|
| 1875 |
{
|
| 1876 |
"epoch": 2.74,
|
| 1877 |
"learning_rate": 2.737881508078995e-05,
|
| 1878 |
+
"loss": 0.294,
|
| 1879 |
"step": 3050
|
| 1880 |
},
|
| 1881 |
{
|
| 1882 |
"epoch": 2.75,
|
| 1883 |
"learning_rate": 2.746858168761221e-05,
|
| 1884 |
+
"loss": 0.3119,
|
| 1885 |
"step": 3060
|
| 1886 |
},
|
| 1887 |
{
|
| 1888 |
"epoch": 2.75,
|
| 1889 |
"learning_rate": 2.755834829443447e-05,
|
| 1890 |
+
"loss": 0.2956,
|
| 1891 |
"step": 3070
|
| 1892 |
},
|
| 1893 |
{
|
| 1894 |
"epoch": 2.76,
|
| 1895 |
"learning_rate": 2.7648114901256732e-05,
|
| 1896 |
+
"loss": 0.2744,
|
| 1897 |
"step": 3080
|
| 1898 |
},
|
| 1899 |
{
|
| 1900 |
"epoch": 2.77,
|
| 1901 |
"learning_rate": 2.7737881508078994e-05,
|
| 1902 |
+
"loss": 0.279,
|
| 1903 |
"step": 3090
|
| 1904 |
},
|
| 1905 |
{
|
| 1906 |
"epoch": 2.78,
|
| 1907 |
"learning_rate": 2.7827648114901255e-05,
|
| 1908 |
+
"loss": 0.379,
|
| 1909 |
"step": 3100
|
| 1910 |
},
|
| 1911 |
{
|
| 1912 |
"epoch": 2.79,
|
| 1913 |
"learning_rate": 2.7917414721723523e-05,
|
| 1914 |
+
"loss": 0.2667,
|
| 1915 |
"step": 3110
|
| 1916 |
},
|
| 1917 |
{
|
| 1918 |
"epoch": 2.8,
|
| 1919 |
"learning_rate": 2.8007181328545784e-05,
|
| 1920 |
+
"loss": 0.2732,
|
| 1921 |
"step": 3120
|
| 1922 |
},
|
| 1923 |
{
|
| 1924 |
"epoch": 2.81,
|
| 1925 |
"learning_rate": 2.8096947935368045e-05,
|
| 1926 |
+
"loss": 0.311,
|
| 1927 |
"step": 3130
|
| 1928 |
},
|
| 1929 |
{
|
| 1930 |
"epoch": 2.82,
|
| 1931 |
"learning_rate": 2.8186714542190307e-05,
|
| 1932 |
+
"loss": 0.3404,
|
| 1933 |
"step": 3140
|
| 1934 |
},
|
| 1935 |
{
|
| 1936 |
"epoch": 2.83,
|
| 1937 |
"learning_rate": 2.8276481149012568e-05,
|
| 1938 |
+
"loss": 0.2479,
|
| 1939 |
"step": 3150
|
| 1940 |
},
|
| 1941 |
{
|
| 1942 |
"epoch": 2.83,
|
| 1943 |
"learning_rate": 2.836624775583483e-05,
|
| 1944 |
+
"loss": 0.3057,
|
| 1945 |
"step": 3160
|
| 1946 |
},
|
| 1947 |
{
|
| 1948 |
"epoch": 2.84,
|
| 1949 |
"learning_rate": 2.845601436265709e-05,
|
| 1950 |
+
"loss": 0.3075,
|
| 1951 |
"step": 3170
|
| 1952 |
},
|
| 1953 |
{
|
| 1954 |
"epoch": 2.85,
|
| 1955 |
"learning_rate": 2.8545780969479358e-05,
|
| 1956 |
+
"loss": 0.3008,
|
| 1957 |
"step": 3180
|
| 1958 |
},
|
| 1959 |
{
|
| 1960 |
"epoch": 2.86,
|
| 1961 |
"learning_rate": 2.863554757630162e-05,
|
| 1962 |
+
"loss": 0.2372,
|
| 1963 |
"step": 3190
|
| 1964 |
},
|
| 1965 |
{
|
| 1966 |
"epoch": 2.87,
|
| 1967 |
"learning_rate": 2.872531418312388e-05,
|
| 1968 |
+
"loss": 0.2617,
|
| 1969 |
"step": 3200
|
| 1970 |
},
|
| 1971 |
{
|
| 1972 |
"epoch": 2.88,
|
| 1973 |
"learning_rate": 2.8815080789946142e-05,
|
| 1974 |
+
"loss": 0.3,
|
| 1975 |
"step": 3210
|
| 1976 |
},
|
| 1977 |
{
|
| 1978 |
"epoch": 2.89,
|
| 1979 |
"learning_rate": 2.8904847396768403e-05,
|
| 1980 |
+
"loss": 0.2749,
|
| 1981 |
"step": 3220
|
| 1982 |
},
|
| 1983 |
{
|
| 1984 |
"epoch": 2.9,
|
| 1985 |
"learning_rate": 2.8994614003590664e-05,
|
| 1986 |
+
"loss": 0.2742,
|
| 1987 |
"step": 3230
|
| 1988 |
},
|
| 1989 |
{
|
| 1990 |
"epoch": 2.91,
|
| 1991 |
"learning_rate": 2.9084380610412926e-05,
|
| 1992 |
+
"loss": 0.2844,
|
| 1993 |
"step": 3240
|
| 1994 |
},
|
| 1995 |
{
|
| 1996 |
"epoch": 2.92,
|
| 1997 |
"learning_rate": 2.9174147217235194e-05,
|
| 1998 |
+
"loss": 0.2909,
|
| 1999 |
"step": 3250
|
| 2000 |
},
|
| 2001 |
{
|
| 2002 |
"epoch": 2.92,
|
| 2003 |
"learning_rate": 2.9263913824057455e-05,
|
| 2004 |
+
"loss": 0.3045,
|
| 2005 |
"step": 3260
|
| 2006 |
},
|
| 2007 |
{
|
| 2008 |
"epoch": 2.93,
|
| 2009 |
"learning_rate": 2.9353680430879716e-05,
|
| 2010 |
+
"loss": 0.2875,
|
| 2011 |
"step": 3270
|
| 2012 |
},
|
| 2013 |
{
|
| 2014 |
"epoch": 2.94,
|
| 2015 |
"learning_rate": 2.9443447037701977e-05,
|
| 2016 |
+
"loss": 0.2963,
|
| 2017 |
"step": 3280
|
| 2018 |
},
|
| 2019 |
{
|
| 2020 |
"epoch": 2.95,
|
| 2021 |
"learning_rate": 2.953321364452424e-05,
|
| 2022 |
+
"loss": 0.314,
|
| 2023 |
"step": 3290
|
| 2024 |
},
|
| 2025 |
{
|
| 2026 |
"epoch": 2.96,
|
| 2027 |
"learning_rate": 2.96229802513465e-05,
|
| 2028 |
+
"loss": 0.3005,
|
| 2029 |
"step": 3300
|
| 2030 |
},
|
| 2031 |
{
|
| 2032 |
"epoch": 2.97,
|
| 2033 |
"learning_rate": 2.971274685816876e-05,
|
| 2034 |
+
"loss": 0.3127,
|
| 2035 |
"step": 3310
|
| 2036 |
},
|
| 2037 |
{
|
| 2038 |
"epoch": 2.98,
|
| 2039 |
"learning_rate": 2.9802513464991022e-05,
|
| 2040 |
+
"loss": 0.3113,
|
| 2041 |
"step": 3320
|
| 2042 |
},
|
| 2043 |
{
|
| 2044 |
"epoch": 2.99,
|
| 2045 |
"learning_rate": 2.989228007181329e-05,
|
| 2046 |
+
"loss": 0.2886,
|
| 2047 |
"step": 3330
|
| 2048 |
},
|
| 2049 |
{
|
| 2050 |
"epoch": 3.0,
|
| 2051 |
"learning_rate": 2.998204667863555e-05,
|
| 2052 |
+
"loss": 0.321,
|
| 2053 |
"step": 3340
|
| 2054 |
},
|
| 2055 |
{
|
| 2056 |
"epoch": 3.0,
|
| 2057 |
"eval_accuracy": {
|
| 2058 |
+
"accuracy": 0.9410742318905584
|
| 2059 |
},
|
| 2060 |
"eval_f1": {
|
| 2061 |
+
"f1": 0.9378351696354423
|
| 2062 |
},
|
| 2063 |
+
"eval_loss": 0.18490658700466156,
|
| 2064 |
"eval_precision": {
|
| 2065 |
+
"precision": 0.9376663800198108
|
| 2066 |
},
|
| 2067 |
"eval_recall": {
|
| 2068 |
+
"recall": 0.9381803360170522
|
| 2069 |
},
|
| 2070 |
+
"eval_runtime": 165.2584,
|
| 2071 |
+
"eval_samples_per_second": 107.928,
|
| 2072 |
+
"eval_steps_per_second": 6.747,
|
| 2073 |
"step": 3344
|
| 2074 |
},
|
| 2075 |
{
|
| 2076 |
"epoch": 3.01,
|
| 2077 |
"learning_rate": 3.0071813285457812e-05,
|
| 2078 |
+
"loss": 0.3252,
|
| 2079 |
"step": 3350
|
| 2080 |
},
|
| 2081 |
{
|
| 2082 |
"epoch": 3.01,
|
| 2083 |
"learning_rate": 3.0161579892280074e-05,
|
| 2084 |
+
"loss": 0.2945,
|
| 2085 |
"step": 3360
|
| 2086 |
},
|
| 2087 |
{
|
| 2088 |
"epoch": 3.02,
|
| 2089 |
"learning_rate": 3.0251346499102335e-05,
|
| 2090 |
+
"loss": 0.2584,
|
| 2091 |
"step": 3370
|
| 2092 |
},
|
| 2093 |
{
|
| 2094 |
"epoch": 3.03,
|
| 2095 |
"learning_rate": 3.0341113105924596e-05,
|
| 2096 |
+
"loss": 0.2878,
|
| 2097 |
"step": 3380
|
| 2098 |
},
|
| 2099 |
{
|
| 2100 |
"epoch": 3.04,
|
| 2101 |
"learning_rate": 3.0430879712746857e-05,
|
| 2102 |
+
"loss": 0.2618,
|
| 2103 |
"step": 3390
|
| 2104 |
},
|
| 2105 |
{
|
| 2106 |
"epoch": 3.05,
|
| 2107 |
"learning_rate": 3.0520646319569125e-05,
|
| 2108 |
+
"loss": 0.3356,
|
| 2109 |
"step": 3400
|
| 2110 |
},
|
| 2111 |
{
|
| 2112 |
"epoch": 3.06,
|
| 2113 |
"learning_rate": 3.061041292639138e-05,
|
| 2114 |
+
"loss": 0.2706,
|
| 2115 |
"step": 3410
|
| 2116 |
},
|
| 2117 |
{
|
| 2118 |
"epoch": 3.07,
|
| 2119 |
"learning_rate": 3.070017953321365e-05,
|
| 2120 |
+
"loss": 0.2821,
|
| 2121 |
"step": 3420
|
| 2122 |
},
|
| 2123 |
{
|
| 2124 |
"epoch": 3.08,
|
| 2125 |
"learning_rate": 3.0789946140035906e-05,
|
| 2126 |
+
"loss": 0.2634,
|
| 2127 |
"step": 3430
|
| 2128 |
},
|
| 2129 |
{
|
| 2130 |
"epoch": 3.09,
|
| 2131 |
"learning_rate": 3.087971274685817e-05,
|
| 2132 |
+
"loss": 0.2927,
|
| 2133 |
"step": 3440
|
| 2134 |
},
|
| 2135 |
{
|
| 2136 |
"epoch": 3.09,
|
| 2137 |
"learning_rate": 3.096947935368043e-05,
|
| 2138 |
+
"loss": 0.2406,
|
| 2139 |
"step": 3450
|
| 2140 |
},
|
| 2141 |
{
|
| 2142 |
"epoch": 3.1,
|
| 2143 |
"learning_rate": 3.105924596050269e-05,
|
| 2144 |
+
"loss": 0.2871,
|
| 2145 |
"step": 3460
|
| 2146 |
},
|
| 2147 |
{
|
| 2148 |
"epoch": 3.11,
|
| 2149 |
"learning_rate": 3.114901256732496e-05,
|
| 2150 |
+
"loss": 0.2389,
|
| 2151 |
"step": 3470
|
| 2152 |
},
|
| 2153 |
{
|
| 2154 |
"epoch": 3.12,
|
| 2155 |
"learning_rate": 3.123877917414722e-05,
|
| 2156 |
+
"loss": 0.2782,
|
| 2157 |
"step": 3480
|
| 2158 |
},
|
| 2159 |
{
|
| 2160 |
"epoch": 3.13,
|
| 2161 |
"learning_rate": 3.132854578096948e-05,
|
| 2162 |
+
"loss": 0.2864,
|
| 2163 |
"step": 3490
|
| 2164 |
},
|
| 2165 |
{
|
| 2166 |
"epoch": 3.14,
|
| 2167 |
"learning_rate": 3.1418312387791744e-05,
|
| 2168 |
+
"loss": 0.29,
|
| 2169 |
"step": 3500
|
| 2170 |
},
|
| 2171 |
{
|
| 2172 |
"epoch": 3.15,
|
| 2173 |
"learning_rate": 3.1508078994614e-05,
|
| 2174 |
+
"loss": 0.2089,
|
| 2175 |
"step": 3510
|
| 2176 |
},
|
| 2177 |
{
|
| 2178 |
"epoch": 3.16,
|
| 2179 |
"learning_rate": 3.159784560143627e-05,
|
| 2180 |
+
"loss": 0.2716,
|
| 2181 |
"step": 3520
|
| 2182 |
},
|
| 2183 |
{
|
| 2184 |
"epoch": 3.17,
|
| 2185 |
"learning_rate": 3.1687612208258525e-05,
|
| 2186 |
+
"loss": 0.2981,
|
| 2187 |
"step": 3530
|
| 2188 |
},
|
| 2189 |
{
|
| 2190 |
"epoch": 3.18,
|
| 2191 |
"learning_rate": 3.1777378815080796e-05,
|
| 2192 |
+
"loss": 0.2265,
|
| 2193 |
"step": 3540
|
| 2194 |
},
|
| 2195 |
{
|
| 2196 |
"epoch": 3.18,
|
| 2197 |
"learning_rate": 3.1867145421903054e-05,
|
| 2198 |
+
"loss": 0.1997,
|
| 2199 |
"step": 3550
|
| 2200 |
},
|
| 2201 |
{
|
| 2202 |
"epoch": 3.19,
|
| 2203 |
"learning_rate": 3.195691202872532e-05,
|
| 2204 |
+
"loss": 0.2707,
|
| 2205 |
"step": 3560
|
| 2206 |
},
|
| 2207 |
{
|
| 2208 |
"epoch": 3.2,
|
| 2209 |
"learning_rate": 3.2046678635547576e-05,
|
| 2210 |
+
"loss": 0.2917,
|
| 2211 |
"step": 3570
|
| 2212 |
},
|
| 2213 |
{
|
| 2214 |
"epoch": 3.21,
|
| 2215 |
"learning_rate": 3.213644524236984e-05,
|
| 2216 |
+
"loss": 0.2966,
|
| 2217 |
"step": 3580
|
| 2218 |
},
|
| 2219 |
{
|
| 2220 |
"epoch": 3.22,
|
| 2221 |
"learning_rate": 3.22262118491921e-05,
|
| 2222 |
+
"loss": 0.265,
|
| 2223 |
"step": 3590
|
| 2224 |
},
|
| 2225 |
{
|
| 2226 |
"epoch": 3.23,
|
| 2227 |
"learning_rate": 3.231597845601436e-05,
|
| 2228 |
+
"loss": 0.2386,
|
| 2229 |
"step": 3600
|
| 2230 |
},
|
| 2231 |
{
|
| 2232 |
"epoch": 3.24,
|
| 2233 |
"learning_rate": 3.240574506283663e-05,
|
| 2234 |
+
"loss": 0.2613,
|
| 2235 |
"step": 3610
|
| 2236 |
},
|
| 2237 |
{
|
| 2238 |
"epoch": 3.25,
|
| 2239 |
"learning_rate": 3.249551166965889e-05,
|
| 2240 |
+
"loss": 0.2737,
|
| 2241 |
"step": 3620
|
| 2242 |
},
|
| 2243 |
{
|
| 2244 |
"epoch": 3.26,
|
| 2245 |
"learning_rate": 3.258527827648115e-05,
|
| 2246 |
+
"loss": 0.2965,
|
| 2247 |
"step": 3630
|
| 2248 |
},
|
| 2249 |
{
|
| 2250 |
"epoch": 3.27,
|
| 2251 |
"learning_rate": 3.2675044883303415e-05,
|
| 2252 |
+
"loss": 0.2683,
|
| 2253 |
"step": 3640
|
| 2254 |
},
|
| 2255 |
{
|
| 2256 |
"epoch": 3.27,
|
| 2257 |
"learning_rate": 3.276481149012567e-05,
|
| 2258 |
+
"loss": 0.2379,
|
| 2259 |
"step": 3650
|
| 2260 |
},
|
| 2261 |
{
|
| 2262 |
"epoch": 3.28,
|
| 2263 |
"learning_rate": 3.285457809694794e-05,
|
| 2264 |
+
"loss": 0.3282,
|
| 2265 |
"step": 3660
|
| 2266 |
},
|
| 2267 |
{
|
| 2268 |
"epoch": 3.29,
|
| 2269 |
"learning_rate": 3.2944344703770195e-05,
|
| 2270 |
+
"loss": 0.2589,
|
| 2271 |
"step": 3670
|
| 2272 |
},
|
| 2273 |
{
|
| 2274 |
"epoch": 3.3,
|
| 2275 |
"learning_rate": 3.3034111310592466e-05,
|
| 2276 |
+
"loss": 0.2955,
|
| 2277 |
"step": 3680
|
| 2278 |
},
|
| 2279 |
{
|
| 2280 |
"epoch": 3.31,
|
| 2281 |
"learning_rate": 3.3123877917414724e-05,
|
| 2282 |
+
"loss": 0.2584,
|
| 2283 |
"step": 3690
|
| 2284 |
},
|
| 2285 |
{
|
| 2286 |
"epoch": 3.32,
|
| 2287 |
"learning_rate": 3.321364452423699e-05,
|
| 2288 |
+
"loss": 0.2643,
|
| 2289 |
"step": 3700
|
| 2290 |
},
|
| 2291 |
{
|
| 2292 |
"epoch": 3.33,
|
| 2293 |
"learning_rate": 3.330341113105925e-05,
|
| 2294 |
+
"loss": 0.2987,
|
| 2295 |
"step": 3710
|
| 2296 |
},
|
| 2297 |
{
|
| 2298 |
"epoch": 3.34,
|
| 2299 |
"learning_rate": 3.339317773788151e-05,
|
| 2300 |
+
"loss": 0.3,
|
| 2301 |
"step": 3720
|
| 2302 |
},
|
| 2303 |
{
|
| 2304 |
"epoch": 3.35,
|
| 2305 |
"learning_rate": 3.348294434470377e-05,
|
| 2306 |
+
"loss": 0.2488,
|
| 2307 |
"step": 3730
|
| 2308 |
},
|
| 2309 |
{
|
| 2310 |
"epoch": 3.36,
|
| 2311 |
"learning_rate": 3.3572710951526034e-05,
|
| 2312 |
+
"loss": 0.252,
|
| 2313 |
"step": 3740
|
| 2314 |
},
|
| 2315 |
{
|
| 2316 |
"epoch": 3.36,
|
| 2317 |
"learning_rate": 3.36624775583483e-05,
|
| 2318 |
+
"loss": 0.286,
|
| 2319 |
"step": 3750
|
| 2320 |
},
|
| 2321 |
{
|
| 2322 |
"epoch": 3.37,
|
| 2323 |
"learning_rate": 3.375224416517056e-05,
|
| 2324 |
+
"loss": 0.2801,
|
| 2325 |
"step": 3760
|
| 2326 |
},
|
| 2327 |
{
|
| 2328 |
"epoch": 3.38,
|
| 2329 |
"learning_rate": 3.384201077199282e-05,
|
| 2330 |
+
"loss": 0.2463,
|
| 2331 |
"step": 3770
|
| 2332 |
},
|
| 2333 |
{
|
| 2334 |
"epoch": 3.39,
|
| 2335 |
"learning_rate": 3.3931777378815085e-05,
|
| 2336 |
+
"loss": 0.2824,
|
| 2337 |
"step": 3780
|
| 2338 |
},
|
| 2339 |
{
|
| 2340 |
"epoch": 3.4,
|
| 2341 |
"learning_rate": 3.402154398563734e-05,
|
| 2342 |
+
"loss": 0.2847,
|
| 2343 |
"step": 3790
|
| 2344 |
},
|
| 2345 |
{
|
| 2346 |
"epoch": 3.41,
|
| 2347 |
"learning_rate": 3.411131059245961e-05,
|
| 2348 |
+
"loss": 0.3061,
|
| 2349 |
"step": 3800
|
| 2350 |
},
|
| 2351 |
{
|
| 2352 |
"epoch": 3.42,
|
| 2353 |
"learning_rate": 3.4201077199281866e-05,
|
| 2354 |
+
"loss": 0.2799,
|
| 2355 |
"step": 3810
|
| 2356 |
},
|
| 2357 |
{
|
| 2358 |
"epoch": 3.43,
|
| 2359 |
"learning_rate": 3.429084380610413e-05,
|
| 2360 |
+
"loss": 0.2532,
|
| 2361 |
"step": 3820
|
| 2362 |
},
|
| 2363 |
{
|
| 2364 |
"epoch": 3.44,
|
| 2365 |
"learning_rate": 3.4380610412926395e-05,
|
| 2366 |
+
"loss": 0.2761,
|
| 2367 |
"step": 3830
|
| 2368 |
},
|
| 2369 |
{
|
| 2370 |
"epoch": 3.44,
|
| 2371 |
"learning_rate": 3.447037701974866e-05,
|
| 2372 |
+
"loss": 0.3087,
|
| 2373 |
"step": 3840
|
| 2374 |
},
|
| 2375 |
{
|
| 2376 |
"epoch": 3.45,
|
| 2377 |
"learning_rate": 3.456014362657092e-05,
|
| 2378 |
+
"loss": 0.2841,
|
| 2379 |
"step": 3850
|
| 2380 |
},
|
| 2381 |
{
|
| 2382 |
"epoch": 3.46,
|
| 2383 |
"learning_rate": 3.464991023339318e-05,
|
| 2384 |
+
"loss": 0.2191,
|
| 2385 |
"step": 3860
|
| 2386 |
},
|
| 2387 |
{
|
| 2388 |
"epoch": 3.47,
|
| 2389 |
"learning_rate": 3.473967684021544e-05,
|
| 2390 |
+
"loss": 0.328,
|
| 2391 |
"step": 3870
|
| 2392 |
},
|
| 2393 |
{
|
| 2394 |
"epoch": 3.48,
|
| 2395 |
"learning_rate": 3.4829443447037704e-05,
|
| 2396 |
+
"loss": 0.2377,
|
| 2397 |
"step": 3880
|
| 2398 |
},
|
| 2399 |
{
|
| 2400 |
"epoch": 3.49,
|
| 2401 |
"learning_rate": 3.491921005385996e-05,
|
| 2402 |
+
"loss": 0.2844,
|
| 2403 |
"step": 3890
|
| 2404 |
},
|
| 2405 |
{
|
| 2406 |
"epoch": 3.5,
|
| 2407 |
"learning_rate": 3.500897666068223e-05,
|
| 2408 |
+
"loss": 0.2944,
|
| 2409 |
"step": 3900
|
| 2410 |
},
|
| 2411 |
{
|
| 2412 |
"epoch": 3.51,
|
| 2413 |
"learning_rate": 3.509874326750449e-05,
|
| 2414 |
+
"loss": 0.2371,
|
| 2415 |
"step": 3910
|
| 2416 |
},
|
| 2417 |
{
|
| 2418 |
"epoch": 3.52,
|
| 2419 |
"learning_rate": 3.5188509874326756e-05,
|
| 2420 |
+
"loss": 0.2552,
|
| 2421 |
"step": 3920
|
| 2422 |
},
|
| 2423 |
{
|
| 2424 |
"epoch": 3.53,
|
| 2425 |
"learning_rate": 3.5278276481149014e-05,
|
| 2426 |
+
"loss": 0.2492,
|
| 2427 |
"step": 3930
|
| 2428 |
},
|
| 2429 |
{
|
| 2430 |
"epoch": 3.53,
|
| 2431 |
"learning_rate": 3.536804308797128e-05,
|
| 2432 |
+
"loss": 0.253,
|
| 2433 |
"step": 3940
|
| 2434 |
},
|
| 2435 |
{
|
| 2436 |
"epoch": 3.54,
|
| 2437 |
"learning_rate": 3.5457809694793536e-05,
|
| 2438 |
+
"loss": 0.2794,
|
| 2439 |
"step": 3950
|
| 2440 |
},
|
| 2441 |
{
|
| 2442 |
"epoch": 3.55,
|
| 2443 |
"learning_rate": 3.55475763016158e-05,
|
| 2444 |
+
"loss": 0.2253,
|
| 2445 |
"step": 3960
|
| 2446 |
},
|
| 2447 |
{
|
| 2448 |
"epoch": 3.56,
|
| 2449 |
"learning_rate": 3.5637342908438065e-05,
|
| 2450 |
+
"loss": 0.2894,
|
| 2451 |
"step": 3970
|
| 2452 |
},
|
| 2453 |
{
|
| 2454 |
"epoch": 3.57,
|
| 2455 |
"learning_rate": 3.572710951526032e-05,
|
| 2456 |
+
"loss": 0.2076,
|
| 2457 |
"step": 3980
|
| 2458 |
},
|
| 2459 |
{
|
| 2460 |
"epoch": 3.58,
|
| 2461 |
"learning_rate": 3.581687612208259e-05,
|
| 2462 |
+
"loss": 0.2241,
|
| 2463 |
"step": 3990
|
| 2464 |
},
|
| 2465 |
{
|
| 2466 |
"epoch": 3.59,
|
| 2467 |
"learning_rate": 3.5906642728904846e-05,
|
| 2468 |
+
"loss": 0.2547,
|
| 2469 |
"step": 4000
|
| 2470 |
},
|
| 2471 |
{
|
| 2472 |
"epoch": 3.6,
|
| 2473 |
"learning_rate": 3.599640933572711e-05,
|
| 2474 |
+
"loss": 0.2471,
|
| 2475 |
"step": 4010
|
| 2476 |
},
|
| 2477 |
{
|
| 2478 |
"epoch": 3.61,
|
| 2479 |
"learning_rate": 3.6086175942549375e-05,
|
| 2480 |
+
"loss": 0.2585,
|
| 2481 |
"step": 4020
|
| 2482 |
},
|
| 2483 |
{
|
| 2484 |
"epoch": 3.62,
|
| 2485 |
"learning_rate": 3.617594254937163e-05,
|
| 2486 |
+
"loss": 0.2617,
|
| 2487 |
"step": 4030
|
| 2488 |
},
|
| 2489 |
{
|
| 2490 |
"epoch": 3.62,
|
| 2491 |
"learning_rate": 3.62657091561939e-05,
|
| 2492 |
+
"loss": 0.2724,
|
| 2493 |
"step": 4040
|
| 2494 |
},
|
| 2495 |
{
|
| 2496 |
"epoch": 3.63,
|
| 2497 |
"learning_rate": 3.635547576301616e-05,
|
| 2498 |
+
"loss": 0.2353,
|
| 2499 |
"step": 4050
|
| 2500 |
},
|
| 2501 |
{
|
| 2502 |
"epoch": 3.64,
|
| 2503 |
"learning_rate": 3.644524236983842e-05,
|
| 2504 |
+
"loss": 0.2288,
|
| 2505 |
"step": 4060
|
| 2506 |
},
|
| 2507 |
{
|
| 2508 |
"epoch": 3.65,
|
| 2509 |
"learning_rate": 3.6535008976660684e-05,
|
| 2510 |
+
"loss": 0.2431,
|
| 2511 |
"step": 4070
|
| 2512 |
},
|
| 2513 |
{
|
| 2514 |
"epoch": 3.66,
|
| 2515 |
"learning_rate": 3.662477558348294e-05,
|
| 2516 |
+
"loss": 0.2693,
|
| 2517 |
"step": 4080
|
| 2518 |
},
|
| 2519 |
{
|
| 2520 |
"epoch": 3.67,
|
| 2521 |
"learning_rate": 3.671454219030521e-05,
|
| 2522 |
+
"loss": 0.2903,
|
| 2523 |
"step": 4090
|
| 2524 |
},
|
| 2525 |
{
|
| 2526 |
"epoch": 3.68,
|
| 2527 |
"learning_rate": 3.6804308797127465e-05,
|
| 2528 |
+
"loss": 0.2658,
|
| 2529 |
"step": 4100
|
| 2530 |
},
|
| 2531 |
{
|
| 2532 |
"epoch": 3.69,
|
| 2533 |
"learning_rate": 3.6894075403949736e-05,
|
| 2534 |
+
"loss": 0.2175,
|
| 2535 |
"step": 4110
|
| 2536 |
},
|
| 2537 |
{
|
| 2538 |
"epoch": 3.7,
|
| 2539 |
"learning_rate": 3.6983842010771994e-05,
|
| 2540 |
+
"loss": 0.3072,
|
| 2541 |
"step": 4120
|
| 2542 |
},
|
| 2543 |
{
|
| 2544 |
"epoch": 3.7,
|
| 2545 |
"learning_rate": 3.707360861759426e-05,
|
| 2546 |
+
"loss": 0.305,
|
| 2547 |
"step": 4130
|
| 2548 |
},
|
| 2549 |
{
|
| 2550 |
"epoch": 3.71,
|
| 2551 |
"learning_rate": 3.7163375224416516e-05,
|
| 2552 |
+
"loss": 0.2475,
|
| 2553 |
"step": 4140
|
| 2554 |
},
|
| 2555 |
{
|
| 2556 |
"epoch": 3.72,
|
| 2557 |
"learning_rate": 3.725314183123878e-05,
|
| 2558 |
+
"loss": 0.2682,
|
| 2559 |
"step": 4150
|
| 2560 |
},
|
| 2561 |
{
|
| 2562 |
"epoch": 3.73,
|
| 2563 |
"learning_rate": 3.734290843806104e-05,
|
| 2564 |
+
"loss": 0.2593,
|
| 2565 |
"step": 4160
|
| 2566 |
},
|
| 2567 |
{
|
| 2568 |
"epoch": 3.74,
|
| 2569 |
"learning_rate": 3.74326750448833e-05,
|
| 2570 |
+
"loss": 0.3293,
|
| 2571 |
"step": 4170
|
| 2572 |
},
|
| 2573 |
{
|
| 2574 |
"epoch": 3.75,
|
| 2575 |
"learning_rate": 3.752244165170557e-05,
|
| 2576 |
+
"loss": 0.2422,
|
| 2577 |
"step": 4180
|
| 2578 |
},
|
| 2579 |
{
|
| 2580 |
"epoch": 3.76,
|
| 2581 |
"learning_rate": 3.761220825852783e-05,
|
| 2582 |
+
"loss": 0.2018,
|
| 2583 |
"step": 4190
|
| 2584 |
},
|
| 2585 |
{
|
| 2586 |
"epoch": 3.77,
|
| 2587 |
"learning_rate": 3.770197486535009e-05,
|
| 2588 |
+
"loss": 0.2802,
|
| 2589 |
"step": 4200
|
| 2590 |
},
|
| 2591 |
{
|
| 2592 |
"epoch": 3.78,
|
| 2593 |
"learning_rate": 3.7791741472172355e-05,
|
| 2594 |
+
"loss": 0.2298,
|
| 2595 |
"step": 4210
|
| 2596 |
},
|
| 2597 |
{
|
| 2598 |
"epoch": 3.79,
|
| 2599 |
"learning_rate": 3.788150807899461e-05,
|
| 2600 |
+
"loss": 0.2512,
|
| 2601 |
"step": 4220
|
| 2602 |
},
|
| 2603 |
{
|
| 2604 |
"epoch": 3.79,
|
| 2605 |
"learning_rate": 3.797127468581688e-05,
|
| 2606 |
+
"loss": 0.2503,
|
| 2607 |
"step": 4230
|
| 2608 |
},
|
| 2609 |
{
|
| 2610 |
"epoch": 3.8,
|
| 2611 |
"learning_rate": 3.8061041292639135e-05,
|
| 2612 |
+
"loss": 0.2244,
|
| 2613 |
"step": 4240
|
| 2614 |
},
|
| 2615 |
{
|
| 2616 |
"epoch": 3.81,
|
| 2617 |
"learning_rate": 3.815080789946141e-05,
|
| 2618 |
+
"loss": 0.2456,
|
| 2619 |
"step": 4250
|
| 2620 |
},
|
| 2621 |
{
|
| 2622 |
"epoch": 3.82,
|
| 2623 |
"learning_rate": 3.8240574506283664e-05,
|
| 2624 |
+
"loss": 0.2085,
|
| 2625 |
"step": 4260
|
| 2626 |
},
|
| 2627 |
{
|
| 2628 |
"epoch": 3.83,
|
| 2629 |
"learning_rate": 3.833034111310593e-05,
|
| 2630 |
+
"loss": 0.2367,
|
| 2631 |
"step": 4270
|
| 2632 |
},
|
| 2633 |
{
|
| 2634 |
"epoch": 3.84,
|
| 2635 |
"learning_rate": 3.842010771992819e-05,
|
| 2636 |
+
"loss": 0.2174,
|
| 2637 |
"step": 4280
|
| 2638 |
},
|
| 2639 |
{
|
| 2640 |
"epoch": 3.85,
|
| 2641 |
"learning_rate": 3.850987432675045e-05,
|
| 2642 |
+
"loss": 0.2256,
|
| 2643 |
"step": 4290
|
| 2644 |
},
|
| 2645 |
{
|
| 2646 |
"epoch": 3.86,
|
| 2647 |
"learning_rate": 3.859964093357271e-05,
|
| 2648 |
+
"loss": 0.2344,
|
| 2649 |
"step": 4300
|
| 2650 |
},
|
| 2651 |
{
|
| 2652 |
"epoch": 3.87,
|
| 2653 |
"learning_rate": 3.8689407540394974e-05,
|
| 2654 |
+
"loss": 0.2107,
|
| 2655 |
"step": 4310
|
| 2656 |
},
|
| 2657 |
{
|
| 2658 |
"epoch": 3.88,
|
| 2659 |
"learning_rate": 3.877917414721723e-05,
|
| 2660 |
+
"loss": 0.2797,
|
| 2661 |
"step": 4320
|
| 2662 |
},
|
| 2663 |
{
|
| 2664 |
"epoch": 3.88,
|
| 2665 |
"learning_rate": 3.88689407540395e-05,
|
| 2666 |
+
"loss": 0.2568,
|
| 2667 |
"step": 4330
|
| 2668 |
},
|
| 2669 |
{
|
| 2670 |
"epoch": 3.89,
|
| 2671 |
"learning_rate": 3.895870736086176e-05,
|
| 2672 |
+
"loss": 0.2312,
|
| 2673 |
"step": 4340
|
| 2674 |
},
|
| 2675 |
{
|
| 2676 |
"epoch": 3.9,
|
| 2677 |
"learning_rate": 3.9048473967684026e-05,
|
| 2678 |
+
"loss": 0.2269,
|
| 2679 |
"step": 4350
|
| 2680 |
},
|
| 2681 |
{
|
| 2682 |
"epoch": 3.91,
|
| 2683 |
"learning_rate": 3.9138240574506283e-05,
|
| 2684 |
+
"loss": 0.2412,
|
| 2685 |
"step": 4360
|
| 2686 |
},
|
| 2687 |
{
|
| 2688 |
"epoch": 3.92,
|
| 2689 |
"learning_rate": 3.922800718132855e-05,
|
| 2690 |
+
"loss": 0.2626,
|
| 2691 |
"step": 4370
|
| 2692 |
},
|
| 2693 |
{
|
| 2694 |
"epoch": 3.93,
|
| 2695 |
"learning_rate": 3.9317773788150806e-05,
|
| 2696 |
+
"loss": 0.2995,
|
| 2697 |
"step": 4380
|
| 2698 |
},
|
| 2699 |
{
|
| 2700 |
"epoch": 3.94,
|
| 2701 |
"learning_rate": 3.940754039497307e-05,
|
| 2702 |
+
"loss": 0.2435,
|
| 2703 |
"step": 4390
|
| 2704 |
},
|
| 2705 |
{
|
| 2706 |
"epoch": 3.95,
|
| 2707 |
"learning_rate": 3.9497307001795335e-05,
|
| 2708 |
+
"loss": 0.2317,
|
| 2709 |
"step": 4400
|
| 2710 |
},
|
| 2711 |
{
|
| 2712 |
"epoch": 3.96,
|
| 2713 |
"learning_rate": 3.95870736086176e-05,
|
| 2714 |
+
"loss": 0.1932,
|
| 2715 |
"step": 4410
|
| 2716 |
},
|
| 2717 |
{
|
| 2718 |
"epoch": 3.97,
|
| 2719 |
"learning_rate": 3.967684021543986e-05,
|
| 2720 |
+
"loss": 0.2498,
|
| 2721 |
"step": 4420
|
| 2722 |
},
|
| 2723 |
{
|
| 2724 |
"epoch": 3.97,
|
| 2725 |
"learning_rate": 3.976660682226212e-05,
|
| 2726 |
+
"loss": 0.2792,
|
| 2727 |
"step": 4430
|
| 2728 |
},
|
| 2729 |
{
|
| 2730 |
"epoch": 3.98,
|
| 2731 |
"learning_rate": 3.985637342908438e-05,
|
| 2732 |
+
"loss": 0.2554,
|
| 2733 |
"step": 4440
|
| 2734 |
},
|
| 2735 |
{
|
| 2736 |
"epoch": 3.99,
|
| 2737 |
"learning_rate": 3.9946140035906645e-05,
|
| 2738 |
+
"loss": 0.3403,
|
| 2739 |
"step": 4450
|
| 2740 |
},
|
| 2741 |
{
|
| 2742 |
"epoch": 4.0,
|
| 2743 |
"eval_accuracy": {
|
| 2744 |
+
"accuracy": 0.9382148463781117
|
| 2745 |
},
|
| 2746 |
"eval_f1": {
|
| 2747 |
+
"f1": 0.9358183038090213
|
| 2748 |
},
|
| 2749 |
+
"eval_loss": 0.18279214203357697,
|
| 2750 |
"eval_precision": {
|
| 2751 |
+
"precision": 0.9351989305945833
|
| 2752 |
},
|
| 2753 |
"eval_recall": {
|
| 2754 |
+
"recall": 0.9374254756058774
|
| 2755 |
},
|
| 2756 |
+
"eval_runtime": 165.3018,
|
| 2757 |
+
"eval_samples_per_second": 107.9,
|
| 2758 |
+
"eval_steps_per_second": 6.745,
|
| 2759 |
"step": 4459
|
| 2760 |
},
|
| 2761 |
{
|
| 2762 |
"epoch": 4.0,
|
| 2763 |
"learning_rate": 4.00359066427289e-05,
|
| 2764 |
+
"loss": 0.2936,
|
| 2765 |
"step": 4460
|
| 2766 |
},
|
| 2767 |
{
|
| 2768 |
"epoch": 4.01,
|
| 2769 |
"learning_rate": 4.0125673249551174e-05,
|
| 2770 |
+
"loss": 0.2723,
|
| 2771 |
"step": 4470
|
| 2772 |
},
|
| 2773 |
{
|
| 2774 |
"epoch": 4.02,
|
| 2775 |
"learning_rate": 4.021543985637343e-05,
|
| 2776 |
+
"loss": 0.2307,
|
| 2777 |
"step": 4480
|
| 2778 |
},
|
| 2779 |
{
|
| 2780 |
"epoch": 4.03,
|
| 2781 |
"learning_rate": 4.0305206463195696e-05,
|
| 2782 |
+
"loss": 0.2567,
|
| 2783 |
"step": 4490
|
| 2784 |
},
|
| 2785 |
{
|
| 2786 |
"epoch": 4.04,
|
| 2787 |
"learning_rate": 4.0394973070017954e-05,
|
| 2788 |
+
"loss": 0.2273,
|
| 2789 |
"step": 4500
|
| 2790 |
},
|
| 2791 |
{
|
| 2792 |
"epoch": 4.05,
|
| 2793 |
"learning_rate": 4.048473967684022e-05,
|
| 2794 |
+
"loss": 0.2638,
|
| 2795 |
"step": 4510
|
| 2796 |
},
|
| 2797 |
{
|
| 2798 |
"epoch": 4.05,
|
| 2799 |
"learning_rate": 4.0574506283662476e-05,
|
| 2800 |
+
"loss": 0.2625,
|
| 2801 |
"step": 4520
|
| 2802 |
},
|
| 2803 |
{
|
| 2804 |
"epoch": 4.06,
|
| 2805 |
"learning_rate": 4.066427289048474e-05,
|
| 2806 |
+
"loss": 0.2432,
|
| 2807 |
"step": 4530
|
| 2808 |
},
|
| 2809 |
{
|
| 2810 |
"epoch": 4.07,
|
| 2811 |
"learning_rate": 4.0754039497307006e-05,
|
| 2812 |
+
"loss": 0.2345,
|
| 2813 |
"step": 4540
|
| 2814 |
},
|
| 2815 |
{
|
| 2816 |
"epoch": 4.08,
|
| 2817 |
"learning_rate": 4.084380610412927e-05,
|
| 2818 |
+
"loss": 0.2272,
|
| 2819 |
"step": 4550
|
| 2820 |
},
|
| 2821 |
{
|
| 2822 |
"epoch": 4.09,
|
| 2823 |
"learning_rate": 4.093357271095153e-05,
|
| 2824 |
+
"loss": 0.2217,
|
| 2825 |
"step": 4560
|
| 2826 |
},
|
| 2827 |
{
|
| 2828 |
"epoch": 4.1,
|
| 2829 |
"learning_rate": 4.102333931777379e-05,
|
| 2830 |
+
"loss": 0.2195,
|
| 2831 |
"step": 4570
|
| 2832 |
},
|
| 2833 |
{
|
| 2834 |
"epoch": 4.11,
|
| 2835 |
"learning_rate": 4.111310592459605e-05,
|
| 2836 |
+
"loss": 0.2637,
|
| 2837 |
"step": 4580
|
| 2838 |
},
|
| 2839 |
{
|
| 2840 |
"epoch": 4.12,
|
| 2841 |
"learning_rate": 4.1202872531418315e-05,
|
| 2842 |
+
"loss": 0.2668,
|
| 2843 |
"step": 4590
|
| 2844 |
},
|
| 2845 |
{
|
| 2846 |
"epoch": 4.13,
|
| 2847 |
"learning_rate": 4.129263913824057e-05,
|
| 2848 |
+
"loss": 0.2742,
|
| 2849 |
"step": 4600
|
| 2850 |
},
|
| 2851 |
{
|
| 2852 |
"epoch": 4.14,
|
| 2853 |
"learning_rate": 4.1382405745062844e-05,
|
| 2854 |
+
"loss": 0.2128,
|
| 2855 |
"step": 4610
|
| 2856 |
},
|
| 2857 |
{
|
| 2858 |
"epoch": 4.14,
|
| 2859 |
"learning_rate": 4.14721723518851e-05,
|
| 2860 |
+
"loss": 0.241,
|
| 2861 |
"step": 4620
|
| 2862 |
},
|
| 2863 |
{
|
| 2864 |
"epoch": 4.15,
|
| 2865 |
"learning_rate": 4.156193895870737e-05,
|
| 2866 |
+
"loss": 0.2332,
|
| 2867 |
"step": 4630
|
| 2868 |
},
|
| 2869 |
{
|
| 2870 |
"epoch": 4.16,
|
| 2871 |
"learning_rate": 4.1651705565529625e-05,
|
| 2872 |
+
"loss": 0.2359,
|
| 2873 |
"step": 4640
|
| 2874 |
},
|
| 2875 |
{
|
| 2876 |
"epoch": 4.17,
|
| 2877 |
"learning_rate": 4.174147217235189e-05,
|
| 2878 |
+
"loss": 0.2219,
|
| 2879 |
"step": 4650
|
| 2880 |
},
|
| 2881 |
{
|
| 2882 |
"epoch": 4.18,
|
| 2883 |
"learning_rate": 4.183123877917415e-05,
|
| 2884 |
+
"loss": 0.2325,
|
| 2885 |
"step": 4660
|
| 2886 |
},
|
| 2887 |
{
|
| 2888 |
"epoch": 4.19,
|
| 2889 |
"learning_rate": 4.192100538599641e-05,
|
| 2890 |
+
"loss": 0.2627,
|
| 2891 |
"step": 4670
|
| 2892 |
},
|
| 2893 |
{
|
| 2894 |
"epoch": 4.2,
|
| 2895 |
"learning_rate": 4.2010771992818676e-05,
|
| 2896 |
+
"loss": 0.196,
|
| 2897 |
"step": 4680
|
| 2898 |
},
|
| 2899 |
{
|
| 2900 |
"epoch": 4.21,
|
| 2901 |
"learning_rate": 4.2100538599640934e-05,
|
| 2902 |
+
"loss": 0.234,
|
| 2903 |
"step": 4690
|
| 2904 |
},
|
| 2905 |
{
|
| 2906 |
"epoch": 4.22,
|
| 2907 |
"learning_rate": 4.21903052064632e-05,
|
| 2908 |
+
"loss": 0.1998,
|
| 2909 |
"step": 4700
|
| 2910 |
},
|
| 2911 |
{
|
| 2912 |
"epoch": 4.23,
|
| 2913 |
"learning_rate": 4.228007181328546e-05,
|
| 2914 |
+
"loss": 0.2579,
|
| 2915 |
"step": 4710
|
| 2916 |
},
|
| 2917 |
{
|
| 2918 |
"epoch": 4.23,
|
| 2919 |
"learning_rate": 4.236983842010772e-05,
|
| 2920 |
+
"loss": 0.2678,
|
| 2921 |
"step": 4720
|
| 2922 |
},
|
| 2923 |
{
|
| 2924 |
"epoch": 4.24,
|
| 2925 |
"learning_rate": 4.2459605026929986e-05,
|
| 2926 |
+
"loss": 0.2431,
|
| 2927 |
"step": 4730
|
| 2928 |
},
|
| 2929 |
{
|
| 2930 |
"epoch": 4.25,
|
| 2931 |
"learning_rate": 4.2549371633752244e-05,
|
| 2932 |
+
"loss": 0.2274,
|
| 2933 |
"step": 4740
|
| 2934 |
},
|
| 2935 |
{
|
| 2936 |
"epoch": 4.26,
|
| 2937 |
"learning_rate": 4.263913824057451e-05,
|
| 2938 |
+
"loss": 0.2325,
|
| 2939 |
"step": 4750
|
| 2940 |
},
|
| 2941 |
{
|
| 2942 |
"epoch": 4.27,
|
| 2943 |
"learning_rate": 4.272890484739677e-05,
|
| 2944 |
+
"loss": 0.2121,
|
| 2945 |
"step": 4760
|
| 2946 |
},
|
| 2947 |
{
|
| 2948 |
"epoch": 4.28,
|
| 2949 |
"learning_rate": 4.281867145421903e-05,
|
| 2950 |
+
"loss": 0.2466,
|
| 2951 |
"step": 4770
|
| 2952 |
},
|
| 2953 |
{
|
| 2954 |
"epoch": 4.29,
|
| 2955 |
"learning_rate": 4.2908438061041295e-05,
|
| 2956 |
+
"loss": 0.2331,
|
| 2957 |
"step": 4780
|
| 2958 |
},
|
| 2959 |
{
|
| 2960 |
"epoch": 4.3,
|
| 2961 |
"learning_rate": 4.299820466786356e-05,
|
| 2962 |
+
"loss": 0.221,
|
| 2963 |
"step": 4790
|
| 2964 |
},
|
| 2965 |
{
|
| 2966 |
"epoch": 4.31,
|
| 2967 |
"learning_rate": 4.308797127468582e-05,
|
| 2968 |
+
"loss": 0.2563,
|
| 2969 |
"step": 4800
|
| 2970 |
},
|
| 2971 |
{
|
| 2972 |
"epoch": 4.31,
|
| 2973 |
"learning_rate": 4.317773788150808e-05,
|
| 2974 |
+
"loss": 0.2564,
|
| 2975 |
"step": 4810
|
| 2976 |
},
|
| 2977 |
{
|
| 2978 |
"epoch": 4.32,
|
| 2979 |
"learning_rate": 4.326750448833034e-05,
|
| 2980 |
+
"loss": 0.2369,
|
| 2981 |
"step": 4820
|
| 2982 |
},
|
| 2983 |
{
|
| 2984 |
"epoch": 4.33,
|
| 2985 |
"learning_rate": 4.3357271095152605e-05,
|
| 2986 |
+
"loss": 0.1929,
|
| 2987 |
"step": 4830
|
| 2988 |
},
|
| 2989 |
{
|
| 2990 |
"epoch": 4.34,
|
| 2991 |
"learning_rate": 4.344703770197487e-05,
|
| 2992 |
+
"loss": 0.2639,
|
| 2993 |
"step": 4840
|
| 2994 |
},
|
| 2995 |
{
|
| 2996 |
"epoch": 4.35,
|
| 2997 |
"learning_rate": 4.353680430879713e-05,
|
| 2998 |
+
"loss": 0.232,
|
| 2999 |
"step": 4850
|
| 3000 |
},
|
| 3001 |
{
|
| 3002 |
"epoch": 4.36,
|
| 3003 |
"learning_rate": 4.362657091561939e-05,
|
| 3004 |
+
"loss": 0.2453,
|
| 3005 |
"step": 4860
|
| 3006 |
},
|
| 3007 |
{
|
| 3008 |
"epoch": 4.37,
|
| 3009 |
"learning_rate": 4.371633752244165e-05,
|
| 3010 |
+
"loss": 0.2145,
|
| 3011 |
"step": 4870
|
| 3012 |
},
|
| 3013 |
{
|
| 3014 |
"epoch": 4.38,
|
| 3015 |
"learning_rate": 4.3806104129263914e-05,
|
| 3016 |
+
"loss": 0.2249,
|
| 3017 |
"step": 4880
|
| 3018 |
},
|
| 3019 |
{
|
| 3020 |
"epoch": 4.39,
|
| 3021 |
"learning_rate": 4.389587073608618e-05,
|
| 3022 |
+
"loss": 0.2861,
|
| 3023 |
"step": 4890
|
| 3024 |
},
|
| 3025 |
{
|
| 3026 |
"epoch": 4.4,
|
| 3027 |
"learning_rate": 4.398563734290844e-05,
|
| 3028 |
+
"loss": 0.2452,
|
| 3029 |
"step": 4900
|
| 3030 |
},
|
| 3031 |
{
|
| 3032 |
"epoch": 4.4,
|
| 3033 |
"learning_rate": 4.40754039497307e-05,
|
| 3034 |
+
"loss": 0.2512,
|
| 3035 |
"step": 4910
|
| 3036 |
},
|
| 3037 |
{
|
| 3038 |
"epoch": 4.41,
|
| 3039 |
"learning_rate": 4.4165170556552966e-05,
|
| 3040 |
+
"loss": 0.2373,
|
| 3041 |
"step": 4920
|
| 3042 |
},
|
| 3043 |
{
|
| 3044 |
"epoch": 4.42,
|
| 3045 |
"learning_rate": 4.4254937163375224e-05,
|
| 3046 |
+
"loss": 0.2262,
|
| 3047 |
"step": 4930
|
| 3048 |
},
|
| 3049 |
{
|
| 3050 |
"epoch": 4.43,
|
| 3051 |
"learning_rate": 4.434470377019749e-05,
|
| 3052 |
+
"loss": 0.2243,
|
| 3053 |
"step": 4940
|
| 3054 |
},
|
| 3055 |
{
|
| 3056 |
"epoch": 4.44,
|
| 3057 |
"learning_rate": 4.4434470377019746e-05,
|
| 3058 |
+
"loss": 0.2228,
|
| 3059 |
"step": 4950
|
| 3060 |
},
|
| 3061 |
{
|
| 3062 |
"epoch": 4.45,
|
| 3063 |
"learning_rate": 4.452423698384201e-05,
|
| 3064 |
+
"loss": 0.2331,
|
| 3065 |
"step": 4960
|
| 3066 |
},
|
| 3067 |
{
|
| 3068 |
"epoch": 4.46,
|
| 3069 |
"learning_rate": 4.4614003590664275e-05,
|
| 3070 |
+
"loss": 0.2706,
|
| 3071 |
"step": 4970
|
| 3072 |
},
|
| 3073 |
{
|
| 3074 |
"epoch": 4.47,
|
| 3075 |
"learning_rate": 4.470377019748654e-05,
|
| 3076 |
+
"loss": 0.2326,
|
| 3077 |
"step": 4980
|
| 3078 |
},
|
| 3079 |
{
|
| 3080 |
"epoch": 4.48,
|
| 3081 |
"learning_rate": 4.47935368043088e-05,
|
| 3082 |
+
"loss": 0.3095,
|
| 3083 |
"step": 4990
|
| 3084 |
},
|
| 3085 |
{
|
| 3086 |
"epoch": 4.49,
|
| 3087 |
"learning_rate": 4.488330341113106e-05,
|
| 3088 |
+
"loss": 0.2106,
|
| 3089 |
"step": 5000
|
| 3090 |
},
|
| 3091 |
{
|
| 3092 |
"epoch": 4.49,
|
| 3093 |
"learning_rate": 4.497307001795332e-05,
|
| 3094 |
+
"loss": 0.2494,
|
| 3095 |
"step": 5010
|
| 3096 |
},
|
| 3097 |
{
|
| 3098 |
"epoch": 4.5,
|
| 3099 |
"learning_rate": 4.5062836624775585e-05,
|
| 3100 |
+
"loss": 0.234,
|
| 3101 |
"step": 5020
|
| 3102 |
},
|
| 3103 |
{
|
| 3104 |
"epoch": 4.51,
|
| 3105 |
"learning_rate": 4.515260323159784e-05,
|
| 3106 |
+
"loss": 0.283,
|
| 3107 |
"step": 5030
|
| 3108 |
},
|
| 3109 |
{
|
| 3110 |
"epoch": 4.52,
|
| 3111 |
"learning_rate": 4.5242369838420114e-05,
|
| 3112 |
+
"loss": 0.2626,
|
| 3113 |
"step": 5040
|
| 3114 |
},
|
| 3115 |
{
|
| 3116 |
"epoch": 4.53,
|
| 3117 |
"learning_rate": 4.533213644524237e-05,
|
| 3118 |
+
"loss": 0.2352,
|
| 3119 |
"step": 5050
|
| 3120 |
},
|
| 3121 |
{
|
| 3122 |
"epoch": 4.54,
|
| 3123 |
"learning_rate": 4.5421903052064636e-05,
|
| 3124 |
+
"loss": 0.2748,
|
| 3125 |
"step": 5060
|
| 3126 |
},
|
| 3127 |
{
|
| 3128 |
"epoch": 4.55,
|
| 3129 |
"learning_rate": 4.5511669658886894e-05,
|
| 3130 |
+
"loss": 0.2849,
|
| 3131 |
"step": 5070
|
| 3132 |
},
|
| 3133 |
{
|
| 3134 |
"epoch": 4.56,
|
| 3135 |
"learning_rate": 4.560143626570916e-05,
|
| 3136 |
+
"loss": 0.2296,
|
| 3137 |
"step": 5080
|
| 3138 |
},
|
| 3139 |
{
|
| 3140 |
"epoch": 4.57,
|
| 3141 |
"learning_rate": 4.5691202872531417e-05,
|
| 3142 |
+
"loss": 0.2231,
|
| 3143 |
"step": 5090
|
| 3144 |
},
|
| 3145 |
{
|
| 3146 |
"epoch": 4.58,
|
| 3147 |
"learning_rate": 4.578096947935368e-05,
|
| 3148 |
+
"loss": 0.2526,
|
| 3149 |
"step": 5100
|
| 3150 |
},
|
| 3151 |
{
|
| 3152 |
"epoch": 4.58,
|
| 3153 |
"learning_rate": 4.5870736086175946e-05,
|
| 3154 |
+
"loss": 0.2441,
|
| 3155 |
"step": 5110
|
| 3156 |
},
|
| 3157 |
{
|
| 3158 |
"epoch": 4.59,
|
| 3159 |
"learning_rate": 4.596050269299821e-05,
|
| 3160 |
+
"loss": 0.236,
|
| 3161 |
"step": 5120
|
| 3162 |
},
|
| 3163 |
{
|
| 3164 |
"epoch": 4.6,
|
| 3165 |
"learning_rate": 4.605026929982047e-05,
|
| 3166 |
+
"loss": 0.2485,
|
| 3167 |
"step": 5130
|
| 3168 |
},
|
| 3169 |
{
|
| 3170 |
"epoch": 4.61,
|
| 3171 |
"learning_rate": 4.614003590664273e-05,
|
| 3172 |
+
"loss": 0.1969,
|
| 3173 |
"step": 5140
|
| 3174 |
},
|
| 3175 |
{
|
| 3176 |
"epoch": 4.62,
|
| 3177 |
"learning_rate": 4.622980251346499e-05,
|
| 3178 |
+
"loss": 0.2454,
|
| 3179 |
"step": 5150
|
| 3180 |
},
|
| 3181 |
{
|
| 3182 |
"epoch": 4.63,
|
| 3183 |
"learning_rate": 4.6319569120287255e-05,
|
| 3184 |
+
"loss": 0.2486,
|
| 3185 |
"step": 5160
|
| 3186 |
},
|
| 3187 |
{
|
| 3188 |
"epoch": 4.64,
|
| 3189 |
"learning_rate": 4.640933572710951e-05,
|
| 3190 |
+
"loss": 0.2304,
|
| 3191 |
"step": 5170
|
| 3192 |
},
|
| 3193 |
{
|
| 3194 |
"epoch": 4.65,
|
| 3195 |
"learning_rate": 4.6499102333931784e-05,
|
| 3196 |
+
"loss": 0.2321,
|
| 3197 |
"step": 5180
|
| 3198 |
},
|
| 3199 |
{
|
| 3200 |
"epoch": 4.66,
|
| 3201 |
"learning_rate": 4.658886894075404e-05,
|
| 3202 |
+
"loss": 0.2205,
|
| 3203 |
"step": 5190
|
| 3204 |
},
|
| 3205 |
{
|
| 3206 |
"epoch": 4.66,
|
| 3207 |
"learning_rate": 4.667863554757631e-05,
|
| 3208 |
+
"loss": 0.2107,
|
| 3209 |
"step": 5200
|
| 3210 |
},
|
| 3211 |
{
|
| 3212 |
"epoch": 4.67,
|
| 3213 |
"learning_rate": 4.6768402154398565e-05,
|
| 3214 |
+
"loss": 0.2215,
|
| 3215 |
"step": 5210
|
| 3216 |
},
|
| 3217 |
{
|
| 3218 |
"epoch": 4.68,
|
| 3219 |
"learning_rate": 4.685816876122083e-05,
|
| 3220 |
+
"loss": 0.222,
|
| 3221 |
"step": 5220
|
| 3222 |
},
|
| 3223 |
{
|
| 3224 |
"epoch": 4.69,
|
| 3225 |
"learning_rate": 4.694793536804309e-05,
|
| 3226 |
+
"loss": 0.2271,
|
| 3227 |
"step": 5230
|
| 3228 |
},
|
| 3229 |
{
|
| 3230 |
"epoch": 4.7,
|
| 3231 |
"learning_rate": 4.703770197486535e-05,
|
| 3232 |
+
"loss": 0.2342,
|
| 3233 |
"step": 5240
|
| 3234 |
},
|
| 3235 |
{
|
| 3236 |
"epoch": 4.71,
|
| 3237 |
"learning_rate": 4.7127468581687616e-05,
|
| 3238 |
+
"loss": 0.2478,
|
| 3239 |
"step": 5250
|
| 3240 |
},
|
| 3241 |
{
|
| 3242 |
"epoch": 4.72,
|
| 3243 |
"learning_rate": 4.721723518850988e-05,
|
| 3244 |
+
"loss": 0.2691,
|
| 3245 |
"step": 5260
|
| 3246 |
},
|
| 3247 |
{
|
| 3248 |
"epoch": 4.73,
|
| 3249 |
"learning_rate": 4.730700179533214e-05,
|
| 3250 |
+
"loss": 0.2059,
|
| 3251 |
"step": 5270
|
| 3252 |
},
|
| 3253 |
{
|
| 3254 |
"epoch": 4.74,
|
| 3255 |
"learning_rate": 4.7396768402154403e-05,
|
| 3256 |
+
"loss": 0.2085,
|
| 3257 |
"step": 5280
|
| 3258 |
},
|
| 3259 |
{
|
| 3260 |
"epoch": 4.75,
|
| 3261 |
"learning_rate": 4.748653500897666e-05,
|
| 3262 |
+
"loss": 0.1888,
|
| 3263 |
"step": 5290
|
| 3264 |
},
|
| 3265 |
{
|
| 3266 |
"epoch": 4.75,
|
| 3267 |
"learning_rate": 4.7576301615798926e-05,
|
| 3268 |
+
"loss": 0.2933,
|
| 3269 |
"step": 5300
|
| 3270 |
},
|
| 3271 |
{
|
| 3272 |
"epoch": 4.76,
|
| 3273 |
"learning_rate": 4.7666068222621184e-05,
|
| 3274 |
+
"loss": 0.2519,
|
| 3275 |
"step": 5310
|
| 3276 |
},
|
| 3277 |
{
|
| 3278 |
"epoch": 4.77,
|
| 3279 |
"learning_rate": 4.775583482944345e-05,
|
| 3280 |
+
"loss": 0.2142,
|
| 3281 |
"step": 5320
|
| 3282 |
},
|
| 3283 |
{
|
| 3284 |
"epoch": 4.78,
|
| 3285 |
"learning_rate": 4.784560143626571e-05,
|
| 3286 |
+
"loss": 0.1972,
|
| 3287 |
"step": 5330
|
| 3288 |
},
|
| 3289 |
{
|
| 3290 |
"epoch": 4.79,
|
| 3291 |
"learning_rate": 4.793536804308798e-05,
|
| 3292 |
+
"loss": 0.2134,
|
| 3293 |
"step": 5340
|
| 3294 |
},
|
| 3295 |
{
|
| 3296 |
"epoch": 4.8,
|
| 3297 |
"learning_rate": 4.8025134649910235e-05,
|
| 3298 |
+
"loss": 0.2924,
|
| 3299 |
"step": 5350
|
| 3300 |
},
|
| 3301 |
{
|
| 3302 |
"epoch": 4.81,
|
| 3303 |
"learning_rate": 4.81149012567325e-05,
|
| 3304 |
+
"loss": 0.2207,
|
| 3305 |
"step": 5360
|
| 3306 |
},
|
| 3307 |
{
|
| 3308 |
"epoch": 4.82,
|
| 3309 |
"learning_rate": 4.820466786355476e-05,
|
| 3310 |
+
"loss": 0.2369,
|
| 3311 |
"step": 5370
|
| 3312 |
},
|
| 3313 |
{
|
| 3314 |
"epoch": 4.83,
|
| 3315 |
"learning_rate": 4.829443447037702e-05,
|
| 3316 |
+
"loss": 0.242,
|
| 3317 |
"step": 5380
|
| 3318 |
},
|
| 3319 |
{
|
| 3320 |
"epoch": 4.84,
|
| 3321 |
"learning_rate": 4.838420107719928e-05,
|
| 3322 |
+
"loss": 0.2023,
|
| 3323 |
"step": 5390
|
| 3324 |
},
|
| 3325 |
{
|
| 3326 |
"epoch": 4.84,
|
| 3327 |
"learning_rate": 4.847396768402155e-05,
|
| 3328 |
+
"loss": 0.2413,
|
| 3329 |
"step": 5400
|
| 3330 |
},
|
| 3331 |
{
|
| 3332 |
"epoch": 4.85,
|
| 3333 |
"learning_rate": 4.856373429084381e-05,
|
| 3334 |
+
"loss": 0.2757,
|
| 3335 |
"step": 5410
|
| 3336 |
},
|
| 3337 |
{
|
| 3338 |
"epoch": 4.86,
|
| 3339 |
"learning_rate": 4.8653500897666074e-05,
|
| 3340 |
+
"loss": 0.2481,
|
| 3341 |
"step": 5420
|
| 3342 |
},
|
| 3343 |
{
|
| 3344 |
"epoch": 4.87,
|
| 3345 |
"learning_rate": 4.874326750448833e-05,
|
| 3346 |
+
"loss": 0.2487,
|
| 3347 |
"step": 5430
|
| 3348 |
},
|
| 3349 |
{
|
| 3350 |
"epoch": 4.88,
|
| 3351 |
"learning_rate": 4.8833034111310596e-05,
|
| 3352 |
+
"loss": 0.1747,
|
| 3353 |
"step": 5440
|
| 3354 |
},
|
| 3355 |
{
|
| 3356 |
"epoch": 4.89,
|
| 3357 |
"learning_rate": 4.8922800718132854e-05,
|
| 3358 |
+
"loss": 0.2537,
|
| 3359 |
"step": 5450
|
| 3360 |
},
|
| 3361 |
{
|
| 3362 |
"epoch": 4.9,
|
| 3363 |
"learning_rate": 4.901256732495512e-05,
|
| 3364 |
+
"loss": 0.2722,
|
| 3365 |
"step": 5460
|
| 3366 |
},
|
| 3367 |
{
|
| 3368 |
"epoch": 4.91,
|
| 3369 |
"learning_rate": 4.9102333931777383e-05,
|
| 3370 |
+
"loss": 0.2225,
|
| 3371 |
"step": 5470
|
| 3372 |
},
|
| 3373 |
{
|
| 3374 |
"epoch": 4.92,
|
| 3375 |
"learning_rate": 4.919210053859965e-05,
|
| 3376 |
+
"loss": 0.1949,
|
| 3377 |
"step": 5480
|
| 3378 |
},
|
| 3379 |
{
|
| 3380 |
"epoch": 4.92,
|
| 3381 |
"learning_rate": 4.9281867145421906e-05,
|
| 3382 |
+
"loss": 0.2416,
|
| 3383 |
"step": 5490
|
| 3384 |
},
|
| 3385 |
{
|
| 3386 |
"epoch": 4.93,
|
| 3387 |
"learning_rate": 4.937163375224417e-05,
|
| 3388 |
+
"loss": 0.2197,
|
| 3389 |
"step": 5500
|
| 3390 |
},
|
| 3391 |
{
|
| 3392 |
"epoch": 4.94,
|
| 3393 |
"learning_rate": 4.946140035906643e-05,
|
| 3394 |
+
"loss": 0.2939,
|
| 3395 |
"step": 5510
|
| 3396 |
},
|
| 3397 |
{
|
| 3398 |
"epoch": 4.95,
|
| 3399 |
"learning_rate": 4.955116696588869e-05,
|
| 3400 |
+
"loss": 0.2786,
|
| 3401 |
"step": 5520
|
| 3402 |
},
|
| 3403 |
{
|
| 3404 |
"epoch": 4.96,
|
| 3405 |
"learning_rate": 4.964093357271095e-05,
|
| 3406 |
+
"loss": 0.2187,
|
| 3407 |
"step": 5530
|
| 3408 |
},
|
| 3409 |
{
|
| 3410 |
"epoch": 4.97,
|
| 3411 |
"learning_rate": 4.9730700179533215e-05,
|
| 3412 |
+
"loss": 0.2518,
|
| 3413 |
"step": 5540
|
| 3414 |
},
|
| 3415 |
{
|
| 3416 |
"epoch": 4.98,
|
| 3417 |
"learning_rate": 4.982046678635548e-05,
|
| 3418 |
+
"loss": 0.205,
|
| 3419 |
"step": 5550
|
| 3420 |
},
|
| 3421 |
{
|
| 3422 |
"epoch": 4.99,
|
| 3423 |
"learning_rate": 4.991023339317774e-05,
|
| 3424 |
+
"loss": 0.2277,
|
| 3425 |
"step": 5560
|
| 3426 |
},
|
| 3427 |
{
|
| 3428 |
"epoch": 5.0,
|
| 3429 |
"learning_rate": 5e-05,
|
| 3430 |
+
"loss": 0.2399,
|
| 3431 |
"step": 5570
|
| 3432 |
},
|
| 3433 |
{
|
| 3434 |
"epoch": 5.0,
|
| 3435 |
"eval_accuracy": {
|
| 3436 |
+
"accuracy": 0.9539694998878673
|
| 3437 |
},
|
| 3438 |
"eval_f1": {
|
| 3439 |
+
"f1": 0.9516195088904471
|
| 3440 |
},
|
| 3441 |
+
"eval_loss": 0.1283472627401352,
|
| 3442 |
"eval_precision": {
|
| 3443 |
+
"precision": 0.9525864748624857
|
| 3444 |
},
|
| 3445 |
"eval_recall": {
|
| 3446 |
+
"recall": 0.9506985041313745
|
| 3447 |
},
|
| 3448 |
+
"eval_runtime": 164.7789,
|
| 3449 |
+
"eval_samples_per_second": 108.242,
|
| 3450 |
+
"eval_steps_per_second": 6.767,
|
| 3451 |
"step": 5573
|
| 3452 |
}
|
| 3453 |
],
|
tmp-checkpoint-5573/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 sha256:5c55fb57ba4dc2c9b2c813516c4846a21f1047b19b6cf48a97b9224ced523b9b
|
| 3 |
size 4792
|