Delete trainer_state.json with huggingface_hub
Browse files- trainer_state.json +0 -4006
trainer_state.json
DELETED
|
@@ -1,4006 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"best_metric": null,
|
| 3 |
-
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 2.999207397622193,
|
| 5 |
-
"eval_steps": 500,
|
| 6 |
-
"global_step": 2838,
|
| 7 |
-
"is_hyper_param_search": false,
|
| 8 |
-
"is_local_process_zero": true,
|
| 9 |
-
"is_world_process_zero": true,
|
| 10 |
-
"log_history": [
|
| 11 |
-
{
|
| 12 |
-
"epoch": 0.0,
|
| 13 |
-
"grad_norm": 54.60578280011217,
|
| 14 |
-
"learning_rate": 3.4482758620689656e-07,
|
| 15 |
-
"loss": 2.0852,
|
| 16 |
-
"step": 1
|
| 17 |
-
},
|
| 18 |
-
{
|
| 19 |
-
"epoch": 0.01,
|
| 20 |
-
"grad_norm": 42.54745106429926,
|
| 21 |
-
"learning_rate": 1.724137931034483e-06,
|
| 22 |
-
"loss": 2.0381,
|
| 23 |
-
"step": 5
|
| 24 |
-
},
|
| 25 |
-
{
|
| 26 |
-
"epoch": 0.01,
|
| 27 |
-
"grad_norm": 9.212853479586382,
|
| 28 |
-
"learning_rate": 3.448275862068966e-06,
|
| 29 |
-
"loss": 1.6632,
|
| 30 |
-
"step": 10
|
| 31 |
-
},
|
| 32 |
-
{
|
| 33 |
-
"epoch": 0.02,
|
| 34 |
-
"grad_norm": 5.314912275996213,
|
| 35 |
-
"learning_rate": 5.172413793103449e-06,
|
| 36 |
-
"loss": 1.3468,
|
| 37 |
-
"step": 15
|
| 38 |
-
},
|
| 39 |
-
{
|
| 40 |
-
"epoch": 0.02,
|
| 41 |
-
"grad_norm": 4.03433890035307,
|
| 42 |
-
"learning_rate": 6.896551724137932e-06,
|
| 43 |
-
"loss": 1.2538,
|
| 44 |
-
"step": 20
|
| 45 |
-
},
|
| 46 |
-
{
|
| 47 |
-
"epoch": 0.03,
|
| 48 |
-
"grad_norm": 2.6447411660183016,
|
| 49 |
-
"learning_rate": 8.620689655172414e-06,
|
| 50 |
-
"loss": 1.2604,
|
| 51 |
-
"step": 25
|
| 52 |
-
},
|
| 53 |
-
{
|
| 54 |
-
"epoch": 0.03,
|
| 55 |
-
"grad_norm": 2.1857065034640795,
|
| 56 |
-
"learning_rate": 9.999996872939885e-06,
|
| 57 |
-
"loss": 1.1935,
|
| 58 |
-
"step": 30
|
| 59 |
-
},
|
| 60 |
-
{
|
| 61 |
-
"epoch": 0.04,
|
| 62 |
-
"grad_norm": 1.7526909326796276,
|
| 63 |
-
"learning_rate": 9.999887426246524e-06,
|
| 64 |
-
"loss": 1.1939,
|
| 65 |
-
"step": 35
|
| 66 |
-
},
|
| 67 |
-
{
|
| 68 |
-
"epoch": 0.04,
|
| 69 |
-
"grad_norm": 1.5378704654179458,
|
| 70 |
-
"learning_rate": 9.999621630458743e-06,
|
| 71 |
-
"loss": 1.1626,
|
| 72 |
-
"step": 40
|
| 73 |
-
},
|
| 74 |
-
{
|
| 75 |
-
"epoch": 0.05,
|
| 76 |
-
"grad_norm": 1.3071739468687105,
|
| 77 |
-
"learning_rate": 9.999199493888118e-06,
|
| 78 |
-
"loss": 1.122,
|
| 79 |
-
"step": 45
|
| 80 |
-
},
|
| 81 |
-
{
|
| 82 |
-
"epoch": 0.05,
|
| 83 |
-
"grad_norm": 1.1982413397341172,
|
| 84 |
-
"learning_rate": 9.998621029735082e-06,
|
| 85 |
-
"loss": 1.1191,
|
| 86 |
-
"step": 50
|
| 87 |
-
},
|
| 88 |
-
{
|
| 89 |
-
"epoch": 0.06,
|
| 90 |
-
"grad_norm": 1.0956550608235818,
|
| 91 |
-
"learning_rate": 9.997886256088507e-06,
|
| 92 |
-
"loss": 1.123,
|
| 93 |
-
"step": 55
|
| 94 |
-
},
|
| 95 |
-
{
|
| 96 |
-
"epoch": 0.06,
|
| 97 |
-
"grad_norm": 1.0821639142212824,
|
| 98 |
-
"learning_rate": 9.996995195925152e-06,
|
| 99 |
-
"loss": 1.0751,
|
| 100 |
-
"step": 60
|
| 101 |
-
},
|
| 102 |
-
{
|
| 103 |
-
"epoch": 0.07,
|
| 104 |
-
"grad_norm": 1.0865870956103822,
|
| 105 |
-
"learning_rate": 9.995947877108933e-06,
|
| 106 |
-
"loss": 1.114,
|
| 107 |
-
"step": 65
|
| 108 |
-
},
|
| 109 |
-
{
|
| 110 |
-
"epoch": 0.07,
|
| 111 |
-
"grad_norm": 1.0582468505041362,
|
| 112 |
-
"learning_rate": 9.99474433239006e-06,
|
| 113 |
-
"loss": 1.0609,
|
| 114 |
-
"step": 70
|
| 115 |
-
},
|
| 116 |
-
{
|
| 117 |
-
"epoch": 0.08,
|
| 118 |
-
"grad_norm": 1.0556893008132,
|
| 119 |
-
"learning_rate": 9.993384599404001e-06,
|
| 120 |
-
"loss": 1.0861,
|
| 121 |
-
"step": 75
|
| 122 |
-
},
|
| 123 |
-
{
|
| 124 |
-
"epoch": 0.08,
|
| 125 |
-
"grad_norm": 1.053104954690895,
|
| 126 |
-
"learning_rate": 9.991868720670322e-06,
|
| 127 |
-
"loss": 1.0792,
|
| 128 |
-
"step": 80
|
| 129 |
-
},
|
| 130 |
-
{
|
| 131 |
-
"epoch": 0.09,
|
| 132 |
-
"grad_norm": 1.069931327549148,
|
| 133 |
-
"learning_rate": 9.990196743591341e-06,
|
| 134 |
-
"loss": 1.0722,
|
| 135 |
-
"step": 85
|
| 136 |
-
},
|
| 137 |
-
{
|
| 138 |
-
"epoch": 0.1,
|
| 139 |
-
"grad_norm": 1.0793525694046096,
|
| 140 |
-
"learning_rate": 9.988368720450656e-06,
|
| 141 |
-
"loss": 1.0561,
|
| 142 |
-
"step": 90
|
| 143 |
-
},
|
| 144 |
-
{
|
| 145 |
-
"epoch": 0.1,
|
| 146 |
-
"grad_norm": 1.0449093375062701,
|
| 147 |
-
"learning_rate": 9.986384708411507e-06,
|
| 148 |
-
"loss": 1.0675,
|
| 149 |
-
"step": 95
|
| 150 |
-
},
|
| 151 |
-
{
|
| 152 |
-
"epoch": 0.11,
|
| 153 |
-
"grad_norm": 0.9947694351867048,
|
| 154 |
-
"learning_rate": 9.984244769514988e-06,
|
| 155 |
-
"loss": 1.0395,
|
| 156 |
-
"step": 100
|
| 157 |
-
},
|
| 158 |
-
{
|
| 159 |
-
"epoch": 0.11,
|
| 160 |
-
"grad_norm": 1.0281907914775885,
|
| 161 |
-
"learning_rate": 9.981948970678107e-06,
|
| 162 |
-
"loss": 1.0455,
|
| 163 |
-
"step": 105
|
| 164 |
-
},
|
| 165 |
-
{
|
| 166 |
-
"epoch": 0.12,
|
| 167 |
-
"grad_norm": 1.0429607059919377,
|
| 168 |
-
"learning_rate": 9.979497383691695e-06,
|
| 169 |
-
"loss": 1.014,
|
| 170 |
-
"step": 110
|
| 171 |
-
},
|
| 172 |
-
{
|
| 173 |
-
"epoch": 0.12,
|
| 174 |
-
"grad_norm": 1.1551790238118493,
|
| 175 |
-
"learning_rate": 9.976890085218157e-06,
|
| 176 |
-
"loss": 1.0335,
|
| 177 |
-
"step": 115
|
| 178 |
-
},
|
| 179 |
-
{
|
| 180 |
-
"epoch": 0.13,
|
| 181 |
-
"grad_norm": 1.0958017153997939,
|
| 182 |
-
"learning_rate": 9.974127156789082e-06,
|
| 183 |
-
"loss": 1.0566,
|
| 184 |
-
"step": 120
|
| 185 |
-
},
|
| 186 |
-
{
|
| 187 |
-
"epoch": 0.13,
|
| 188 |
-
"grad_norm": 1.0545164230640043,
|
| 189 |
-
"learning_rate": 9.971208684802686e-06,
|
| 190 |
-
"loss": 1.0234,
|
| 191 |
-
"step": 125
|
| 192 |
-
},
|
| 193 |
-
{
|
| 194 |
-
"epoch": 0.14,
|
| 195 |
-
"grad_norm": 1.1119539198620567,
|
| 196 |
-
"learning_rate": 9.968134760521114e-06,
|
| 197 |
-
"loss": 0.9956,
|
| 198 |
-
"step": 130
|
| 199 |
-
},
|
| 200 |
-
{
|
| 201 |
-
"epoch": 0.14,
|
| 202 |
-
"grad_norm": 1.045532676163788,
|
| 203 |
-
"learning_rate": 9.964905480067585e-06,
|
| 204 |
-
"loss": 1.0103,
|
| 205 |
-
"step": 135
|
| 206 |
-
},
|
| 207 |
-
{
|
| 208 |
-
"epoch": 0.15,
|
| 209 |
-
"grad_norm": 1.0818099938062198,
|
| 210 |
-
"learning_rate": 9.96152094442339e-06,
|
| 211 |
-
"loss": 0.987,
|
| 212 |
-
"step": 140
|
| 213 |
-
},
|
| 214 |
-
{
|
| 215 |
-
"epoch": 0.15,
|
| 216 |
-
"grad_norm": 1.06916616510137,
|
| 217 |
-
"learning_rate": 9.957981259424724e-06,
|
| 218 |
-
"loss": 1.0189,
|
| 219 |
-
"step": 145
|
| 220 |
-
},
|
| 221 |
-
{
|
| 222 |
-
"epoch": 0.16,
|
| 223 |
-
"grad_norm": 1.1000812098052206,
|
| 224 |
-
"learning_rate": 9.954286535759394e-06,
|
| 225 |
-
"loss": 1.0025,
|
| 226 |
-
"step": 150
|
| 227 |
-
},
|
| 228 |
-
{
|
| 229 |
-
"epoch": 0.16,
|
| 230 |
-
"grad_norm": 1.0740685860653156,
|
| 231 |
-
"learning_rate": 9.950436888963337e-06,
|
| 232 |
-
"loss": 1.0394,
|
| 233 |
-
"step": 155
|
| 234 |
-
},
|
| 235 |
-
{
|
| 236 |
-
"epoch": 0.17,
|
| 237 |
-
"grad_norm": 1.0578416601226404,
|
| 238 |
-
"learning_rate": 9.946432439417021e-06,
|
| 239 |
-
"loss": 1.0419,
|
| 240 |
-
"step": 160
|
| 241 |
-
},
|
| 242 |
-
{
|
| 243 |
-
"epoch": 0.17,
|
| 244 |
-
"grad_norm": 1.1378367934770748,
|
| 245 |
-
"learning_rate": 9.942273312341679e-06,
|
| 246 |
-
"loss": 1.04,
|
| 247 |
-
"step": 165
|
| 248 |
-
},
|
| 249 |
-
{
|
| 250 |
-
"epoch": 0.18,
|
| 251 |
-
"grad_norm": 1.106141894903122,
|
| 252 |
-
"learning_rate": 9.937959637795389e-06,
|
| 253 |
-
"loss": 1.0112,
|
| 254 |
-
"step": 170
|
| 255 |
-
},
|
| 256 |
-
{
|
| 257 |
-
"epoch": 0.18,
|
| 258 |
-
"grad_norm": 1.0459501547982482,
|
| 259 |
-
"learning_rate": 9.93349155066901e-06,
|
| 260 |
-
"loss": 0.9959,
|
| 261 |
-
"step": 175
|
| 262 |
-
},
|
| 263 |
-
{
|
| 264 |
-
"epoch": 0.19,
|
| 265 |
-
"grad_norm": 1.1420602608538855,
|
| 266 |
-
"learning_rate": 9.928869190681964e-06,
|
| 267 |
-
"loss": 0.9952,
|
| 268 |
-
"step": 180
|
| 269 |
-
},
|
| 270 |
-
{
|
| 271 |
-
"epoch": 0.2,
|
| 272 |
-
"grad_norm": 1.0748374838181862,
|
| 273 |
-
"learning_rate": 9.924092702377863e-06,
|
| 274 |
-
"loss": 1.0094,
|
| 275 |
-
"step": 185
|
| 276 |
-
},
|
| 277 |
-
{
|
| 278 |
-
"epoch": 0.2,
|
| 279 |
-
"grad_norm": 1.0535011085546289,
|
| 280 |
-
"learning_rate": 9.919162235119996e-06,
|
| 281 |
-
"loss": 1.0054,
|
| 282 |
-
"step": 190
|
| 283 |
-
},
|
| 284 |
-
{
|
| 285 |
-
"epoch": 0.21,
|
| 286 |
-
"grad_norm": 1.0310625793824704,
|
| 287 |
-
"learning_rate": 9.91407794308665e-06,
|
| 288 |
-
"loss": 1.0117,
|
| 289 |
-
"step": 195
|
| 290 |
-
},
|
| 291 |
-
{
|
| 292 |
-
"epoch": 0.21,
|
| 293 |
-
"grad_norm": 1.0359842004906923,
|
| 294 |
-
"learning_rate": 9.908839985266297e-06,
|
| 295 |
-
"loss": 0.9982,
|
| 296 |
-
"step": 200
|
| 297 |
-
},
|
| 298 |
-
{
|
| 299 |
-
"epoch": 0.22,
|
| 300 |
-
"grad_norm": 1.084059570369228,
|
| 301 |
-
"learning_rate": 9.903448525452618e-06,
|
| 302 |
-
"loss": 1.0127,
|
| 303 |
-
"step": 205
|
| 304 |
-
},
|
| 305 |
-
{
|
| 306 |
-
"epoch": 0.22,
|
| 307 |
-
"grad_norm": 1.1227120329409497,
|
| 308 |
-
"learning_rate": 9.89790373223938e-06,
|
| 309 |
-
"loss": 1.048,
|
| 310 |
-
"step": 210
|
| 311 |
-
},
|
| 312 |
-
{
|
| 313 |
-
"epoch": 0.23,
|
| 314 |
-
"grad_norm": 1.0615021518173307,
|
| 315 |
-
"learning_rate": 9.892205779015167e-06,
|
| 316 |
-
"loss": 1.0021,
|
| 317 |
-
"step": 215
|
| 318 |
-
},
|
| 319 |
-
{
|
| 320 |
-
"epoch": 0.23,
|
| 321 |
-
"grad_norm": 1.070318683802529,
|
| 322 |
-
"learning_rate": 9.886354843957953e-06,
|
| 323 |
-
"loss": 1.0043,
|
| 324 |
-
"step": 220
|
| 325 |
-
},
|
| 326 |
-
{
|
| 327 |
-
"epoch": 0.24,
|
| 328 |
-
"grad_norm": 1.0419755132096296,
|
| 329 |
-
"learning_rate": 9.88035111002954e-06,
|
| 330 |
-
"loss": 0.9743,
|
| 331 |
-
"step": 225
|
| 332 |
-
},
|
| 333 |
-
{
|
| 334 |
-
"epoch": 0.24,
|
| 335 |
-
"grad_norm": 1.1796172322040084,
|
| 336 |
-
"learning_rate": 9.874194764969827e-06,
|
| 337 |
-
"loss": 0.9957,
|
| 338 |
-
"step": 230
|
| 339 |
-
},
|
| 340 |
-
{
|
| 341 |
-
"epoch": 0.25,
|
| 342 |
-
"grad_norm": 1.0933963352790785,
|
| 343 |
-
"learning_rate": 9.867886001290943e-06,
|
| 344 |
-
"loss": 0.9814,
|
| 345 |
-
"step": 235
|
| 346 |
-
},
|
| 347 |
-
{
|
| 348 |
-
"epoch": 0.25,
|
| 349 |
-
"grad_norm": 1.1066280030775704,
|
| 350 |
-
"learning_rate": 9.861425016271227e-06,
|
| 351 |
-
"loss": 0.9832,
|
| 352 |
-
"step": 240
|
| 353 |
-
},
|
| 354 |
-
{
|
| 355 |
-
"epoch": 0.26,
|
| 356 |
-
"grad_norm": 1.1138948008724274,
|
| 357 |
-
"learning_rate": 9.854812011949059e-06,
|
| 358 |
-
"loss": 0.9871,
|
| 359 |
-
"step": 245
|
| 360 |
-
},
|
| 361 |
-
{
|
| 362 |
-
"epoch": 0.26,
|
| 363 |
-
"grad_norm": 1.0644401239508805,
|
| 364 |
-
"learning_rate": 9.848047195116543e-06,
|
| 365 |
-
"loss": 0.9951,
|
| 366 |
-
"step": 250
|
| 367 |
-
},
|
| 368 |
-
{
|
| 369 |
-
"epoch": 0.27,
|
| 370 |
-
"grad_norm": 1.1880183474724784,
|
| 371 |
-
"learning_rate": 9.841130777313039e-06,
|
| 372 |
-
"loss": 0.9902,
|
| 373 |
-
"step": 255
|
| 374 |
-
},
|
| 375 |
-
{
|
| 376 |
-
"epoch": 0.27,
|
| 377 |
-
"grad_norm": 1.0747113009717828,
|
| 378 |
-
"learning_rate": 9.834062974818547e-06,
|
| 379 |
-
"loss": 0.9433,
|
| 380 |
-
"step": 260
|
| 381 |
-
},
|
| 382 |
-
{
|
| 383 |
-
"epoch": 0.28,
|
| 384 |
-
"grad_norm": 1.1442114734348945,
|
| 385 |
-
"learning_rate": 9.826844008646949e-06,
|
| 386 |
-
"loss": 0.9703,
|
| 387 |
-
"step": 265
|
| 388 |
-
},
|
| 389 |
-
{
|
| 390 |
-
"epoch": 0.29,
|
| 391 |
-
"grad_norm": 1.0895758630826766,
|
| 392 |
-
"learning_rate": 9.81947410453909e-06,
|
| 393 |
-
"loss": 1.0236,
|
| 394 |
-
"step": 270
|
| 395 |
-
},
|
| 396 |
-
{
|
| 397 |
-
"epoch": 0.29,
|
| 398 |
-
"grad_norm": 0.996075250542336,
|
| 399 |
-
"learning_rate": 9.811953492955728e-06,
|
| 400 |
-
"loss": 0.9577,
|
| 401 |
-
"step": 275
|
| 402 |
-
},
|
| 403 |
-
{
|
| 404 |
-
"epoch": 0.3,
|
| 405 |
-
"grad_norm": 1.1734623195649692,
|
| 406 |
-
"learning_rate": 9.80428240907032e-06,
|
| 407 |
-
"loss": 0.9752,
|
| 408 |
-
"step": 280
|
| 409 |
-
},
|
| 410 |
-
{
|
| 411 |
-
"epoch": 0.3,
|
| 412 |
-
"grad_norm": 1.282701051609298,
|
| 413 |
-
"learning_rate": 9.796461092761668e-06,
|
| 414 |
-
"loss": 0.987,
|
| 415 |
-
"step": 285
|
| 416 |
-
},
|
| 417 |
-
{
|
| 418 |
-
"epoch": 0.31,
|
| 419 |
-
"grad_norm": 1.0721992980205135,
|
| 420 |
-
"learning_rate": 9.788489788606423e-06,
|
| 421 |
-
"loss": 0.944,
|
| 422 |
-
"step": 290
|
| 423 |
-
},
|
| 424 |
-
{
|
| 425 |
-
"epoch": 0.31,
|
| 426 |
-
"grad_norm": 1.105694230535082,
|
| 427 |
-
"learning_rate": 9.780368745871438e-06,
|
| 428 |
-
"loss": 0.9804,
|
| 429 |
-
"step": 295
|
| 430 |
-
},
|
| 431 |
-
{
|
| 432 |
-
"epoch": 0.32,
|
| 433 |
-
"grad_norm": 1.1121587653939105,
|
| 434 |
-
"learning_rate": 9.772098218505963e-06,
|
| 435 |
-
"loss": 1.0099,
|
| 436 |
-
"step": 300
|
| 437 |
-
},
|
| 438 |
-
{
|
| 439 |
-
"epoch": 0.32,
|
| 440 |
-
"grad_norm": 1.1073177873687883,
|
| 441 |
-
"learning_rate": 9.763678465133712e-06,
|
| 442 |
-
"loss": 0.9887,
|
| 443 |
-
"step": 305
|
| 444 |
-
},
|
| 445 |
-
{
|
| 446 |
-
"epoch": 0.33,
|
| 447 |
-
"grad_norm": 1.1986141459298305,
|
| 448 |
-
"learning_rate": 9.755109749044781e-06,
|
| 449 |
-
"loss": 0.9749,
|
| 450 |
-
"step": 310
|
| 451 |
-
},
|
| 452 |
-
{
|
| 453 |
-
"epoch": 0.33,
|
| 454 |
-
"grad_norm": 1.0864391212895972,
|
| 455 |
-
"learning_rate": 9.7463923381874e-06,
|
| 456 |
-
"loss": 0.9767,
|
| 457 |
-
"step": 315
|
| 458 |
-
},
|
| 459 |
-
{
|
| 460 |
-
"epoch": 0.34,
|
| 461 |
-
"grad_norm": 1.0595953209575595,
|
| 462 |
-
"learning_rate": 9.737526505159564e-06,
|
| 463 |
-
"loss": 0.9297,
|
| 464 |
-
"step": 320
|
| 465 |
-
},
|
| 466 |
-
{
|
| 467 |
-
"epoch": 0.34,
|
| 468 |
-
"grad_norm": 1.083224438455533,
|
| 469 |
-
"learning_rate": 9.728512527200509e-06,
|
| 470 |
-
"loss": 0.9498,
|
| 471 |
-
"step": 325
|
| 472 |
-
},
|
| 473 |
-
{
|
| 474 |
-
"epoch": 0.35,
|
| 475 |
-
"grad_norm": 1.1306776282190978,
|
| 476 |
-
"learning_rate": 9.719350686182041e-06,
|
| 477 |
-
"loss": 0.982,
|
| 478 |
-
"step": 330
|
| 479 |
-
},
|
| 480 |
-
{
|
| 481 |
-
"epoch": 0.35,
|
| 482 |
-
"grad_norm": 1.07939319367538,
|
| 483 |
-
"learning_rate": 9.710041268599718e-06,
|
| 484 |
-
"loss": 0.9669,
|
| 485 |
-
"step": 335
|
| 486 |
-
},
|
| 487 |
-
{
|
| 488 |
-
"epoch": 0.36,
|
| 489 |
-
"grad_norm": 1.1100410279851476,
|
| 490 |
-
"learning_rate": 9.700584565563897e-06,
|
| 491 |
-
"loss": 0.956,
|
| 492 |
-
"step": 340
|
| 493 |
-
},
|
| 494 |
-
{
|
| 495 |
-
"epoch": 0.36,
|
| 496 |
-
"grad_norm": 1.0917533373255544,
|
| 497 |
-
"learning_rate": 9.690980872790627e-06,
|
| 498 |
-
"loss": 0.9878,
|
| 499 |
-
"step": 345
|
| 500 |
-
},
|
| 501 |
-
{
|
| 502 |
-
"epoch": 0.37,
|
| 503 |
-
"grad_norm": 1.1287494016251205,
|
| 504 |
-
"learning_rate": 9.681230490592403e-06,
|
| 505 |
-
"loss": 0.9604,
|
| 506 |
-
"step": 350
|
| 507 |
-
},
|
| 508 |
-
{
|
| 509 |
-
"epoch": 0.38,
|
| 510 |
-
"grad_norm": 1.0366025693971206,
|
| 511 |
-
"learning_rate": 9.671333723868773e-06,
|
| 512 |
-
"loss": 0.9809,
|
| 513 |
-
"step": 355
|
| 514 |
-
},
|
| 515 |
-
{
|
| 516 |
-
"epoch": 0.38,
|
| 517 |
-
"grad_norm": 1.1876939558601538,
|
| 518 |
-
"learning_rate": 9.66129088209681e-06,
|
| 519 |
-
"loss": 0.9324,
|
| 520 |
-
"step": 360
|
| 521 |
-
},
|
| 522 |
-
{
|
| 523 |
-
"epoch": 0.39,
|
| 524 |
-
"grad_norm": 1.1296469706806582,
|
| 525 |
-
"learning_rate": 9.651102279321429e-06,
|
| 526 |
-
"loss": 0.98,
|
| 527 |
-
"step": 365
|
| 528 |
-
},
|
| 529 |
-
{
|
| 530 |
-
"epoch": 0.39,
|
| 531 |
-
"grad_norm": 1.0920615981549329,
|
| 532 |
-
"learning_rate": 9.640768234145563e-06,
|
| 533 |
-
"loss": 0.9474,
|
| 534 |
-
"step": 370
|
| 535 |
-
},
|
| 536 |
-
{
|
| 537 |
-
"epoch": 0.4,
|
| 538 |
-
"grad_norm": 1.045353192143218,
|
| 539 |
-
"learning_rate": 9.630289069720213e-06,
|
| 540 |
-
"loss": 0.9416,
|
| 541 |
-
"step": 375
|
| 542 |
-
},
|
| 543 |
-
{
|
| 544 |
-
"epoch": 0.4,
|
| 545 |
-
"grad_norm": 1.0546831730532094,
|
| 546 |
-
"learning_rate": 9.619665113734327e-06,
|
| 547 |
-
"loss": 0.9583,
|
| 548 |
-
"step": 380
|
| 549 |
-
},
|
| 550 |
-
{
|
| 551 |
-
"epoch": 0.41,
|
| 552 |
-
"grad_norm": 1.120397617115956,
|
| 553 |
-
"learning_rate": 9.608896698404567e-06,
|
| 554 |
-
"loss": 0.9739,
|
| 555 |
-
"step": 385
|
| 556 |
-
},
|
| 557 |
-
{
|
| 558 |
-
"epoch": 0.41,
|
| 559 |
-
"grad_norm": 1.0897789727469696,
|
| 560 |
-
"learning_rate": 9.597984160464908e-06,
|
| 561 |
-
"loss": 0.9882,
|
| 562 |
-
"step": 390
|
| 563 |
-
},
|
| 564 |
-
{
|
| 565 |
-
"epoch": 0.42,
|
| 566 |
-
"grad_norm": 1.0655227440534312,
|
| 567 |
-
"learning_rate": 9.586927841156121e-06,
|
| 568 |
-
"loss": 0.973,
|
| 569 |
-
"step": 395
|
| 570 |
-
},
|
| 571 |
-
{
|
| 572 |
-
"epoch": 0.42,
|
| 573 |
-
"grad_norm": 1.024445190271631,
|
| 574 |
-
"learning_rate": 9.575728086215093e-06,
|
| 575 |
-
"loss": 0.9488,
|
| 576 |
-
"step": 400
|
| 577 |
-
},
|
| 578 |
-
{
|
| 579 |
-
"epoch": 0.43,
|
| 580 |
-
"grad_norm": 1.0957551302719917,
|
| 581 |
-
"learning_rate": 9.564385245864015e-06,
|
| 582 |
-
"loss": 0.9395,
|
| 583 |
-
"step": 405
|
| 584 |
-
},
|
| 585 |
-
{
|
| 586 |
-
"epoch": 0.43,
|
| 587 |
-
"grad_norm": 1.0348921383964815,
|
| 588 |
-
"learning_rate": 9.552899674799438e-06,
|
| 589 |
-
"loss": 0.9618,
|
| 590 |
-
"step": 410
|
| 591 |
-
},
|
| 592 |
-
{
|
| 593 |
-
"epoch": 0.44,
|
| 594 |
-
"grad_norm": 1.1320917241343242,
|
| 595 |
-
"learning_rate": 9.541271732181174e-06,
|
| 596 |
-
"loss": 0.9737,
|
| 597 |
-
"step": 415
|
| 598 |
-
},
|
| 599 |
-
{
|
| 600 |
-
"epoch": 0.44,
|
| 601 |
-
"grad_norm": 1.0955620287950987,
|
| 602 |
-
"learning_rate": 9.52950178162107e-06,
|
| 603 |
-
"loss": 0.9765,
|
| 604 |
-
"step": 420
|
| 605 |
-
},
|
| 606 |
-
{
|
| 607 |
-
"epoch": 0.45,
|
| 608 |
-
"grad_norm": 1.0865957472837047,
|
| 609 |
-
"learning_rate": 9.517590191171638e-06,
|
| 610 |
-
"loss": 0.9402,
|
| 611 |
-
"step": 425
|
| 612 |
-
},
|
| 613 |
-
{
|
| 614 |
-
"epoch": 0.45,
|
| 615 |
-
"grad_norm": 1.0608004961340336,
|
| 616 |
-
"learning_rate": 9.505537333314534e-06,
|
| 617 |
-
"loss": 0.938,
|
| 618 |
-
"step": 430
|
| 619 |
-
},
|
| 620 |
-
{
|
| 621 |
-
"epoch": 0.46,
|
| 622 |
-
"grad_norm": 1.0436288259170787,
|
| 623 |
-
"learning_rate": 9.493343584948931e-06,
|
| 624 |
-
"loss": 0.9495,
|
| 625 |
-
"step": 435
|
| 626 |
-
},
|
| 627 |
-
{
|
| 628 |
-
"epoch": 0.46,
|
| 629 |
-
"grad_norm": 1.0827000850655668,
|
| 630 |
-
"learning_rate": 9.481009327379714e-06,
|
| 631 |
-
"loss": 0.9505,
|
| 632 |
-
"step": 440
|
| 633 |
-
},
|
| 634 |
-
{
|
| 635 |
-
"epoch": 0.47,
|
| 636 |
-
"grad_norm": 1.0958366892000795,
|
| 637 |
-
"learning_rate": 9.46853494630557e-06,
|
| 638 |
-
"loss": 0.9536,
|
| 639 |
-
"step": 445
|
| 640 |
-
},
|
| 641 |
-
{
|
| 642 |
-
"epoch": 0.48,
|
| 643 |
-
"grad_norm": 1.0431220913897328,
|
| 644 |
-
"learning_rate": 9.455920831806917e-06,
|
| 645 |
-
"loss": 0.942,
|
| 646 |
-
"step": 450
|
| 647 |
-
},
|
| 648 |
-
{
|
| 649 |
-
"epoch": 0.48,
|
| 650 |
-
"grad_norm": 1.1372655798293543,
|
| 651 |
-
"learning_rate": 9.443167378333711e-06,
|
| 652 |
-
"loss": 0.9447,
|
| 653 |
-
"step": 455
|
| 654 |
-
},
|
| 655 |
-
{
|
| 656 |
-
"epoch": 0.49,
|
| 657 |
-
"grad_norm": 1.0890187843066097,
|
| 658 |
-
"learning_rate": 9.43027498469311e-06,
|
| 659 |
-
"loss": 0.9291,
|
| 660 |
-
"step": 460
|
| 661 |
-
},
|
| 662 |
-
{
|
| 663 |
-
"epoch": 0.49,
|
| 664 |
-
"grad_norm": 1.128255566030822,
|
| 665 |
-
"learning_rate": 9.41724405403701e-06,
|
| 666 |
-
"loss": 0.9418,
|
| 667 |
-
"step": 465
|
| 668 |
-
},
|
| 669 |
-
{
|
| 670 |
-
"epoch": 0.5,
|
| 671 |
-
"grad_norm": 1.0200134644324146,
|
| 672 |
-
"learning_rate": 9.404074993849421e-06,
|
| 673 |
-
"loss": 0.927,
|
| 674 |
-
"step": 470
|
| 675 |
-
},
|
| 676 |
-
{
|
| 677 |
-
"epoch": 0.5,
|
| 678 |
-
"grad_norm": 1.0912622433950008,
|
| 679 |
-
"learning_rate": 9.390768215933746e-06,
|
| 680 |
-
"loss": 0.943,
|
| 681 |
-
"step": 475
|
| 682 |
-
},
|
| 683 |
-
{
|
| 684 |
-
"epoch": 0.51,
|
| 685 |
-
"grad_norm": 1.1784430852167105,
|
| 686 |
-
"learning_rate": 9.377324136399887e-06,
|
| 687 |
-
"loss": 0.9409,
|
| 688 |
-
"step": 480
|
| 689 |
-
},
|
| 690 |
-
{
|
| 691 |
-
"epoch": 0.51,
|
| 692 |
-
"grad_norm": 1.0732445497397998,
|
| 693 |
-
"learning_rate": 9.36374317565124e-06,
|
| 694 |
-
"loss": 0.9401,
|
| 695 |
-
"step": 485
|
| 696 |
-
},
|
| 697 |
-
{
|
| 698 |
-
"epoch": 0.52,
|
| 699 |
-
"grad_norm": 1.1241973380928443,
|
| 700 |
-
"learning_rate": 9.350025758371554e-06,
|
| 701 |
-
"loss": 0.9188,
|
| 702 |
-
"step": 490
|
| 703 |
-
},
|
| 704 |
-
{
|
| 705 |
-
"epoch": 0.52,
|
| 706 |
-
"grad_norm": 1.0680249447424572,
|
| 707 |
-
"learning_rate": 9.336172313511636e-06,
|
| 708 |
-
"loss": 0.9304,
|
| 709 |
-
"step": 495
|
| 710 |
-
},
|
| 711 |
-
{
|
| 712 |
-
"epoch": 0.53,
|
| 713 |
-
"grad_norm": 1.0400938648362148,
|
| 714 |
-
"learning_rate": 9.322183274275954e-06,
|
| 715 |
-
"loss": 0.9465,
|
| 716 |
-
"step": 500
|
| 717 |
-
},
|
| 718 |
-
{
|
| 719 |
-
"epoch": 0.53,
|
| 720 |
-
"grad_norm": 1.1484166178621282,
|
| 721 |
-
"learning_rate": 9.308059078109078e-06,
|
| 722 |
-
"loss": 0.9431,
|
| 723 |
-
"step": 505
|
| 724 |
-
},
|
| 725 |
-
{
|
| 726 |
-
"epoch": 0.54,
|
| 727 |
-
"grad_norm": 1.0928763685485705,
|
| 728 |
-
"learning_rate": 9.29380016668201e-06,
|
| 729 |
-
"loss": 0.9368,
|
| 730 |
-
"step": 510
|
| 731 |
-
},
|
| 732 |
-
{
|
| 733 |
-
"epoch": 0.54,
|
| 734 |
-
"grad_norm": 1.0470334802413224,
|
| 735 |
-
"learning_rate": 9.279406985878367e-06,
|
| 736 |
-
"loss": 0.9529,
|
| 737 |
-
"step": 515
|
| 738 |
-
},
|
| 739 |
-
{
|
| 740 |
-
"epoch": 0.55,
|
| 741 |
-
"grad_norm": 1.055693577627048,
|
| 742 |
-
"learning_rate": 9.264879985780436e-06,
|
| 743 |
-
"loss": 0.9237,
|
| 744 |
-
"step": 520
|
| 745 |
-
},
|
| 746 |
-
{
|
| 747 |
-
"epoch": 0.55,
|
| 748 |
-
"grad_norm": 1.0582407523485609,
|
| 749 |
-
"learning_rate": 9.250219620655112e-06,
|
| 750 |
-
"loss": 0.9455,
|
| 751 |
-
"step": 525
|
| 752 |
-
},
|
| 753 |
-
{
|
| 754 |
-
"epoch": 0.56,
|
| 755 |
-
"grad_norm": 1.0392740863841614,
|
| 756 |
-
"learning_rate": 9.235426348939674e-06,
|
| 757 |
-
"loss": 0.9866,
|
| 758 |
-
"step": 530
|
| 759 |
-
},
|
| 760 |
-
{
|
| 761 |
-
"epoch": 0.57,
|
| 762 |
-
"grad_norm": 1.087021743413759,
|
| 763 |
-
"learning_rate": 9.220500633227467e-06,
|
| 764 |
-
"loss": 0.9797,
|
| 765 |
-
"step": 535
|
| 766 |
-
},
|
| 767 |
-
{
|
| 768 |
-
"epoch": 0.57,
|
| 769 |
-
"grad_norm": 1.0905659766649087,
|
| 770 |
-
"learning_rate": 9.205442940253426e-06,
|
| 771 |
-
"loss": 0.9231,
|
| 772 |
-
"step": 540
|
| 773 |
-
},
|
| 774 |
-
{
|
| 775 |
-
"epoch": 0.58,
|
| 776 |
-
"grad_norm": 1.0838061353931883,
|
| 777 |
-
"learning_rate": 9.190253740879484e-06,
|
| 778 |
-
"loss": 0.9155,
|
| 779 |
-
"step": 545
|
| 780 |
-
},
|
| 781 |
-
{
|
| 782 |
-
"epoch": 0.58,
|
| 783 |
-
"grad_norm": 1.1721559515157844,
|
| 784 |
-
"learning_rate": 9.174933510079847e-06,
|
| 785 |
-
"loss": 0.9132,
|
| 786 |
-
"step": 550
|
| 787 |
-
},
|
| 788 |
-
{
|
| 789 |
-
"epoch": 0.59,
|
| 790 |
-
"grad_norm": 1.0711291424853389,
|
| 791 |
-
"learning_rate": 9.159482726926147e-06,
|
| 792 |
-
"loss": 0.9368,
|
| 793 |
-
"step": 555
|
| 794 |
-
},
|
| 795 |
-
{
|
| 796 |
-
"epoch": 0.59,
|
| 797 |
-
"grad_norm": 1.0906836737125443,
|
| 798 |
-
"learning_rate": 9.14390187457245e-06,
|
| 799 |
-
"loss": 0.9652,
|
| 800 |
-
"step": 560
|
| 801 |
-
},
|
| 802 |
-
{
|
| 803 |
-
"epoch": 0.6,
|
| 804 |
-
"grad_norm": 1.2147816750505283,
|
| 805 |
-
"learning_rate": 9.128191440240159e-06,
|
| 806 |
-
"loss": 0.922,
|
| 807 |
-
"step": 565
|
| 808 |
-
},
|
| 809 |
-
{
|
| 810 |
-
"epoch": 0.6,
|
| 811 |
-
"grad_norm": 1.0745698856829782,
|
| 812 |
-
"learning_rate": 9.11235191520277e-06,
|
| 813 |
-
"loss": 0.9267,
|
| 814 |
-
"step": 570
|
| 815 |
-
},
|
| 816 |
-
{
|
| 817 |
-
"epoch": 0.61,
|
| 818 |
-
"grad_norm": 1.1107563079565528,
|
| 819 |
-
"learning_rate": 9.096383794770513e-06,
|
| 820 |
-
"loss": 0.9403,
|
| 821 |
-
"step": 575
|
| 822 |
-
},
|
| 823 |
-
{
|
| 824 |
-
"epoch": 0.61,
|
| 825 |
-
"grad_norm": 1.0645734678937102,
|
| 826 |
-
"learning_rate": 9.080287578274866e-06,
|
| 827 |
-
"loss": 0.9149,
|
| 828 |
-
"step": 580
|
| 829 |
-
},
|
| 830 |
-
{
|
| 831 |
-
"epoch": 0.62,
|
| 832 |
-
"grad_norm": 1.1729380707889032,
|
| 833 |
-
"learning_rate": 9.064063769052933e-06,
|
| 834 |
-
"loss": 0.9236,
|
| 835 |
-
"step": 585
|
| 836 |
-
},
|
| 837 |
-
{
|
| 838 |
-
"epoch": 0.62,
|
| 839 |
-
"grad_norm": 1.0634029251400858,
|
| 840 |
-
"learning_rate": 9.047712874431716e-06,
|
| 841 |
-
"loss": 0.9264,
|
| 842 |
-
"step": 590
|
| 843 |
-
},
|
| 844 |
-
{
|
| 845 |
-
"epoch": 0.63,
|
| 846 |
-
"grad_norm": 1.185148731024843,
|
| 847 |
-
"learning_rate": 9.031235405712239e-06,
|
| 848 |
-
"loss": 0.9632,
|
| 849 |
-
"step": 595
|
| 850 |
-
},
|
| 851 |
-
{
|
| 852 |
-
"epoch": 0.63,
|
| 853 |
-
"grad_norm": 1.1238661801404854,
|
| 854 |
-
"learning_rate": 9.014631878153564e-06,
|
| 855 |
-
"loss": 0.9364,
|
| 856 |
-
"step": 600
|
| 857 |
-
},
|
| 858 |
-
{
|
| 859 |
-
"epoch": 0.64,
|
| 860 |
-
"grad_norm": 1.1101591200426506,
|
| 861 |
-
"learning_rate": 8.997902810956682e-06,
|
| 862 |
-
"loss": 0.9121,
|
| 863 |
-
"step": 605
|
| 864 |
-
},
|
| 865 |
-
{
|
| 866 |
-
"epoch": 0.64,
|
| 867 |
-
"grad_norm": 1.1328306862765927,
|
| 868 |
-
"learning_rate": 8.98104872724827e-06,
|
| 869 |
-
"loss": 0.9637,
|
| 870 |
-
"step": 610
|
| 871 |
-
},
|
| 872 |
-
{
|
| 873 |
-
"epoch": 0.65,
|
| 874 |
-
"grad_norm": 1.1182389860600772,
|
| 875 |
-
"learning_rate": 8.964070154064343e-06,
|
| 876 |
-
"loss": 0.9431,
|
| 877 |
-
"step": 615
|
| 878 |
-
},
|
| 879 |
-
{
|
| 880 |
-
"epoch": 0.66,
|
| 881 |
-
"grad_norm": 1.2315329373588069,
|
| 882 |
-
"learning_rate": 8.94696762233376e-06,
|
| 883 |
-
"loss": 0.9261,
|
| 884 |
-
"step": 620
|
| 885 |
-
},
|
| 886 |
-
{
|
| 887 |
-
"epoch": 0.66,
|
| 888 |
-
"grad_norm": 1.0785263989248792,
|
| 889 |
-
"learning_rate": 8.92974166686163e-06,
|
| 890 |
-
"loss": 0.9218,
|
| 891 |
-
"step": 625
|
| 892 |
-
},
|
| 893 |
-
{
|
| 894 |
-
"epoch": 0.67,
|
| 895 |
-
"grad_norm": 1.0293877329539916,
|
| 896 |
-
"learning_rate": 8.912392826312595e-06,
|
| 897 |
-
"loss": 0.9516,
|
| 898 |
-
"step": 630
|
| 899 |
-
},
|
| 900 |
-
{
|
| 901 |
-
"epoch": 0.67,
|
| 902 |
-
"grad_norm": 1.0797961930582287,
|
| 903 |
-
"learning_rate": 8.894921643193966e-06,
|
| 904 |
-
"loss": 0.94,
|
| 905 |
-
"step": 635
|
| 906 |
-
},
|
| 907 |
-
{
|
| 908 |
-
"epoch": 0.68,
|
| 909 |
-
"grad_norm": 1.0052477432214972,
|
| 910 |
-
"learning_rate": 8.877328663838776e-06,
|
| 911 |
-
"loss": 0.9207,
|
| 912 |
-
"step": 640
|
| 913 |
-
},
|
| 914 |
-
{
|
| 915 |
-
"epoch": 0.68,
|
| 916 |
-
"grad_norm": 1.0126272743426095,
|
| 917 |
-
"learning_rate": 8.85961443838869e-06,
|
| 918 |
-
"loss": 0.9292,
|
| 919 |
-
"step": 645
|
| 920 |
-
},
|
| 921 |
-
{
|
| 922 |
-
"epoch": 0.69,
|
| 923 |
-
"grad_norm": 1.0166858946265631,
|
| 924 |
-
"learning_rate": 8.841779520776803e-06,
|
| 925 |
-
"loss": 0.9171,
|
| 926 |
-
"step": 650
|
| 927 |
-
},
|
| 928 |
-
{
|
| 929 |
-
"epoch": 0.69,
|
| 930 |
-
"grad_norm": 1.0674058891203713,
|
| 931 |
-
"learning_rate": 8.823824468710312e-06,
|
| 932 |
-
"loss": 0.9238,
|
| 933 |
-
"step": 655
|
| 934 |
-
},
|
| 935 |
-
{
|
| 936 |
-
"epoch": 0.7,
|
| 937 |
-
"grad_norm": 1.0826543746678357,
|
| 938 |
-
"learning_rate": 8.805749843653086e-06,
|
| 939 |
-
"loss": 0.8903,
|
| 940 |
-
"step": 660
|
| 941 |
-
},
|
| 942 |
-
{
|
| 943 |
-
"epoch": 0.7,
|
| 944 |
-
"grad_norm": 1.0474293060948185,
|
| 945 |
-
"learning_rate": 8.787556210808101e-06,
|
| 946 |
-
"loss": 0.8952,
|
| 947 |
-
"step": 665
|
| 948 |
-
},
|
| 949 |
-
{
|
| 950 |
-
"epoch": 0.71,
|
| 951 |
-
"grad_norm": 1.1092322508696293,
|
| 952 |
-
"learning_rate": 8.769244139099774e-06,
|
| 953 |
-
"loss": 0.9191,
|
| 954 |
-
"step": 670
|
| 955 |
-
},
|
| 956 |
-
{
|
| 957 |
-
"epoch": 0.71,
|
| 958 |
-
"grad_norm": 1.0453618423472522,
|
| 959 |
-
"learning_rate": 8.750814201156157e-06,
|
| 960 |
-
"loss": 0.9287,
|
| 961 |
-
"step": 675
|
| 962 |
-
},
|
| 963 |
-
{
|
| 964 |
-
"epoch": 0.72,
|
| 965 |
-
"grad_norm": 1.0150902528617922,
|
| 966 |
-
"learning_rate": 8.732266973291053e-06,
|
| 967 |
-
"loss": 0.9005,
|
| 968 |
-
"step": 680
|
| 969 |
-
},
|
| 970 |
-
{
|
| 971 |
-
"epoch": 0.72,
|
| 972 |
-
"grad_norm": 1.111573072134849,
|
| 973 |
-
"learning_rate": 8.713603035485972e-06,
|
| 974 |
-
"loss": 0.9061,
|
| 975 |
-
"step": 685
|
| 976 |
-
},
|
| 977 |
-
{
|
| 978 |
-
"epoch": 0.73,
|
| 979 |
-
"grad_norm": 1.0266552996471214,
|
| 980 |
-
"learning_rate": 8.694822971372012e-06,
|
| 981 |
-
"loss": 0.8981,
|
| 982 |
-
"step": 690
|
| 983 |
-
},
|
| 984 |
-
{
|
| 985 |
-
"epoch": 0.73,
|
| 986 |
-
"grad_norm": 1.026959416886306,
|
| 987 |
-
"learning_rate": 8.675927368211599e-06,
|
| 988 |
-
"loss": 0.9119,
|
| 989 |
-
"step": 695
|
| 990 |
-
},
|
| 991 |
-
{
|
| 992 |
-
"epoch": 0.74,
|
| 993 |
-
"grad_norm": 0.990879098356618,
|
| 994 |
-
"learning_rate": 8.656916816880122e-06,
|
| 995 |
-
"loss": 0.934,
|
| 996 |
-
"step": 700
|
| 997 |
-
},
|
| 998 |
-
{
|
| 999 |
-
"epoch": 0.75,
|
| 1000 |
-
"grad_norm": 1.016936193517629,
|
| 1001 |
-
"learning_rate": 8.637791911847462e-06,
|
| 1002 |
-
"loss": 0.9031,
|
| 1003 |
-
"step": 705
|
| 1004 |
-
},
|
| 1005 |
-
{
|
| 1006 |
-
"epoch": 0.75,
|
| 1007 |
-
"grad_norm": 1.0105346034407392,
|
| 1008 |
-
"learning_rate": 8.618553251159405e-06,
|
| 1009 |
-
"loss": 0.8918,
|
| 1010 |
-
"step": 710
|
| 1011 |
-
},
|
| 1012 |
-
{
|
| 1013 |
-
"epoch": 0.76,
|
| 1014 |
-
"grad_norm": 1.0219526658502593,
|
| 1015 |
-
"learning_rate": 8.599201436418927e-06,
|
| 1016 |
-
"loss": 0.9202,
|
| 1017 |
-
"step": 715
|
| 1018 |
-
},
|
| 1019 |
-
{
|
| 1020 |
-
"epoch": 0.76,
|
| 1021 |
-
"grad_norm": 1.0611008297726183,
|
| 1022 |
-
"learning_rate": 8.579737072767396e-06,
|
| 1023 |
-
"loss": 0.8956,
|
| 1024 |
-
"step": 720
|
| 1025 |
-
},
|
| 1026 |
-
{
|
| 1027 |
-
"epoch": 0.77,
|
| 1028 |
-
"grad_norm": 1.0532525094762688,
|
| 1029 |
-
"learning_rate": 8.560160768865642e-06,
|
| 1030 |
-
"loss": 0.8782,
|
| 1031 |
-
"step": 725
|
| 1032 |
-
},
|
| 1033 |
-
{
|
| 1034 |
-
"epoch": 0.77,
|
| 1035 |
-
"grad_norm": 1.0472370063073,
|
| 1036 |
-
"learning_rate": 8.540473136874926e-06,
|
| 1037 |
-
"loss": 0.9215,
|
| 1038 |
-
"step": 730
|
| 1039 |
-
},
|
| 1040 |
-
{
|
| 1041 |
-
"epoch": 0.78,
|
| 1042 |
-
"grad_norm": 1.0503901600633805,
|
| 1043 |
-
"learning_rate": 8.520674792437793e-06,
|
| 1044 |
-
"loss": 0.905,
|
| 1045 |
-
"step": 735
|
| 1046 |
-
},
|
| 1047 |
-
{
|
| 1048 |
-
"epoch": 0.78,
|
| 1049 |
-
"grad_norm": 1.0699401745712223,
|
| 1050 |
-
"learning_rate": 8.50076635465883e-06,
|
| 1051 |
-
"loss": 0.8914,
|
| 1052 |
-
"step": 740
|
| 1053 |
-
},
|
| 1054 |
-
{
|
| 1055 |
-
"epoch": 0.79,
|
| 1056 |
-
"grad_norm": 1.1604934245734189,
|
| 1057 |
-
"learning_rate": 8.480748446085293e-06,
|
| 1058 |
-
"loss": 0.923,
|
| 1059 |
-
"step": 745
|
| 1060 |
-
},
|
| 1061 |
-
{
|
| 1062 |
-
"epoch": 0.79,
|
| 1063 |
-
"grad_norm": 1.0575469862405844,
|
| 1064 |
-
"learning_rate": 8.460621692687656e-06,
|
| 1065 |
-
"loss": 0.91,
|
| 1066 |
-
"step": 750
|
| 1067 |
-
},
|
| 1068 |
-
{
|
| 1069 |
-
"epoch": 0.8,
|
| 1070 |
-
"grad_norm": 1.1861862918344839,
|
| 1071 |
-
"learning_rate": 8.44038672384002e-06,
|
| 1072 |
-
"loss": 0.9183,
|
| 1073 |
-
"step": 755
|
| 1074 |
-
},
|
| 1075 |
-
{
|
| 1076 |
-
"epoch": 0.8,
|
| 1077 |
-
"grad_norm": 1.0866238920331526,
|
| 1078 |
-
"learning_rate": 8.420044172300443e-06,
|
| 1079 |
-
"loss": 0.9012,
|
| 1080 |
-
"step": 760
|
| 1081 |
-
},
|
| 1082 |
-
{
|
| 1083 |
-
"epoch": 0.81,
|
| 1084 |
-
"grad_norm": 1.0963030089254635,
|
| 1085 |
-
"learning_rate": 8.399594674191147e-06,
|
| 1086 |
-
"loss": 0.8867,
|
| 1087 |
-
"step": 765
|
| 1088 |
-
},
|
| 1089 |
-
{
|
| 1090 |
-
"epoch": 0.81,
|
| 1091 |
-
"grad_norm": 1.0516263694748806,
|
| 1092 |
-
"learning_rate": 8.379038868978635e-06,
|
| 1093 |
-
"loss": 0.9204,
|
| 1094 |
-
"step": 770
|
| 1095 |
-
},
|
| 1096 |
-
{
|
| 1097 |
-
"epoch": 0.82,
|
| 1098 |
-
"grad_norm": 1.0602404388082067,
|
| 1099 |
-
"learning_rate": 8.358377399453684e-06,
|
| 1100 |
-
"loss": 0.8975,
|
| 1101 |
-
"step": 775
|
| 1102 |
-
},
|
| 1103 |
-
{
|
| 1104 |
-
"epoch": 0.82,
|
| 1105 |
-
"grad_norm": 1.0524212623827451,
|
| 1106 |
-
"learning_rate": 8.337610911711248e-06,
|
| 1107 |
-
"loss": 0.9182,
|
| 1108 |
-
"step": 780
|
| 1109 |
-
},
|
| 1110 |
-
{
|
| 1111 |
-
"epoch": 0.83,
|
| 1112 |
-
"grad_norm": 1.0486851629524967,
|
| 1113 |
-
"learning_rate": 8.316740055130263e-06,
|
| 1114 |
-
"loss": 0.8996,
|
| 1115 |
-
"step": 785
|
| 1116 |
-
},
|
| 1117 |
-
{
|
| 1118 |
-
"epoch": 0.83,
|
| 1119 |
-
"grad_norm": 1.0382393662171674,
|
| 1120 |
-
"learning_rate": 8.295765482353326e-06,
|
| 1121 |
-
"loss": 0.8898,
|
| 1122 |
-
"step": 790
|
| 1123 |
-
},
|
| 1124 |
-
{
|
| 1125 |
-
"epoch": 0.84,
|
| 1126 |
-
"grad_norm": 1.0801053233779676,
|
| 1127 |
-
"learning_rate": 8.274687849266295e-06,
|
| 1128 |
-
"loss": 0.8942,
|
| 1129 |
-
"step": 795
|
| 1130 |
-
},
|
| 1131 |
-
{
|
| 1132 |
-
"epoch": 0.85,
|
| 1133 |
-
"grad_norm": 1.082914632918619,
|
| 1134 |
-
"learning_rate": 8.253507814977779e-06,
|
| 1135 |
-
"loss": 0.9335,
|
| 1136 |
-
"step": 800
|
| 1137 |
-
},
|
| 1138 |
-
{
|
| 1139 |
-
"epoch": 0.85,
|
| 1140 |
-
"grad_norm": 1.115797305584172,
|
| 1141 |
-
"learning_rate": 8.232226041798528e-06,
|
| 1142 |
-
"loss": 0.8733,
|
| 1143 |
-
"step": 805
|
| 1144 |
-
},
|
| 1145 |
-
{
|
| 1146 |
-
"epoch": 0.86,
|
| 1147 |
-
"grad_norm": 1.0758274816242523,
|
| 1148 |
-
"learning_rate": 8.210843195220717e-06,
|
| 1149 |
-
"loss": 0.9121,
|
| 1150 |
-
"step": 810
|
| 1151 |
-
},
|
| 1152 |
-
{
|
| 1153 |
-
"epoch": 0.86,
|
| 1154 |
-
"grad_norm": 0.9966437564306923,
|
| 1155 |
-
"learning_rate": 8.189359943897137e-06,
|
| 1156 |
-
"loss": 0.9126,
|
| 1157 |
-
"step": 815
|
| 1158 |
-
},
|
| 1159 |
-
{
|
| 1160 |
-
"epoch": 0.87,
|
| 1161 |
-
"grad_norm": 1.1254388184304862,
|
| 1162 |
-
"learning_rate": 8.167776959620298e-06,
|
| 1163 |
-
"loss": 0.9113,
|
| 1164 |
-
"step": 820
|
| 1165 |
-
},
|
| 1166 |
-
{
|
| 1167 |
-
"epoch": 0.87,
|
| 1168 |
-
"grad_norm": 1.033615919920944,
|
| 1169 |
-
"learning_rate": 8.1460949173014e-06,
|
| 1170 |
-
"loss": 0.8863,
|
| 1171 |
-
"step": 825
|
| 1172 |
-
},
|
| 1173 |
-
{
|
| 1174 |
-
"epoch": 0.88,
|
| 1175 |
-
"grad_norm": 1.0126421627367477,
|
| 1176 |
-
"learning_rate": 8.124314494949247e-06,
|
| 1177 |
-
"loss": 0.9044,
|
| 1178 |
-
"step": 830
|
| 1179 |
-
},
|
| 1180 |
-
{
|
| 1181 |
-
"epoch": 0.88,
|
| 1182 |
-
"grad_norm": 1.0545539629522227,
|
| 1183 |
-
"learning_rate": 8.102436373649029e-06,
|
| 1184 |
-
"loss": 0.8942,
|
| 1185 |
-
"step": 835
|
| 1186 |
-
},
|
| 1187 |
-
{
|
| 1188 |
-
"epoch": 0.89,
|
| 1189 |
-
"grad_norm": 1.004956283976033,
|
| 1190 |
-
"learning_rate": 8.080461237541049e-06,
|
| 1191 |
-
"loss": 0.9255,
|
| 1192 |
-
"step": 840
|
| 1193 |
-
},
|
| 1194 |
-
{
|
| 1195 |
-
"epoch": 0.89,
|
| 1196 |
-
"grad_norm": 1.0862660155528163,
|
| 1197 |
-
"learning_rate": 8.0583897737993e-06,
|
| 1198 |
-
"loss": 0.9275,
|
| 1199 |
-
"step": 845
|
| 1200 |
-
},
|
| 1201 |
-
{
|
| 1202 |
-
"epoch": 0.9,
|
| 1203 |
-
"grad_norm": 1.0697124134441602,
|
| 1204 |
-
"learning_rate": 8.036222672609994e-06,
|
| 1205 |
-
"loss": 0.9161,
|
| 1206 |
-
"step": 850
|
| 1207 |
-
},
|
| 1208 |
-
{
|
| 1209 |
-
"epoch": 0.9,
|
| 1210 |
-
"grad_norm": 1.0639070724236763,
|
| 1211 |
-
"learning_rate": 8.013960627149981e-06,
|
| 1212 |
-
"loss": 0.8874,
|
| 1213 |
-
"step": 855
|
| 1214 |
-
},
|
| 1215 |
-
{
|
| 1216 |
-
"epoch": 0.91,
|
| 1217 |
-
"grad_norm": 1.166900094582672,
|
| 1218 |
-
"learning_rate": 7.991604333565062e-06,
|
| 1219 |
-
"loss": 0.8897,
|
| 1220 |
-
"step": 860
|
| 1221 |
-
},
|
| 1222 |
-
{
|
| 1223 |
-
"epoch": 0.91,
|
| 1224 |
-
"grad_norm": 1.1335592965754175,
|
| 1225 |
-
"learning_rate": 7.969154490948225e-06,
|
| 1226 |
-
"loss": 0.8964,
|
| 1227 |
-
"step": 865
|
| 1228 |
-
},
|
| 1229 |
-
{
|
| 1230 |
-
"epoch": 0.92,
|
| 1231 |
-
"grad_norm": 1.0520381511921073,
|
| 1232 |
-
"learning_rate": 7.946611801317794e-06,
|
| 1233 |
-
"loss": 0.8736,
|
| 1234 |
-
"step": 870
|
| 1235 |
-
},
|
| 1236 |
-
{
|
| 1237 |
-
"epoch": 0.92,
|
| 1238 |
-
"grad_norm": 1.16753848747216,
|
| 1239 |
-
"learning_rate": 7.923976969595459e-06,
|
| 1240 |
-
"loss": 0.9112,
|
| 1241 |
-
"step": 875
|
| 1242 |
-
},
|
| 1243 |
-
{
|
| 1244 |
-
"epoch": 0.93,
|
| 1245 |
-
"grad_norm": 1.0772133099773151,
|
| 1246 |
-
"learning_rate": 7.901250703584245e-06,
|
| 1247 |
-
"loss": 0.9155,
|
| 1248 |
-
"step": 880
|
| 1249 |
-
},
|
| 1250 |
-
{
|
| 1251 |
-
"epoch": 0.94,
|
| 1252 |
-
"grad_norm": 1.1464686627860388,
|
| 1253 |
-
"learning_rate": 7.878433713946373e-06,
|
| 1254 |
-
"loss": 0.8962,
|
| 1255 |
-
"step": 885
|
| 1256 |
-
},
|
| 1257 |
-
{
|
| 1258 |
-
"epoch": 0.94,
|
| 1259 |
-
"grad_norm": 1.0835779136854178,
|
| 1260 |
-
"learning_rate": 7.855526714181041e-06,
|
| 1261 |
-
"loss": 0.9058,
|
| 1262 |
-
"step": 890
|
| 1263 |
-
},
|
| 1264 |
-
{
|
| 1265 |
-
"epoch": 0.95,
|
| 1266 |
-
"grad_norm": 1.171366478493349,
|
| 1267 |
-
"learning_rate": 7.832530420602113e-06,
|
| 1268 |
-
"loss": 0.8756,
|
| 1269 |
-
"step": 895
|
| 1270 |
-
},
|
| 1271 |
-
{
|
| 1272 |
-
"epoch": 0.95,
|
| 1273 |
-
"grad_norm": 1.040168900901505,
|
| 1274 |
-
"learning_rate": 7.809445552315714e-06,
|
| 1275 |
-
"loss": 0.8594,
|
| 1276 |
-
"step": 900
|
| 1277 |
-
},
|
| 1278 |
-
{
|
| 1279 |
-
"epoch": 0.96,
|
| 1280 |
-
"grad_norm": 1.02166560480321,
|
| 1281 |
-
"learning_rate": 7.786272831197745e-06,
|
| 1282 |
-
"loss": 0.8935,
|
| 1283 |
-
"step": 905
|
| 1284 |
-
},
|
| 1285 |
-
{
|
| 1286 |
-
"epoch": 0.96,
|
| 1287 |
-
"grad_norm": 1.1107392454183416,
|
| 1288 |
-
"learning_rate": 7.763012981871314e-06,
|
| 1289 |
-
"loss": 0.904,
|
| 1290 |
-
"step": 910
|
| 1291 |
-
},
|
| 1292 |
-
{
|
| 1293 |
-
"epoch": 0.97,
|
| 1294 |
-
"grad_norm": 0.9896358057101541,
|
| 1295 |
-
"learning_rate": 7.739666731684073e-06,
|
| 1296 |
-
"loss": 0.9068,
|
| 1297 |
-
"step": 915
|
| 1298 |
-
},
|
| 1299 |
-
{
|
| 1300 |
-
"epoch": 0.97,
|
| 1301 |
-
"grad_norm": 0.9788741930391702,
|
| 1302 |
-
"learning_rate": 7.716234810685476e-06,
|
| 1303 |
-
"loss": 0.8846,
|
| 1304 |
-
"step": 920
|
| 1305 |
-
},
|
| 1306 |
-
{
|
| 1307 |
-
"epoch": 0.98,
|
| 1308 |
-
"grad_norm": 0.9931045191442167,
|
| 1309 |
-
"learning_rate": 7.692717951603942e-06,
|
| 1310 |
-
"loss": 0.8584,
|
| 1311 |
-
"step": 925
|
| 1312 |
-
},
|
| 1313 |
-
{
|
| 1314 |
-
"epoch": 0.98,
|
| 1315 |
-
"grad_norm": 1.0645481368236074,
|
| 1316 |
-
"learning_rate": 7.669116889823955e-06,
|
| 1317 |
-
"loss": 0.8992,
|
| 1318 |
-
"step": 930
|
| 1319 |
-
},
|
| 1320 |
-
{
|
| 1321 |
-
"epoch": 0.99,
|
| 1322 |
-
"grad_norm": 0.9816731950451545,
|
| 1323 |
-
"learning_rate": 7.645432363363057e-06,
|
| 1324 |
-
"loss": 0.8851,
|
| 1325 |
-
"step": 935
|
| 1326 |
-
},
|
| 1327 |
-
{
|
| 1328 |
-
"epoch": 0.99,
|
| 1329 |
-
"grad_norm": 0.9899142833993008,
|
| 1330 |
-
"learning_rate": 7.621665112848776e-06,
|
| 1331 |
-
"loss": 0.8845,
|
| 1332 |
-
"step": 940
|
| 1333 |
-
},
|
| 1334 |
-
{
|
| 1335 |
-
"epoch": 1.0,
|
| 1336 |
-
"grad_norm": 1.0638888300871174,
|
| 1337 |
-
"learning_rate": 7.597815881495465e-06,
|
| 1338 |
-
"loss": 0.8773,
|
| 1339 |
-
"step": 945
|
| 1340 |
-
},
|
| 1341 |
-
{
|
| 1342 |
-
"epoch": 1.0,
|
| 1343 |
-
"grad_norm": 1.031662431521578,
|
| 1344 |
-
"learning_rate": 7.573885415081059e-06,
|
| 1345 |
-
"loss": 0.8258,
|
| 1346 |
-
"step": 950
|
| 1347 |
-
},
|
| 1348 |
-
{
|
| 1349 |
-
"epoch": 1.01,
|
| 1350 |
-
"grad_norm": 1.040426497974828,
|
| 1351 |
-
"learning_rate": 7.54987446192376e-06,
|
| 1352 |
-
"loss": 0.7907,
|
| 1353 |
-
"step": 955
|
| 1354 |
-
},
|
| 1355 |
-
{
|
| 1356 |
-
"epoch": 1.01,
|
| 1357 |
-
"grad_norm": 0.9887566903005512,
|
| 1358 |
-
"learning_rate": 7.525783772858624e-06,
|
| 1359 |
-
"loss": 0.8091,
|
| 1360 |
-
"step": 960
|
| 1361 |
-
},
|
| 1362 |
-
{
|
| 1363 |
-
"epoch": 1.02,
|
| 1364 |
-
"grad_norm": 1.0542179478307365,
|
| 1365 |
-
"learning_rate": 7.5016141012141e-06,
|
| 1366 |
-
"loss": 0.7815,
|
| 1367 |
-
"step": 965
|
| 1368 |
-
},
|
| 1369 |
-
{
|
| 1370 |
-
"epoch": 1.03,
|
| 1371 |
-
"grad_norm": 1.0738731959256824,
|
| 1372 |
-
"learning_rate": 7.477366202788456e-06,
|
| 1373 |
-
"loss": 0.7734,
|
| 1374 |
-
"step": 970
|
| 1375 |
-
},
|
| 1376 |
-
{
|
| 1377 |
-
"epoch": 1.03,
|
| 1378 |
-
"grad_norm": 0.9975806760235982,
|
| 1379 |
-
"learning_rate": 7.45304083582616e-06,
|
| 1380 |
-
"loss": 0.7824,
|
| 1381 |
-
"step": 975
|
| 1382 |
-
},
|
| 1383 |
-
{
|
| 1384 |
-
"epoch": 1.04,
|
| 1385 |
-
"grad_norm": 1.005274019925314,
|
| 1386 |
-
"learning_rate": 7.4286387609941544e-06,
|
| 1387 |
-
"loss": 0.769,
|
| 1388 |
-
"step": 980
|
| 1389 |
-
},
|
| 1390 |
-
{
|
| 1391 |
-
"epoch": 1.04,
|
| 1392 |
-
"grad_norm": 1.0937329481520819,
|
| 1393 |
-
"learning_rate": 7.40416074135808e-06,
|
| 1394 |
-
"loss": 0.791,
|
| 1395 |
-
"step": 985
|
| 1396 |
-
},
|
| 1397 |
-
{
|
| 1398 |
-
"epoch": 1.05,
|
| 1399 |
-
"grad_norm": 0.9987999174071854,
|
| 1400 |
-
"learning_rate": 7.379607542358414e-06,
|
| 1401 |
-
"loss": 0.7983,
|
| 1402 |
-
"step": 990
|
| 1403 |
-
},
|
| 1404 |
-
{
|
| 1405 |
-
"epoch": 1.05,
|
| 1406 |
-
"grad_norm": 1.074721973505265,
|
| 1407 |
-
"learning_rate": 7.3549799317865235e-06,
|
| 1408 |
-
"loss": 0.8264,
|
| 1409 |
-
"step": 995
|
| 1410 |
-
},
|
| 1411 |
-
{
|
| 1412 |
-
"epoch": 1.06,
|
| 1413 |
-
"grad_norm": 1.0023766389640552,
|
| 1414 |
-
"learning_rate": 7.330278679760673e-06,
|
| 1415 |
-
"loss": 0.8166,
|
| 1416 |
-
"step": 1000
|
| 1417 |
-
},
|
| 1418 |
-
{
|
| 1419 |
-
"epoch": 1.06,
|
| 1420 |
-
"grad_norm": 1.0263488491446793,
|
| 1421 |
-
"learning_rate": 7.3055045587019315e-06,
|
| 1422 |
-
"loss": 0.7756,
|
| 1423 |
-
"step": 1005
|
| 1424 |
-
},
|
| 1425 |
-
{
|
| 1426 |
-
"epoch": 1.07,
|
| 1427 |
-
"grad_norm": 1.222252310199244,
|
| 1428 |
-
"learning_rate": 7.280658343310016e-06,
|
| 1429 |
-
"loss": 0.8113,
|
| 1430 |
-
"step": 1010
|
| 1431 |
-
},
|
| 1432 |
-
{
|
| 1433 |
-
"epoch": 1.07,
|
| 1434 |
-
"grad_norm": 1.0803171037496995,
|
| 1435 |
-
"learning_rate": 7.255740810539078e-06,
|
| 1436 |
-
"loss": 0.7773,
|
| 1437 |
-
"step": 1015
|
| 1438 |
-
},
|
| 1439 |
-
{
|
| 1440 |
-
"epoch": 1.08,
|
| 1441 |
-
"grad_norm": 1.0429385720996782,
|
| 1442 |
-
"learning_rate": 7.230752739573398e-06,
|
| 1443 |
-
"loss": 0.7959,
|
| 1444 |
-
"step": 1020
|
| 1445 |
-
},
|
| 1446 |
-
{
|
| 1447 |
-
"epoch": 1.08,
|
| 1448 |
-
"grad_norm": 1.0525788357504489,
|
| 1449 |
-
"learning_rate": 7.205694911803019e-06,
|
| 1450 |
-
"loss": 0.7962,
|
| 1451 |
-
"step": 1025
|
| 1452 |
-
},
|
| 1453 |
-
{
|
| 1454 |
-
"epoch": 1.09,
|
| 1455 |
-
"grad_norm": 0.986228023483833,
|
| 1456 |
-
"learning_rate": 7.18056811079932e-06,
|
| 1457 |
-
"loss": 0.79,
|
| 1458 |
-
"step": 1030
|
| 1459 |
-
},
|
| 1460 |
-
{
|
| 1461 |
-
"epoch": 1.09,
|
| 1462 |
-
"grad_norm": 1.031179895714868,
|
| 1463 |
-
"learning_rate": 7.155373122290508e-06,
|
| 1464 |
-
"loss": 0.8101,
|
| 1465 |
-
"step": 1035
|
| 1466 |
-
},
|
| 1467 |
-
{
|
| 1468 |
-
"epoch": 1.1,
|
| 1469 |
-
"grad_norm": 1.0379629517770603,
|
| 1470 |
-
"learning_rate": 7.13011073413705e-06,
|
| 1471 |
-
"loss": 0.781,
|
| 1472 |
-
"step": 1040
|
| 1473 |
-
},
|
| 1474 |
-
{
|
| 1475 |
-
"epoch": 1.1,
|
| 1476 |
-
"grad_norm": 1.033153108919124,
|
| 1477 |
-
"learning_rate": 7.1047817363070325e-06,
|
| 1478 |
-
"loss": 0.8418,
|
| 1479 |
-
"step": 1045
|
| 1480 |
-
},
|
| 1481 |
-
{
|
| 1482 |
-
"epoch": 1.11,
|
| 1483 |
-
"grad_norm": 1.0357203376239867,
|
| 1484 |
-
"learning_rate": 7.079386920851466e-06,
|
| 1485 |
-
"loss": 0.8065,
|
| 1486 |
-
"step": 1050
|
| 1487 |
-
},
|
| 1488 |
-
{
|
| 1489 |
-
"epoch": 1.11,
|
| 1490 |
-
"grad_norm": 1.0540192082846203,
|
| 1491 |
-
"learning_rate": 7.053927081879505e-06,
|
| 1492 |
-
"loss": 0.7956,
|
| 1493 |
-
"step": 1055
|
| 1494 |
-
},
|
| 1495 |
-
{
|
| 1496 |
-
"epoch": 1.12,
|
| 1497 |
-
"grad_norm": 1.0552828635725824,
|
| 1498 |
-
"learning_rate": 7.0284030155336315e-06,
|
| 1499 |
-
"loss": 0.7945,
|
| 1500 |
-
"step": 1060
|
| 1501 |
-
},
|
| 1502 |
-
{
|
| 1503 |
-
"epoch": 1.13,
|
| 1504 |
-
"grad_norm": 0.9810627289945896,
|
| 1505 |
-
"learning_rate": 7.002815519964745e-06,
|
| 1506 |
-
"loss": 0.7965,
|
| 1507 |
-
"step": 1065
|
| 1508 |
-
},
|
| 1509 |
-
{
|
| 1510 |
-
"epoch": 1.13,
|
| 1511 |
-
"grad_norm": 1.0916102744452092,
|
| 1512 |
-
"learning_rate": 6.977165395307215e-06,
|
| 1513 |
-
"loss": 0.7991,
|
| 1514 |
-
"step": 1070
|
| 1515 |
-
},
|
| 1516 |
-
{
|
| 1517 |
-
"epoch": 1.14,
|
| 1518 |
-
"grad_norm": 1.1543690326062077,
|
| 1519 |
-
"learning_rate": 6.951453443653852e-06,
|
| 1520 |
-
"loss": 0.7896,
|
| 1521 |
-
"step": 1075
|
| 1522 |
-
},
|
| 1523 |
-
{
|
| 1524 |
-
"epoch": 1.14,
|
| 1525 |
-
"grad_norm": 1.1170103600405488,
|
| 1526 |
-
"learning_rate": 6.9256804690308276e-06,
|
| 1527 |
-
"loss": 0.7828,
|
| 1528 |
-
"step": 1080
|
| 1529 |
-
},
|
| 1530 |
-
{
|
| 1531 |
-
"epoch": 1.15,
|
| 1532 |
-
"grad_norm": 1.0526733296614392,
|
| 1533 |
-
"learning_rate": 6.899847277372538e-06,
|
| 1534 |
-
"loss": 0.7923,
|
| 1535 |
-
"step": 1085
|
| 1536 |
-
},
|
| 1537 |
-
{
|
| 1538 |
-
"epoch": 1.15,
|
| 1539 |
-
"grad_norm": 1.0770254342023697,
|
| 1540 |
-
"learning_rate": 6.873954676496395e-06,
|
| 1541 |
-
"loss": 0.8128,
|
| 1542 |
-
"step": 1090
|
| 1543 |
-
},
|
| 1544 |
-
{
|
| 1545 |
-
"epoch": 1.16,
|
| 1546 |
-
"grad_norm": 1.037705594081886,
|
| 1547 |
-
"learning_rate": 6.848003476077567e-06,
|
| 1548 |
-
"loss": 0.7856,
|
| 1549 |
-
"step": 1095
|
| 1550 |
-
},
|
| 1551 |
-
{
|
| 1552 |
-
"epoch": 1.16,
|
| 1553 |
-
"grad_norm": 1.0319807068181204,
|
| 1554 |
-
"learning_rate": 6.8219944876236645e-06,
|
| 1555 |
-
"loss": 0.7949,
|
| 1556 |
-
"step": 1100
|
| 1557 |
-
},
|
| 1558 |
-
{
|
| 1559 |
-
"epoch": 1.17,
|
| 1560 |
-
"grad_norm": 1.0927555007584646,
|
| 1561 |
-
"learning_rate": 6.795928524449354e-06,
|
| 1562 |
-
"loss": 0.7941,
|
| 1563 |
-
"step": 1105
|
| 1564 |
-
},
|
| 1565 |
-
{
|
| 1566 |
-
"epoch": 1.17,
|
| 1567 |
-
"grad_norm": 0.9869897993273156,
|
| 1568 |
-
"learning_rate": 6.769806401650936e-06,
|
| 1569 |
-
"loss": 0.7667,
|
| 1570 |
-
"step": 1110
|
| 1571 |
-
},
|
| 1572 |
-
{
|
| 1573 |
-
"epoch": 1.18,
|
| 1574 |
-
"grad_norm": 1.0055956062759406,
|
| 1575 |
-
"learning_rate": 6.743628936080852e-06,
|
| 1576 |
-
"loss": 0.7855,
|
| 1577 |
-
"step": 1115
|
| 1578 |
-
},
|
| 1579 |
-
{
|
| 1580 |
-
"epoch": 1.18,
|
| 1581 |
-
"grad_norm": 1.0283367881989096,
|
| 1582 |
-
"learning_rate": 6.717396946322137e-06,
|
| 1583 |
-
"loss": 0.7745,
|
| 1584 |
-
"step": 1120
|
| 1585 |
-
},
|
| 1586 |
-
{
|
| 1587 |
-
"epoch": 1.19,
|
| 1588 |
-
"grad_norm": 1.0345829389670045,
|
| 1589 |
-
"learning_rate": 6.6911112526628295e-06,
|
| 1590 |
-
"loss": 0.7842,
|
| 1591 |
-
"step": 1125
|
| 1592 |
-
},
|
| 1593 |
-
{
|
| 1594 |
-
"epoch": 1.19,
|
| 1595 |
-
"grad_norm": 1.0711135328845822,
|
| 1596 |
-
"learning_rate": 6.664772677070316e-06,
|
| 1597 |
-
"loss": 0.7558,
|
| 1598 |
-
"step": 1130
|
| 1599 |
-
},
|
| 1600 |
-
{
|
| 1601 |
-
"epoch": 1.2,
|
| 1602 |
-
"grad_norm": 0.9877769296594265,
|
| 1603 |
-
"learning_rate": 6.638382043165628e-06,
|
| 1604 |
-
"loss": 0.7788,
|
| 1605 |
-
"step": 1135
|
| 1606 |
-
},
|
| 1607 |
-
{
|
| 1608 |
-
"epoch": 1.2,
|
| 1609 |
-
"grad_norm": 1.131836138091609,
|
| 1610 |
-
"learning_rate": 6.611940176197688e-06,
|
| 1611 |
-
"loss": 0.7901,
|
| 1612 |
-
"step": 1140
|
| 1613 |
-
},
|
| 1614 |
-
{
|
| 1615 |
-
"epoch": 1.21,
|
| 1616 |
-
"grad_norm": 1.058249641590972,
|
| 1617 |
-
"learning_rate": 6.585447903017506e-06,
|
| 1618 |
-
"loss": 0.7936,
|
| 1619 |
-
"step": 1145
|
| 1620 |
-
},
|
| 1621 |
-
{
|
| 1622 |
-
"epoch": 1.22,
|
| 1623 |
-
"grad_norm": 1.073971008814511,
|
| 1624 |
-
"learning_rate": 6.558906052052314e-06,
|
| 1625 |
-
"loss": 0.7835,
|
| 1626 |
-
"step": 1150
|
| 1627 |
-
},
|
| 1628 |
-
{
|
| 1629 |
-
"epoch": 1.22,
|
| 1630 |
-
"grad_norm": 1.0491301969369466,
|
| 1631 |
-
"learning_rate": 6.532315453279673e-06,
|
| 1632 |
-
"loss": 0.7902,
|
| 1633 |
-
"step": 1155
|
| 1634 |
-
},
|
| 1635 |
-
{
|
| 1636 |
-
"epoch": 1.23,
|
| 1637 |
-
"grad_norm": 1.046297097483487,
|
| 1638 |
-
"learning_rate": 6.505676938201512e-06,
|
| 1639 |
-
"loss": 0.7767,
|
| 1640 |
-
"step": 1160
|
| 1641 |
-
},
|
| 1642 |
-
{
|
| 1643 |
-
"epoch": 1.23,
|
| 1644 |
-
"grad_norm": 1.046022517875942,
|
| 1645 |
-
"learning_rate": 6.478991339818128e-06,
|
| 1646 |
-
"loss": 0.8091,
|
| 1647 |
-
"step": 1165
|
| 1648 |
-
},
|
| 1649 |
-
{
|
| 1650 |
-
"epoch": 1.24,
|
| 1651 |
-
"grad_norm": 1.0086633248074561,
|
| 1652 |
-
"learning_rate": 6.4522594926021355e-06,
|
| 1653 |
-
"loss": 0.7797,
|
| 1654 |
-
"step": 1170
|
| 1655 |
-
},
|
| 1656 |
-
{
|
| 1657 |
-
"epoch": 1.24,
|
| 1658 |
-
"grad_norm": 1.0965955454651117,
|
| 1659 |
-
"learning_rate": 6.425482232472377e-06,
|
| 1660 |
-
"loss": 0.7702,
|
| 1661 |
-
"step": 1175
|
| 1662 |
-
},
|
| 1663 |
-
{
|
| 1664 |
-
"epoch": 1.25,
|
| 1665 |
-
"grad_norm": 1.0362189192150881,
|
| 1666 |
-
"learning_rate": 6.3986603967677805e-06,
|
| 1667 |
-
"loss": 0.7931,
|
| 1668 |
-
"step": 1180
|
| 1669 |
-
},
|
| 1670 |
-
{
|
| 1671 |
-
"epoch": 1.25,
|
| 1672 |
-
"grad_norm": 1.110468197330772,
|
| 1673 |
-
"learning_rate": 6.371794824221173e-06,
|
| 1674 |
-
"loss": 0.7917,
|
| 1675 |
-
"step": 1185
|
| 1676 |
-
},
|
| 1677 |
-
{
|
| 1678 |
-
"epoch": 1.26,
|
| 1679 |
-
"grad_norm": 1.0163659020071605,
|
| 1680 |
-
"learning_rate": 6.344886354933058e-06,
|
| 1681 |
-
"loss": 0.7886,
|
| 1682 |
-
"step": 1190
|
| 1683 |
-
},
|
| 1684 |
-
{
|
| 1685 |
-
"epoch": 1.26,
|
| 1686 |
-
"grad_norm": 1.0115549227695064,
|
| 1687 |
-
"learning_rate": 6.3179358303453386e-06,
|
| 1688 |
-
"loss": 0.7511,
|
| 1689 |
-
"step": 1195
|
| 1690 |
-
},
|
| 1691 |
-
{
|
| 1692 |
-
"epoch": 1.27,
|
| 1693 |
-
"grad_norm": 1.0872016119161863,
|
| 1694 |
-
"learning_rate": 6.290944093215016e-06,
|
| 1695 |
-
"loss": 0.8036,
|
| 1696 |
-
"step": 1200
|
| 1697 |
-
},
|
| 1698 |
-
{
|
| 1699 |
-
"epoch": 1.27,
|
| 1700 |
-
"grad_norm": 1.0553500518484338,
|
| 1701 |
-
"learning_rate": 6.263911987587822e-06,
|
| 1702 |
-
"loss": 0.7938,
|
| 1703 |
-
"step": 1205
|
| 1704 |
-
},
|
| 1705 |
-
{
|
| 1706 |
-
"epoch": 1.28,
|
| 1707 |
-
"grad_norm": 0.993815270148442,
|
| 1708 |
-
"learning_rate": 6.236840358771837e-06,
|
| 1709 |
-
"loss": 0.7788,
|
| 1710 |
-
"step": 1210
|
| 1711 |
-
},
|
| 1712 |
-
{
|
| 1713 |
-
"epoch": 1.28,
|
| 1714 |
-
"grad_norm": 1.0605675582324252,
|
| 1715 |
-
"learning_rate": 6.20973005331105e-06,
|
| 1716 |
-
"loss": 0.7781,
|
| 1717 |
-
"step": 1215
|
| 1718 |
-
},
|
| 1719 |
-
{
|
| 1720 |
-
"epoch": 1.29,
|
| 1721 |
-
"grad_norm": 1.0965085071552372,
|
| 1722 |
-
"learning_rate": 6.1825819189588885e-06,
|
| 1723 |
-
"loss": 0.7872,
|
| 1724 |
-
"step": 1220
|
| 1725 |
-
},
|
| 1726 |
-
{
|
| 1727 |
-
"epoch": 1.29,
|
| 1728 |
-
"grad_norm": 1.040866195350916,
|
| 1729 |
-
"learning_rate": 6.155396804651714e-06,
|
| 1730 |
-
"loss": 0.7966,
|
| 1731 |
-
"step": 1225
|
| 1732 |
-
},
|
| 1733 |
-
{
|
| 1734 |
-
"epoch": 1.3,
|
| 1735 |
-
"grad_norm": 1.0593376609536802,
|
| 1736 |
-
"learning_rate": 6.128175560482264e-06,
|
| 1737 |
-
"loss": 0.7832,
|
| 1738 |
-
"step": 1230
|
| 1739 |
-
},
|
| 1740 |
-
{
|
| 1741 |
-
"epoch": 1.31,
|
| 1742 |
-
"grad_norm": 1.0081718313330637,
|
| 1743 |
-
"learning_rate": 6.1009190376730785e-06,
|
| 1744 |
-
"loss": 0.7772,
|
| 1745 |
-
"step": 1235
|
| 1746 |
-
},
|
| 1747 |
-
{
|
| 1748 |
-
"epoch": 1.31,
|
| 1749 |
-
"grad_norm": 0.9892554397828908,
|
| 1750 |
-
"learning_rate": 6.07362808854988e-06,
|
| 1751 |
-
"loss": 0.7856,
|
| 1752 |
-
"step": 1240
|
| 1753 |
-
},
|
| 1754 |
-
{
|
| 1755 |
-
"epoch": 1.32,
|
| 1756 |
-
"grad_norm": 1.0515874983049542,
|
| 1757 |
-
"learning_rate": 6.046303566514919e-06,
|
| 1758 |
-
"loss": 0.7812,
|
| 1759 |
-
"step": 1245
|
| 1760 |
-
},
|
| 1761 |
-
{
|
| 1762 |
-
"epoch": 1.32,
|
| 1763 |
-
"grad_norm": 1.01738547568124,
|
| 1764 |
-
"learning_rate": 6.018946326020287e-06,
|
| 1765 |
-
"loss": 0.7824,
|
| 1766 |
-
"step": 1250
|
| 1767 |
-
},
|
| 1768 |
-
{
|
| 1769 |
-
"epoch": 1.33,
|
| 1770 |
-
"grad_norm": 0.992994982201507,
|
| 1771 |
-
"learning_rate": 5.991557222541201e-06,
|
| 1772 |
-
"loss": 0.7842,
|
| 1773 |
-
"step": 1255
|
| 1774 |
-
},
|
| 1775 |
-
{
|
| 1776 |
-
"epoch": 1.33,
|
| 1777 |
-
"grad_norm": 0.9928822859609259,
|
| 1778 |
-
"learning_rate": 5.964137112549251e-06,
|
| 1779 |
-
"loss": 0.7906,
|
| 1780 |
-
"step": 1260
|
| 1781 |
-
},
|
| 1782 |
-
{
|
| 1783 |
-
"epoch": 1.34,
|
| 1784 |
-
"grad_norm": 1.0673862770846931,
|
| 1785 |
-
"learning_rate": 5.9366868534856115e-06,
|
| 1786 |
-
"loss": 0.7896,
|
| 1787 |
-
"step": 1265
|
| 1788 |
-
},
|
| 1789 |
-
{
|
| 1790 |
-
"epoch": 1.34,
|
| 1791 |
-
"grad_norm": 1.0627251705995355,
|
| 1792 |
-
"learning_rate": 5.909207303734241e-06,
|
| 1793 |
-
"loss": 0.7965,
|
| 1794 |
-
"step": 1270
|
| 1795 |
-
},
|
| 1796 |
-
{
|
| 1797 |
-
"epoch": 1.35,
|
| 1798 |
-
"grad_norm": 1.0050051635503012,
|
| 1799 |
-
"learning_rate": 5.881699322595031e-06,
|
| 1800 |
-
"loss": 0.7775,
|
| 1801 |
-
"step": 1275
|
| 1802 |
-
},
|
| 1803 |
-
{
|
| 1804 |
-
"epoch": 1.35,
|
| 1805 |
-
"grad_norm": 1.0049258262531797,
|
| 1806 |
-
"learning_rate": 5.854163770256934e-06,
|
| 1807 |
-
"loss": 0.7659,
|
| 1808 |
-
"step": 1280
|
| 1809 |
-
},
|
| 1810 |
-
{
|
| 1811 |
-
"epoch": 1.36,
|
| 1812 |
-
"grad_norm": 1.1097225296353777,
|
| 1813 |
-
"learning_rate": 5.826601507771073e-06,
|
| 1814 |
-
"loss": 0.7699,
|
| 1815 |
-
"step": 1285
|
| 1816 |
-
},
|
| 1817 |
-
{
|
| 1818 |
-
"epoch": 1.36,
|
| 1819 |
-
"grad_norm": 1.0610730723756006,
|
| 1820 |
-
"learning_rate": 5.799013397023806e-06,
|
| 1821 |
-
"loss": 0.7996,
|
| 1822 |
-
"step": 1290
|
| 1823 |
-
},
|
| 1824 |
-
{
|
| 1825 |
-
"epoch": 1.37,
|
| 1826 |
-
"grad_norm": 1.0285633823079718,
|
| 1827 |
-
"learning_rate": 5.771400300709785e-06,
|
| 1828 |
-
"loss": 0.7829,
|
| 1829 |
-
"step": 1295
|
| 1830 |
-
},
|
| 1831 |
-
{
|
| 1832 |
-
"epoch": 1.37,
|
| 1833 |
-
"grad_norm": 1.0484599021027985,
|
| 1834 |
-
"learning_rate": 5.743763082304973e-06,
|
| 1835 |
-
"loss": 0.7619,
|
| 1836 |
-
"step": 1300
|
| 1837 |
-
},
|
| 1838 |
-
{
|
| 1839 |
-
"epoch": 1.38,
|
| 1840 |
-
"grad_norm": 1.0137701786577156,
|
| 1841 |
-
"learning_rate": 5.7161026060396375e-06,
|
| 1842 |
-
"loss": 0.798,
|
| 1843 |
-
"step": 1305
|
| 1844 |
-
},
|
| 1845 |
-
{
|
| 1846 |
-
"epoch": 1.38,
|
| 1847 |
-
"grad_norm": 1.0289414598602742,
|
| 1848 |
-
"learning_rate": 5.688419736871341e-06,
|
| 1849 |
-
"loss": 0.7827,
|
| 1850 |
-
"step": 1310
|
| 1851 |
-
},
|
| 1852 |
-
{
|
| 1853 |
-
"epoch": 1.39,
|
| 1854 |
-
"grad_norm": 1.058376335913828,
|
| 1855 |
-
"learning_rate": 5.660715340457874e-06,
|
| 1856 |
-
"loss": 0.7921,
|
| 1857 |
-
"step": 1315
|
| 1858 |
-
},
|
| 1859 |
-
{
|
| 1860 |
-
"epoch": 1.39,
|
| 1861 |
-
"grad_norm": 1.0011219088912342,
|
| 1862 |
-
"learning_rate": 5.632990283130204e-06,
|
| 1863 |
-
"loss": 0.781,
|
| 1864 |
-
"step": 1320
|
| 1865 |
-
},
|
| 1866 |
-
{
|
| 1867 |
-
"epoch": 1.4,
|
| 1868 |
-
"grad_norm": 0.984264955084216,
|
| 1869 |
-
"learning_rate": 5.605245431865368e-06,
|
| 1870 |
-
"loss": 0.7772,
|
| 1871 |
-
"step": 1325
|
| 1872 |
-
},
|
| 1873 |
-
{
|
| 1874 |
-
"epoch": 1.41,
|
| 1875 |
-
"grad_norm": 1.0151072044919451,
|
| 1876 |
-
"learning_rate": 5.577481654259377e-06,
|
| 1877 |
-
"loss": 0.7735,
|
| 1878 |
-
"step": 1330
|
| 1879 |
-
},
|
| 1880 |
-
{
|
| 1881 |
-
"epoch": 1.41,
|
| 1882 |
-
"grad_norm": 1.063533843295668,
|
| 1883 |
-
"learning_rate": 5.549699818500074e-06,
|
| 1884 |
-
"loss": 0.7682,
|
| 1885 |
-
"step": 1335
|
| 1886 |
-
},
|
| 1887 |
-
{
|
| 1888 |
-
"epoch": 1.42,
|
| 1889 |
-
"grad_norm": 1.0434635789190496,
|
| 1890 |
-
"learning_rate": 5.521900793339989e-06,
|
| 1891 |
-
"loss": 0.7915,
|
| 1892 |
-
"step": 1340
|
| 1893 |
-
},
|
| 1894 |
-
{
|
| 1895 |
-
"epoch": 1.42,
|
| 1896 |
-
"grad_norm": 1.0587561050751115,
|
| 1897 |
-
"learning_rate": 5.494085448069181e-06,
|
| 1898 |
-
"loss": 0.7997,
|
| 1899 |
-
"step": 1345
|
| 1900 |
-
},
|
| 1901 |
-
{
|
| 1902 |
-
"epoch": 1.43,
|
| 1903 |
-
"grad_norm": 1.0758864296233028,
|
| 1904 |
-
"learning_rate": 5.466254652488036e-06,
|
| 1905 |
-
"loss": 0.7964,
|
| 1906 |
-
"step": 1350
|
| 1907 |
-
},
|
| 1908 |
-
{
|
| 1909 |
-
"epoch": 1.43,
|
| 1910 |
-
"grad_norm": 1.0556751372323996,
|
| 1911 |
-
"learning_rate": 5.438409276880089e-06,
|
| 1912 |
-
"loss": 0.8062,
|
| 1913 |
-
"step": 1355
|
| 1914 |
-
},
|
| 1915 |
-
{
|
| 1916 |
-
"epoch": 1.44,
|
| 1917 |
-
"grad_norm": 0.9792859835280993,
|
| 1918 |
-
"learning_rate": 5.410550191984798e-06,
|
| 1919 |
-
"loss": 0.787,
|
| 1920 |
-
"step": 1360
|
| 1921 |
-
},
|
| 1922 |
-
{
|
| 1923 |
-
"epoch": 1.44,
|
| 1924 |
-
"grad_norm": 1.0231438624972786,
|
| 1925 |
-
"learning_rate": 5.3826782689703115e-06,
|
| 1926 |
-
"loss": 0.7803,
|
| 1927 |
-
"step": 1365
|
| 1928 |
-
},
|
| 1929 |
-
{
|
| 1930 |
-
"epoch": 1.45,
|
| 1931 |
-
"grad_norm": 1.0660534726358564,
|
| 1932 |
-
"learning_rate": 5.354794379406242e-06,
|
| 1933 |
-
"loss": 0.78,
|
| 1934 |
-
"step": 1370
|
| 1935 |
-
},
|
| 1936 |
-
{
|
| 1937 |
-
"epoch": 1.45,
|
| 1938 |
-
"grad_norm": 0.9527414539128428,
|
| 1939 |
-
"learning_rate": 5.3268993952363936e-06,
|
| 1940 |
-
"loss": 0.796,
|
| 1941 |
-
"step": 1375
|
| 1942 |
-
},
|
| 1943 |
-
{
|
| 1944 |
-
"epoch": 1.46,
|
| 1945 |
-
"grad_norm": 0.9870931434726852,
|
| 1946 |
-
"learning_rate": 5.29899418875151e-06,
|
| 1947 |
-
"loss": 0.7652,
|
| 1948 |
-
"step": 1380
|
| 1949 |
-
},
|
| 1950 |
-
{
|
| 1951 |
-
"epoch": 1.46,
|
| 1952 |
-
"grad_norm": 1.0537299945885146,
|
| 1953 |
-
"learning_rate": 5.271079632561992e-06,
|
| 1954 |
-
"loss": 0.7854,
|
| 1955 |
-
"step": 1385
|
| 1956 |
-
},
|
| 1957 |
-
{
|
| 1958 |
-
"epoch": 1.47,
|
| 1959 |
-
"grad_norm": 1.1396368040574916,
|
| 1960 |
-
"learning_rate": 5.243156599570606e-06,
|
| 1961 |
-
"loss": 0.7617,
|
| 1962 |
-
"step": 1390
|
| 1963 |
-
},
|
| 1964 |
-
{
|
| 1965 |
-
"epoch": 1.47,
|
| 1966 |
-
"grad_norm": 1.0924704024745873,
|
| 1967 |
-
"learning_rate": 5.2152259629451986e-06,
|
| 1968 |
-
"loss": 0.7713,
|
| 1969 |
-
"step": 1395
|
| 1970 |
-
},
|
| 1971 |
-
{
|
| 1972 |
-
"epoch": 1.48,
|
| 1973 |
-
"grad_norm": 1.021493417245078,
|
| 1974 |
-
"learning_rate": 5.18728859609138e-06,
|
| 1975 |
-
"loss": 0.7609,
|
| 1976 |
-
"step": 1400
|
| 1977 |
-
},
|
| 1978 |
-
{
|
| 1979 |
-
"epoch": 1.48,
|
| 1980 |
-
"grad_norm": 1.0148194958691719,
|
| 1981 |
-
"learning_rate": 5.159345372625223e-06,
|
| 1982 |
-
"loss": 0.7788,
|
| 1983 |
-
"step": 1405
|
| 1984 |
-
},
|
| 1985 |
-
{
|
| 1986 |
-
"epoch": 1.49,
|
| 1987 |
-
"grad_norm": 1.0402765811164951,
|
| 1988 |
-
"learning_rate": 5.131397166345938e-06,
|
| 1989 |
-
"loss": 0.7599,
|
| 1990 |
-
"step": 1410
|
| 1991 |
-
},
|
| 1992 |
-
{
|
| 1993 |
-
"epoch": 1.5,
|
| 1994 |
-
"grad_norm": 0.9966250584272072,
|
| 1995 |
-
"learning_rate": 5.103444851208549e-06,
|
| 1996 |
-
"loss": 0.7874,
|
| 1997 |
-
"step": 1415
|
| 1998 |
-
},
|
| 1999 |
-
{
|
| 2000 |
-
"epoch": 1.5,
|
| 2001 |
-
"grad_norm": 0.9871275158697829,
|
| 2002 |
-
"learning_rate": 5.075489301296567e-06,
|
| 2003 |
-
"loss": 0.7566,
|
| 2004 |
-
"step": 1420
|
| 2005 |
-
},
|
| 2006 |
-
{
|
| 2007 |
-
"epoch": 1.51,
|
| 2008 |
-
"grad_norm": 1.0896451679213162,
|
| 2009 |
-
"learning_rate": 5.047531390794661e-06,
|
| 2010 |
-
"loss": 0.7699,
|
| 2011 |
-
"step": 1425
|
| 2012 |
-
},
|
| 2013 |
-
{
|
| 2014 |
-
"epoch": 1.51,
|
| 2015 |
-
"grad_norm": 1.1203863877988638,
|
| 2016 |
-
"learning_rate": 5.019571993961307e-06,
|
| 2017 |
-
"loss": 0.8088,
|
| 2018 |
-
"step": 1430
|
| 2019 |
-
},
|
| 2020 |
-
{
|
| 2021 |
-
"epoch": 1.52,
|
| 2022 |
-
"grad_norm": 1.03311513179617,
|
| 2023 |
-
"learning_rate": 4.9916119851014664e-06,
|
| 2024 |
-
"loss": 0.7739,
|
| 2025 |
-
"step": 1435
|
| 2026 |
-
},
|
| 2027 |
-
{
|
| 2028 |
-
"epoch": 1.52,
|
| 2029 |
-
"grad_norm": 1.0389351009988612,
|
| 2030 |
-
"learning_rate": 4.96365223853924e-06,
|
| 2031 |
-
"loss": 0.7816,
|
| 2032 |
-
"step": 1440
|
| 2033 |
-
},
|
| 2034 |
-
{
|
| 2035 |
-
"epoch": 1.53,
|
| 2036 |
-
"grad_norm": 0.9960641498632878,
|
| 2037 |
-
"learning_rate": 4.93569362859052e-06,
|
| 2038 |
-
"loss": 0.775,
|
| 2039 |
-
"step": 1445
|
| 2040 |
-
},
|
| 2041 |
-
{
|
| 2042 |
-
"epoch": 1.53,
|
| 2043 |
-
"grad_norm": 0.9388823495229471,
|
| 2044 |
-
"learning_rate": 4.907737029535664e-06,
|
| 2045 |
-
"loss": 0.756,
|
| 2046 |
-
"step": 1450
|
| 2047 |
-
},
|
| 2048 |
-
{
|
| 2049 |
-
"epoch": 1.54,
|
| 2050 |
-
"grad_norm": 1.0662538022442485,
|
| 2051 |
-
"learning_rate": 4.8797833155921396e-06,
|
| 2052 |
-
"loss": 0.7992,
|
| 2053 |
-
"step": 1455
|
| 2054 |
-
},
|
| 2055 |
-
{
|
| 2056 |
-
"epoch": 1.54,
|
| 2057 |
-
"grad_norm": 1.0350212904727674,
|
| 2058 |
-
"learning_rate": 4.8518333608872015e-06,
|
| 2059 |
-
"loss": 0.7595,
|
| 2060 |
-
"step": 1460
|
| 2061 |
-
},
|
| 2062 |
-
{
|
| 2063 |
-
"epoch": 1.55,
|
| 2064 |
-
"grad_norm": 0.9967538128228846,
|
| 2065 |
-
"learning_rate": 4.823888039430551e-06,
|
| 2066 |
-
"loss": 0.7582,
|
| 2067 |
-
"step": 1465
|
| 2068 |
-
},
|
| 2069 |
-
{
|
| 2070 |
-
"epoch": 1.55,
|
| 2071 |
-
"grad_norm": 1.0139079612075497,
|
| 2072 |
-
"learning_rate": 4.795948225087001e-06,
|
| 2073 |
-
"loss": 0.7709,
|
| 2074 |
-
"step": 1470
|
| 2075 |
-
},
|
| 2076 |
-
{
|
| 2077 |
-
"epoch": 1.56,
|
| 2078 |
-
"grad_norm": 1.0510044388149635,
|
| 2079 |
-
"learning_rate": 4.7680147915491585e-06,
|
| 2080 |
-
"loss": 0.7692,
|
| 2081 |
-
"step": 1475
|
| 2082 |
-
},
|
| 2083 |
-
{
|
| 2084 |
-
"epoch": 1.56,
|
| 2085 |
-
"grad_norm": 1.0641353890612333,
|
| 2086 |
-
"learning_rate": 4.740088612310096e-06,
|
| 2087 |
-
"loss": 0.7847,
|
| 2088 |
-
"step": 1480
|
| 2089 |
-
},
|
| 2090 |
-
{
|
| 2091 |
-
"epoch": 1.57,
|
| 2092 |
-
"grad_norm": 1.0192435995305715,
|
| 2093 |
-
"learning_rate": 4.7121705606360424e-06,
|
| 2094 |
-
"loss": 0.7732,
|
| 2095 |
-
"step": 1485
|
| 2096 |
-
},
|
| 2097 |
-
{
|
| 2098 |
-
"epoch": 1.57,
|
| 2099 |
-
"grad_norm": 1.0076325415256413,
|
| 2100 |
-
"learning_rate": 4.684261509539072e-06,
|
| 2101 |
-
"loss": 0.7701,
|
| 2102 |
-
"step": 1490
|
| 2103 |
-
},
|
| 2104 |
-
{
|
| 2105 |
-
"epoch": 1.58,
|
| 2106 |
-
"grad_norm": 0.9707102286396411,
|
| 2107 |
-
"learning_rate": 4.65636233174981e-06,
|
| 2108 |
-
"loss": 0.77,
|
| 2109 |
-
"step": 1495
|
| 2110 |
-
},
|
| 2111 |
-
{
|
| 2112 |
-
"epoch": 1.59,
|
| 2113 |
-
"grad_norm": 1.0835636202474823,
|
| 2114 |
-
"learning_rate": 4.628473899690133e-06,
|
| 2115 |
-
"loss": 0.7849,
|
| 2116 |
-
"step": 1500
|
| 2117 |
-
},
|
| 2118 |
-
{
|
| 2119 |
-
"epoch": 1.59,
|
| 2120 |
-
"grad_norm": 1.0157410126136626,
|
| 2121 |
-
"learning_rate": 4.600597085445894e-06,
|
| 2122 |
-
"loss": 0.784,
|
| 2123 |
-
"step": 1505
|
| 2124 |
-
},
|
| 2125 |
-
{
|
| 2126 |
-
"epoch": 1.6,
|
| 2127 |
-
"grad_norm": 1.0616186913926178,
|
| 2128 |
-
"learning_rate": 4.572732760739653e-06,
|
| 2129 |
-
"loss": 0.7785,
|
| 2130 |
-
"step": 1510
|
| 2131 |
-
},
|
| 2132 |
-
{
|
| 2133 |
-
"epoch": 1.6,
|
| 2134 |
-
"grad_norm": 1.006516145178769,
|
| 2135 |
-
"learning_rate": 4.5448817969034165e-06,
|
| 2136 |
-
"loss": 0.7753,
|
| 2137 |
-
"step": 1515
|
| 2138 |
-
},
|
| 2139 |
-
{
|
| 2140 |
-
"epoch": 1.61,
|
| 2141 |
-
"grad_norm": 1.0480529823653495,
|
| 2142 |
-
"learning_rate": 4.517045064851386e-06,
|
| 2143 |
-
"loss": 0.7989,
|
| 2144 |
-
"step": 1520
|
| 2145 |
-
},
|
| 2146 |
-
{
|
| 2147 |
-
"epoch": 1.61,
|
| 2148 |
-
"grad_norm": 1.0432567441250045,
|
| 2149 |
-
"learning_rate": 4.489223435052732e-06,
|
| 2150 |
-
"loss": 0.7946,
|
| 2151 |
-
"step": 1525
|
| 2152 |
-
},
|
| 2153 |
-
{
|
| 2154 |
-
"epoch": 1.62,
|
| 2155 |
-
"grad_norm": 1.0461342178531015,
|
| 2156 |
-
"learning_rate": 4.461417777504363e-06,
|
| 2157 |
-
"loss": 0.7676,
|
| 2158 |
-
"step": 1530
|
| 2159 |
-
},
|
| 2160 |
-
{
|
| 2161 |
-
"epoch": 1.62,
|
| 2162 |
-
"grad_norm": 1.0045382622138492,
|
| 2163 |
-
"learning_rate": 4.433628961703733e-06,
|
| 2164 |
-
"loss": 0.7651,
|
| 2165 |
-
"step": 1535
|
| 2166 |
-
},
|
| 2167 |
-
{
|
| 2168 |
-
"epoch": 1.63,
|
| 2169 |
-
"grad_norm": 0.9890094489435823,
|
| 2170 |
-
"learning_rate": 4.405857856621644e-06,
|
| 2171 |
-
"loss": 0.7943,
|
| 2172 |
-
"step": 1540
|
| 2173 |
-
},
|
| 2174 |
-
{
|
| 2175 |
-
"epoch": 1.63,
|
| 2176 |
-
"grad_norm": 1.0127639919495397,
|
| 2177 |
-
"learning_rate": 4.378105330675074e-06,
|
| 2178 |
-
"loss": 0.7895,
|
| 2179 |
-
"step": 1545
|
| 2180 |
-
},
|
| 2181 |
-
{
|
| 2182 |
-
"epoch": 1.64,
|
| 2183 |
-
"grad_norm": 1.0398544121817734,
|
| 2184 |
-
"learning_rate": 4.350372251700025e-06,
|
| 2185 |
-
"loss": 0.8004,
|
| 2186 |
-
"step": 1550
|
| 2187 |
-
},
|
| 2188 |
-
{
|
| 2189 |
-
"epoch": 1.64,
|
| 2190 |
-
"grad_norm": 1.037857459368961,
|
| 2191 |
-
"learning_rate": 4.322659486924373e-06,
|
| 2192 |
-
"loss": 0.7963,
|
| 2193 |
-
"step": 1555
|
| 2194 |
-
},
|
| 2195 |
-
{
|
| 2196 |
-
"epoch": 1.65,
|
| 2197 |
-
"grad_norm": 1.106103919813531,
|
| 2198 |
-
"learning_rate": 4.294967902940768e-06,
|
| 2199 |
-
"loss": 0.787,
|
| 2200 |
-
"step": 1560
|
| 2201 |
-
},
|
| 2202 |
-
{
|
| 2203 |
-
"epoch": 1.65,
|
| 2204 |
-
"grad_norm": 1.0865617469424886,
|
| 2205 |
-
"learning_rate": 4.267298365679522e-06,
|
| 2206 |
-
"loss": 0.788,
|
| 2207 |
-
"step": 1565
|
| 2208 |
-
},
|
| 2209 |
-
{
|
| 2210 |
-
"epoch": 1.66,
|
| 2211 |
-
"grad_norm": 1.0303226290700802,
|
| 2212 |
-
"learning_rate": 4.239651740381534e-06,
|
| 2213 |
-
"loss": 0.7642,
|
| 2214 |
-
"step": 1570
|
| 2215 |
-
},
|
| 2216 |
-
{
|
| 2217 |
-
"epoch": 1.66,
|
| 2218 |
-
"grad_norm": 1.0512505166055992,
|
| 2219 |
-
"learning_rate": 4.212028891571237e-06,
|
| 2220 |
-
"loss": 0.7832,
|
| 2221 |
-
"step": 1575
|
| 2222 |
-
},
|
| 2223 |
-
{
|
| 2224 |
-
"epoch": 1.67,
|
| 2225 |
-
"grad_norm": 1.0750316874597787,
|
| 2226 |
-
"learning_rate": 4.184430683029552e-06,
|
| 2227 |
-
"loss": 0.7599,
|
| 2228 |
-
"step": 1580
|
| 2229 |
-
},
|
| 2230 |
-
{
|
| 2231 |
-
"epoch": 1.68,
|
| 2232 |
-
"grad_norm": 1.0622608820174235,
|
| 2233 |
-
"learning_rate": 4.156857977766896e-06,
|
| 2234 |
-
"loss": 0.7841,
|
| 2235 |
-
"step": 1585
|
| 2236 |
-
},
|
| 2237 |
-
{
|
| 2238 |
-
"epoch": 1.68,
|
| 2239 |
-
"grad_norm": 1.0023528643121005,
|
| 2240 |
-
"learning_rate": 4.129311637996182e-06,
|
| 2241 |
-
"loss": 0.7845,
|
| 2242 |
-
"step": 1590
|
| 2243 |
-
},
|
| 2244 |
-
{
|
| 2245 |
-
"epoch": 1.69,
|
| 2246 |
-
"grad_norm": 1.0597451506484419,
|
| 2247 |
-
"learning_rate": 4.101792525105857e-06,
|
| 2248 |
-
"loss": 0.7802,
|
| 2249 |
-
"step": 1595
|
| 2250 |
-
},
|
| 2251 |
-
{
|
| 2252 |
-
"epoch": 1.69,
|
| 2253 |
-
"grad_norm": 0.9622973096022323,
|
| 2254 |
-
"learning_rate": 4.0743014996329764e-06,
|
| 2255 |
-
"loss": 0.7678,
|
| 2256 |
-
"step": 1600
|
| 2257 |
-
},
|
| 2258 |
-
{
|
| 2259 |
-
"epoch": 1.7,
|
| 2260 |
-
"grad_norm": 1.051095411122212,
|
| 2261 |
-
"learning_rate": 4.046839421236276e-06,
|
| 2262 |
-
"loss": 0.7972,
|
| 2263 |
-
"step": 1605
|
| 2264 |
-
},
|
| 2265 |
-
{
|
| 2266 |
-
"epoch": 1.7,
|
| 2267 |
-
"grad_norm": 1.0082128589578265,
|
| 2268 |
-
"learning_rate": 4.019407148669312e-06,
|
| 2269 |
-
"loss": 0.7948,
|
| 2270 |
-
"step": 1610
|
| 2271 |
-
},
|
| 2272 |
-
{
|
| 2273 |
-
"epoch": 1.71,
|
| 2274 |
-
"grad_norm": 1.0901759578931909,
|
| 2275 |
-
"learning_rate": 3.992005539753592e-06,
|
| 2276 |
-
"loss": 0.7914,
|
| 2277 |
-
"step": 1615
|
| 2278 |
-
},
|
| 2279 |
-
{
|
| 2280 |
-
"epoch": 1.71,
|
| 2281 |
-
"grad_norm": 1.0584302499373435,
|
| 2282 |
-
"learning_rate": 3.964635451351758e-06,
|
| 2283 |
-
"loss": 0.7821,
|
| 2284 |
-
"step": 1620
|
| 2285 |
-
},
|
| 2286 |
-
{
|
| 2287 |
-
"epoch": 1.72,
|
| 2288 |
-
"grad_norm": 1.043189384648134,
|
| 2289 |
-
"learning_rate": 3.937297739340783e-06,
|
| 2290 |
-
"loss": 0.778,
|
| 2291 |
-
"step": 1625
|
| 2292 |
-
},
|
| 2293 |
-
{
|
| 2294 |
-
"epoch": 1.72,
|
| 2295 |
-
"grad_norm": 1.0245392793145456,
|
| 2296 |
-
"learning_rate": 3.909993258585219e-06,
|
| 2297 |
-
"loss": 0.7908,
|
| 2298 |
-
"step": 1630
|
| 2299 |
-
},
|
| 2300 |
-
{
|
| 2301 |
-
"epoch": 1.73,
|
| 2302 |
-
"grad_norm": 1.0082519645854728,
|
| 2303 |
-
"learning_rate": 3.882722862910458e-06,
|
| 2304 |
-
"loss": 0.7793,
|
| 2305 |
-
"step": 1635
|
| 2306 |
-
},
|
| 2307 |
-
{
|
| 2308 |
-
"epoch": 1.73,
|
| 2309 |
-
"grad_norm": 1.0211341337802105,
|
| 2310 |
-
"learning_rate": 3.8554874050760345e-06,
|
| 2311 |
-
"loss": 0.8042,
|
| 2312 |
-
"step": 1640
|
| 2313 |
-
},
|
| 2314 |
-
{
|
| 2315 |
-
"epoch": 1.74,
|
| 2316 |
-
"grad_norm": 0.9920127978660441,
|
| 2317 |
-
"learning_rate": 3.828287736748957e-06,
|
| 2318 |
-
"loss": 0.758,
|
| 2319 |
-
"step": 1645
|
| 2320 |
-
},
|
| 2321 |
-
{
|
| 2322 |
-
"epoch": 1.74,
|
| 2323 |
-
"grad_norm": 1.0187229111502758,
|
| 2324 |
-
"learning_rate": 3.8011247084770754e-06,
|
| 2325 |
-
"loss": 0.7986,
|
| 2326 |
-
"step": 1650
|
| 2327 |
-
},
|
| 2328 |
-
{
|
| 2329 |
-
"epoch": 1.75,
|
| 2330 |
-
"grad_norm": 0.9982295207578855,
|
| 2331 |
-
"learning_rate": 3.773999169662489e-06,
|
| 2332 |
-
"loss": 0.7623,
|
| 2333 |
-
"step": 1655
|
| 2334 |
-
},
|
| 2335 |
-
{
|
| 2336 |
-
"epoch": 1.75,
|
| 2337 |
-
"grad_norm": 1.025180441312379,
|
| 2338 |
-
"learning_rate": 3.746911968534982e-06,
|
| 2339 |
-
"loss": 0.7454,
|
| 2340 |
-
"step": 1660
|
| 2341 |
-
},
|
| 2342 |
-
{
|
| 2343 |
-
"epoch": 1.76,
|
| 2344 |
-
"grad_norm": 0.9884338430346545,
|
| 2345 |
-
"learning_rate": 3.7198639521254988e-06,
|
| 2346 |
-
"loss": 0.7671,
|
| 2347 |
-
"step": 1665
|
| 2348 |
-
},
|
| 2349 |
-
{
|
| 2350 |
-
"epoch": 1.76,
|
| 2351 |
-
"grad_norm": 0.9685352318412103,
|
| 2352 |
-
"learning_rate": 3.6928559662396574e-06,
|
| 2353 |
-
"loss": 0.7583,
|
| 2354 |
-
"step": 1670
|
| 2355 |
-
},
|
| 2356 |
-
{
|
| 2357 |
-
"epoch": 1.77,
|
| 2358 |
-
"grad_norm": 1.029404957630594,
|
| 2359 |
-
"learning_rate": 3.6658888554312967e-06,
|
| 2360 |
-
"loss": 0.7868,
|
| 2361 |
-
"step": 1675
|
| 2362 |
-
},
|
| 2363 |
-
{
|
| 2364 |
-
"epoch": 1.78,
|
| 2365 |
-
"grad_norm": 0.9921023940146521,
|
| 2366 |
-
"learning_rate": 3.6389634629760763e-06,
|
| 2367 |
-
"loss": 0.7555,
|
| 2368 |
-
"step": 1680
|
| 2369 |
-
},
|
| 2370 |
-
{
|
| 2371 |
-
"epoch": 1.78,
|
| 2372 |
-
"grad_norm": 1.017350986680598,
|
| 2373 |
-
"learning_rate": 3.612080630845096e-06,
|
| 2374 |
-
"loss": 0.7905,
|
| 2375 |
-
"step": 1685
|
| 2376 |
-
},
|
| 2377 |
-
{
|
| 2378 |
-
"epoch": 1.79,
|
| 2379 |
-
"grad_norm": 1.0430603602540587,
|
| 2380 |
-
"learning_rate": 3.5852411996785776e-06,
|
| 2381 |
-
"loss": 0.7947,
|
| 2382 |
-
"step": 1690
|
| 2383 |
-
},
|
| 2384 |
-
{
|
| 2385 |
-
"epoch": 1.79,
|
| 2386 |
-
"grad_norm": 0.9737056004061376,
|
| 2387 |
-
"learning_rate": 3.558446008759569e-06,
|
| 2388 |
-
"loss": 0.7789,
|
| 2389 |
-
"step": 1695
|
| 2390 |
-
},
|
| 2391 |
-
{
|
| 2392 |
-
"epoch": 1.8,
|
| 2393 |
-
"grad_norm": 1.0212119960635129,
|
| 2394 |
-
"learning_rate": 3.5316958959876985e-06,
|
| 2395 |
-
"loss": 0.7671,
|
| 2396 |
-
"step": 1700
|
| 2397 |
-
},
|
| 2398 |
-
{
|
| 2399 |
-
"epoch": 1.8,
|
| 2400 |
-
"grad_norm": 1.0072141418910243,
|
| 2401 |
-
"learning_rate": 3.504991697852983e-06,
|
| 2402 |
-
"loss": 0.7844,
|
| 2403 |
-
"step": 1705
|
| 2404 |
-
},
|
| 2405 |
-
{
|
| 2406 |
-
"epoch": 1.81,
|
| 2407 |
-
"grad_norm": 1.059809521658242,
|
| 2408 |
-
"learning_rate": 3.4783342494096627e-06,
|
| 2409 |
-
"loss": 0.7845,
|
| 2410 |
-
"step": 1710
|
| 2411 |
-
},
|
| 2412 |
-
{
|
| 2413 |
-
"epoch": 1.81,
|
| 2414 |
-
"grad_norm": 1.032182317108509,
|
| 2415 |
-
"learning_rate": 3.451724384250091e-06,
|
| 2416 |
-
"loss": 0.7792,
|
| 2417 |
-
"step": 1715
|
| 2418 |
-
},
|
| 2419 |
-
{
|
| 2420 |
-
"epoch": 1.82,
|
| 2421 |
-
"grad_norm": 0.9779053888998924,
|
| 2422 |
-
"learning_rate": 3.4251629344786675e-06,
|
| 2423 |
-
"loss": 0.7591,
|
| 2424 |
-
"step": 1720
|
| 2425 |
-
},
|
| 2426 |
-
{
|
| 2427 |
-
"epoch": 1.82,
|
| 2428 |
-
"grad_norm": 1.0116163318504925,
|
| 2429 |
-
"learning_rate": 3.398650730685813e-06,
|
| 2430 |
-
"loss": 0.7556,
|
| 2431 |
-
"step": 1725
|
| 2432 |
-
},
|
| 2433 |
-
{
|
| 2434 |
-
"epoch": 1.83,
|
| 2435 |
-
"grad_norm": 1.0511489470052602,
|
| 2436 |
-
"learning_rate": 3.372188601922006e-06,
|
| 2437 |
-
"loss": 0.7637,
|
| 2438 |
-
"step": 1730
|
| 2439 |
-
},
|
| 2440 |
-
{
|
| 2441 |
-
"epoch": 1.83,
|
| 2442 |
-
"grad_norm": 1.0172930500825146,
|
| 2443 |
-
"learning_rate": 3.3457773756718513e-06,
|
| 2444 |
-
"loss": 0.7696,
|
| 2445 |
-
"step": 1735
|
| 2446 |
-
},
|
| 2447 |
-
{
|
| 2448 |
-
"epoch": 1.84,
|
| 2449 |
-
"grad_norm": 1.039493994412079,
|
| 2450 |
-
"learning_rate": 3.3194178778282046e-06,
|
| 2451 |
-
"loss": 0.7931,
|
| 2452 |
-
"step": 1740
|
| 2453 |
-
},
|
| 2454 |
-
{
|
| 2455 |
-
"epoch": 1.84,
|
| 2456 |
-
"grad_norm": 1.033662637919394,
|
| 2457 |
-
"learning_rate": 3.293110932666349e-06,
|
| 2458 |
-
"loss": 0.7692,
|
| 2459 |
-
"step": 1745
|
| 2460 |
-
},
|
| 2461 |
-
{
|
| 2462 |
-
"epoch": 1.85,
|
| 2463 |
-
"grad_norm": 1.0584694868797393,
|
| 2464 |
-
"learning_rate": 3.2668573628182145e-06,
|
| 2465 |
-
"loss": 0.7792,
|
| 2466 |
-
"step": 1750
|
| 2467 |
-
},
|
| 2468 |
-
{
|
| 2469 |
-
"epoch": 1.85,
|
| 2470 |
-
"grad_norm": 0.994626270021195,
|
| 2471 |
-
"learning_rate": 3.2406579892466582e-06,
|
| 2472 |
-
"loss": 0.7682,
|
| 2473 |
-
"step": 1755
|
| 2474 |
-
},
|
| 2475 |
-
{
|
| 2476 |
-
"epoch": 1.86,
|
| 2477 |
-
"grad_norm": 0.9270237802993908,
|
| 2478 |
-
"learning_rate": 3.2145136312197943e-06,
|
| 2479 |
-
"loss": 0.7552,
|
| 2480 |
-
"step": 1760
|
| 2481 |
-
},
|
| 2482 |
-
{
|
| 2483 |
-
"epoch": 1.87,
|
| 2484 |
-
"grad_norm": 2.0595234604236357,
|
| 2485 |
-
"learning_rate": 3.18842510628537e-06,
|
| 2486 |
-
"loss": 0.7749,
|
| 2487 |
-
"step": 1765
|
| 2488 |
-
},
|
| 2489 |
-
{
|
| 2490 |
-
"epoch": 1.87,
|
| 2491 |
-
"grad_norm": 1.0396319816767299,
|
| 2492 |
-
"learning_rate": 3.162393230245203e-06,
|
| 2493 |
-
"loss": 0.804,
|
| 2494 |
-
"step": 1770
|
| 2495 |
-
},
|
| 2496 |
-
{
|
| 2497 |
-
"epoch": 1.88,
|
| 2498 |
-
"grad_norm": 1.0214462086054552,
|
| 2499 |
-
"learning_rate": 3.1364188171296677e-06,
|
| 2500 |
-
"loss": 0.7744,
|
| 2501 |
-
"step": 1775
|
| 2502 |
-
},
|
| 2503 |
-
{
|
| 2504 |
-
"epoch": 1.88,
|
| 2505 |
-
"grad_norm": 1.0145502545771508,
|
| 2506 |
-
"learning_rate": 3.110502679172246e-06,
|
| 2507 |
-
"loss": 0.7824,
|
| 2508 |
-
"step": 1780
|
| 2509 |
-
},
|
| 2510 |
-
{
|
| 2511 |
-
"epoch": 1.89,
|
| 2512 |
-
"grad_norm": 1.0196641711891408,
|
| 2513 |
-
"learning_rate": 3.084645626784124e-06,
|
| 2514 |
-
"loss": 0.7745,
|
| 2515 |
-
"step": 1785
|
| 2516 |
-
},
|
| 2517 |
-
{
|
| 2518 |
-
"epoch": 1.89,
|
| 2519 |
-
"grad_norm": 1.0197064636159427,
|
| 2520 |
-
"learning_rate": 3.058848468528852e-06,
|
| 2521 |
-
"loss": 0.8031,
|
| 2522 |
-
"step": 1790
|
| 2523 |
-
},
|
| 2524 |
-
{
|
| 2525 |
-
"epoch": 1.9,
|
| 2526 |
-
"grad_norm": 0.9907125667454302,
|
| 2527 |
-
"learning_rate": 3.03311201109706e-06,
|
| 2528 |
-
"loss": 0.7919,
|
| 2529 |
-
"step": 1795
|
| 2530 |
-
},
|
| 2531 |
-
{
|
| 2532 |
-
"epoch": 1.9,
|
| 2533 |
-
"grad_norm": 1.017942513059757,
|
| 2534 |
-
"learning_rate": 3.0074370592812286e-06,
|
| 2535 |
-
"loss": 0.7907,
|
| 2536 |
-
"step": 1800
|
| 2537 |
-
},
|
| 2538 |
-
{
|
| 2539 |
-
"epoch": 1.91,
|
| 2540 |
-
"grad_norm": 1.0821499695866912,
|
| 2541 |
-
"learning_rate": 2.9818244159505265e-06,
|
| 2542 |
-
"loss": 0.7901,
|
| 2543 |
-
"step": 1805
|
| 2544 |
-
},
|
| 2545 |
-
{
|
| 2546 |
-
"epoch": 1.91,
|
| 2547 |
-
"grad_norm": 0.9934394662674368,
|
| 2548 |
-
"learning_rate": 2.956274882025706e-06,
|
| 2549 |
-
"loss": 0.7638,
|
| 2550 |
-
"step": 1810
|
| 2551 |
-
},
|
| 2552 |
-
{
|
| 2553 |
-
"epoch": 1.92,
|
| 2554 |
-
"grad_norm": 1.0313411208961847,
|
| 2555 |
-
"learning_rate": 2.930789256454052e-06,
|
| 2556 |
-
"loss": 0.7553,
|
| 2557 |
-
"step": 1815
|
| 2558 |
-
},
|
| 2559 |
-
{
|
| 2560 |
-
"epoch": 1.92,
|
| 2561 |
-
"grad_norm": 0.9950833531614097,
|
| 2562 |
-
"learning_rate": 2.905368336184406e-06,
|
| 2563 |
-
"loss": 0.7576,
|
| 2564 |
-
"step": 1820
|
| 2565 |
-
},
|
| 2566 |
-
{
|
| 2567 |
-
"epoch": 1.93,
|
| 2568 |
-
"grad_norm": 0.9936896686220547,
|
| 2569 |
-
"learning_rate": 2.8800129161422365e-06,
|
| 2570 |
-
"loss": 0.7671,
|
| 2571 |
-
"step": 1825
|
| 2572 |
-
},
|
| 2573 |
-
{
|
| 2574 |
-
"epoch": 1.93,
|
| 2575 |
-
"grad_norm": 0.9909860465997411,
|
| 2576 |
-
"learning_rate": 2.8547237892047852e-06,
|
| 2577 |
-
"loss": 0.74,
|
| 2578 |
-
"step": 1830
|
| 2579 |
-
},
|
| 2580 |
-
{
|
| 2581 |
-
"epoch": 1.94,
|
| 2582 |
-
"grad_norm": 0.9788752840880554,
|
| 2583 |
-
"learning_rate": 2.8295017461762806e-06,
|
| 2584 |
-
"loss": 0.767,
|
| 2585 |
-
"step": 1835
|
| 2586 |
-
},
|
| 2587 |
-
{
|
| 2588 |
-
"epoch": 1.94,
|
| 2589 |
-
"grad_norm": 0.9764110020200104,
|
| 2590 |
-
"learning_rate": 2.804347575763193e-06,
|
| 2591 |
-
"loss": 0.7668,
|
| 2592 |
-
"step": 1840
|
| 2593 |
-
},
|
| 2594 |
-
{
|
| 2595 |
-
"epoch": 1.95,
|
| 2596 |
-
"grad_norm": 0.9772254707929505,
|
| 2597 |
-
"learning_rate": 2.7792620645495917e-06,
|
| 2598 |
-
"loss": 0.7425,
|
| 2599 |
-
"step": 1845
|
| 2600 |
-
},
|
| 2601 |
-
{
|
| 2602 |
-
"epoch": 1.96,
|
| 2603 |
-
"grad_norm": 1.0000854462976456,
|
| 2604 |
-
"learning_rate": 2.7542459969725215e-06,
|
| 2605 |
-
"loss": 0.7466,
|
| 2606 |
-
"step": 1850
|
| 2607 |
-
},
|
| 2608 |
-
{
|
| 2609 |
-
"epoch": 1.96,
|
| 2610 |
-
"grad_norm": 1.0352323998365711,
|
| 2611 |
-
"learning_rate": 2.729300155297504e-06,
|
| 2612 |
-
"loss": 0.771,
|
| 2613 |
-
"step": 1855
|
| 2614 |
-
},
|
| 2615 |
-
{
|
| 2616 |
-
"epoch": 1.97,
|
| 2617 |
-
"grad_norm": 0.9811051893834364,
|
| 2618 |
-
"learning_rate": 2.704425319594049e-06,
|
| 2619 |
-
"loss": 0.7778,
|
| 2620 |
-
"step": 1860
|
| 2621 |
-
},
|
| 2622 |
-
{
|
| 2623 |
-
"epoch": 1.97,
|
| 2624 |
-
"grad_norm": 1.0284677234046133,
|
| 2625 |
-
"learning_rate": 2.6796222677112825e-06,
|
| 2626 |
-
"loss": 0.7796,
|
| 2627 |
-
"step": 1865
|
| 2628 |
-
},
|
| 2629 |
-
{
|
| 2630 |
-
"epoch": 1.98,
|
| 2631 |
-
"grad_norm": 0.9664217044137716,
|
| 2632 |
-
"learning_rate": 2.6548917752535997e-06,
|
| 2633 |
-
"loss": 0.771,
|
| 2634 |
-
"step": 1870
|
| 2635 |
-
},
|
| 2636 |
-
{
|
| 2637 |
-
"epoch": 1.98,
|
| 2638 |
-
"grad_norm": 1.0008524753186703,
|
| 2639 |
-
"learning_rate": 2.6302346155564385e-06,
|
| 2640 |
-
"loss": 0.7963,
|
| 2641 |
-
"step": 1875
|
| 2642 |
-
},
|
| 2643 |
-
{
|
| 2644 |
-
"epoch": 1.99,
|
| 2645 |
-
"grad_norm": 1.0088045948631796,
|
| 2646 |
-
"learning_rate": 2.6056515596620715e-06,
|
| 2647 |
-
"loss": 0.7571,
|
| 2648 |
-
"step": 1880
|
| 2649 |
-
},
|
| 2650 |
-
{
|
| 2651 |
-
"epoch": 1.99,
|
| 2652 |
-
"grad_norm": 0.9727997698934588,
|
| 2653 |
-
"learning_rate": 2.581143376295516e-06,
|
| 2654 |
-
"loss": 0.7968,
|
| 2655 |
-
"step": 1885
|
| 2656 |
-
},
|
| 2657 |
-
{
|
| 2658 |
-
"epoch": 2.0,
|
| 2659 |
-
"grad_norm": 0.9760428822299934,
|
| 2660 |
-
"learning_rate": 2.556710831840481e-06,
|
| 2661 |
-
"loss": 0.7829,
|
| 2662 |
-
"step": 1890
|
| 2663 |
-
},
|
| 2664 |
-
{
|
| 2665 |
-
"epoch": 2.0,
|
| 2666 |
-
"grad_norm": 1.1893585643467264,
|
| 2667 |
-
"learning_rate": 2.5323546903154074e-06,
|
| 2668 |
-
"loss": 0.7363,
|
| 2669 |
-
"step": 1895
|
| 2670 |
-
},
|
| 2671 |
-
{
|
| 2672 |
-
"epoch": 2.01,
|
| 2673 |
-
"grad_norm": 1.0408498899558132,
|
| 2674 |
-
"learning_rate": 2.508075713349575e-06,
|
| 2675 |
-
"loss": 0.683,
|
| 2676 |
-
"step": 1900
|
| 2677 |
-
},
|
| 2678 |
-
{
|
| 2679 |
-
"epoch": 2.01,
|
| 2680 |
-
"grad_norm": 1.0852218097728863,
|
| 2681 |
-
"learning_rate": 2.483874660159294e-06,
|
| 2682 |
-
"loss": 0.6388,
|
| 2683 |
-
"step": 1905
|
| 2684 |
-
},
|
| 2685 |
-
{
|
| 2686 |
-
"epoch": 2.02,
|
| 2687 |
-
"grad_norm": 1.0636193658435114,
|
| 2688 |
-
"learning_rate": 2.45975228752415e-06,
|
| 2689 |
-
"loss": 0.6785,
|
| 2690 |
-
"step": 1910
|
| 2691 |
-
},
|
| 2692 |
-
{
|
| 2693 |
-
"epoch": 2.02,
|
| 2694 |
-
"grad_norm": 1.05164052954354,
|
| 2695 |
-
"learning_rate": 2.435709349763354e-06,
|
| 2696 |
-
"loss": 0.7024,
|
| 2697 |
-
"step": 1915
|
| 2698 |
-
},
|
| 2699 |
-
{
|
| 2700 |
-
"epoch": 2.03,
|
| 2701 |
-
"grad_norm": 1.0744751292672923,
|
| 2702 |
-
"learning_rate": 2.4117465987121357e-06,
|
| 2703 |
-
"loss": 0.6714,
|
| 2704 |
-
"step": 1920
|
| 2705 |
-
},
|
| 2706 |
-
{
|
| 2707 |
-
"epoch": 2.03,
|
| 2708 |
-
"grad_norm": 1.0221167769747221,
|
| 2709 |
-
"learning_rate": 2.387864783698258e-06,
|
| 2710 |
-
"loss": 0.6441,
|
| 2711 |
-
"step": 1925
|
| 2712 |
-
},
|
| 2713 |
-
{
|
| 2714 |
-
"epoch": 2.04,
|
| 2715 |
-
"grad_norm": 1.0453109653021675,
|
| 2716 |
-
"learning_rate": 2.3640646515185596e-06,
|
| 2717 |
-
"loss": 0.6668,
|
| 2718 |
-
"step": 1930
|
| 2719 |
-
},
|
| 2720 |
-
{
|
| 2721 |
-
"epoch": 2.04,
|
| 2722 |
-
"grad_norm": 1.0035196656143317,
|
| 2723 |
-
"learning_rate": 2.3403469464156235e-06,
|
| 2724 |
-
"loss": 0.6711,
|
| 2725 |
-
"step": 1935
|
| 2726 |
-
},
|
| 2727 |
-
{
|
| 2728 |
-
"epoch": 2.05,
|
| 2729 |
-
"grad_norm": 1.0614923887712562,
|
| 2730 |
-
"learning_rate": 2.31671241005449e-06,
|
| 2731 |
-
"loss": 0.6801,
|
| 2732 |
-
"step": 1940
|
| 2733 |
-
},
|
| 2734 |
-
{
|
| 2735 |
-
"epoch": 2.06,
|
| 2736 |
-
"grad_norm": 1.0457688195463548,
|
| 2737 |
-
"learning_rate": 2.2931617814994704e-06,
|
| 2738 |
-
"loss": 0.6676,
|
| 2739 |
-
"step": 1945
|
| 2740 |
-
},
|
| 2741 |
-
{
|
| 2742 |
-
"epoch": 2.06,
|
| 2743 |
-
"grad_norm": 1.094973586743587,
|
| 2744 |
-
"learning_rate": 2.269695797191032e-06,
|
| 2745 |
-
"loss": 0.6467,
|
| 2746 |
-
"step": 1950
|
| 2747 |
-
},
|
| 2748 |
-
{
|
| 2749 |
-
"epoch": 2.07,
|
| 2750 |
-
"grad_norm": 1.0312304548353073,
|
| 2751 |
-
"learning_rate": 2.2463151909227804e-06,
|
| 2752 |
-
"loss": 0.6626,
|
| 2753 |
-
"step": 1955
|
| 2754 |
-
},
|
| 2755 |
-
{
|
| 2756 |
-
"epoch": 2.07,
|
| 2757 |
-
"grad_norm": 1.0435526510546405,
|
| 2758 |
-
"learning_rate": 2.223020693818495e-06,
|
| 2759 |
-
"loss": 0.6565,
|
| 2760 |
-
"step": 1960
|
| 2761 |
-
},
|
| 2762 |
-
{
|
| 2763 |
-
"epoch": 2.08,
|
| 2764 |
-
"grad_norm": 1.0361388218534178,
|
| 2765 |
-
"learning_rate": 2.1998130343092866e-06,
|
| 2766 |
-
"loss": 0.655,
|
| 2767 |
-
"step": 1965
|
| 2768 |
-
},
|
| 2769 |
-
{
|
| 2770 |
-
"epoch": 2.08,
|
| 2771 |
-
"grad_norm": 1.071971382261616,
|
| 2772 |
-
"learning_rate": 2.176692938110801e-06,
|
| 2773 |
-
"loss": 0.6628,
|
| 2774 |
-
"step": 1970
|
| 2775 |
-
},
|
| 2776 |
-
{
|
| 2777 |
-
"epoch": 2.09,
|
| 2778 |
-
"grad_norm": 1.0449189624346316,
|
| 2779 |
-
"learning_rate": 2.1536611282005374e-06,
|
| 2780 |
-
"loss": 0.6742,
|
| 2781 |
-
"step": 1975
|
| 2782 |
-
},
|
| 2783 |
-
{
|
| 2784 |
-
"epoch": 2.09,
|
| 2785 |
-
"grad_norm": 1.0076278447431801,
|
| 2786 |
-
"learning_rate": 2.130718324795234e-06,
|
| 2787 |
-
"loss": 0.6615,
|
| 2788 |
-
"step": 1980
|
| 2789 |
-
},
|
| 2790 |
-
{
|
| 2791 |
-
"epoch": 2.1,
|
| 2792 |
-
"grad_norm": 1.044357139317297,
|
| 2793 |
-
"learning_rate": 2.107865245328354e-06,
|
| 2794 |
-
"loss": 0.6707,
|
| 2795 |
-
"step": 1985
|
| 2796 |
-
},
|
| 2797 |
-
{
|
| 2798 |
-
"epoch": 2.1,
|
| 2799 |
-
"grad_norm": 1.0155250644507565,
|
| 2800 |
-
"learning_rate": 2.0851026044276405e-06,
|
| 2801 |
-
"loss": 0.6701,
|
| 2802 |
-
"step": 1990
|
| 2803 |
-
},
|
| 2804 |
-
{
|
| 2805 |
-
"epoch": 2.11,
|
| 2806 |
-
"grad_norm": 1.012020172763002,
|
| 2807 |
-
"learning_rate": 2.0624311138927795e-06,
|
| 2808 |
-
"loss": 0.6531,
|
| 2809 |
-
"step": 1995
|
| 2810 |
-
},
|
| 2811 |
-
{
|
| 2812 |
-
"epoch": 2.11,
|
| 2813 |
-
"grad_norm": 1.0209851165233697,
|
| 2814 |
-
"learning_rate": 2.0398514826731326e-06,
|
| 2815 |
-
"loss": 0.6685,
|
| 2816 |
-
"step": 2000
|
| 2817 |
-
},
|
| 2818 |
-
{
|
| 2819 |
-
"epoch": 2.12,
|
| 2820 |
-
"grad_norm": 1.0147123852944229,
|
| 2821 |
-
"learning_rate": 2.017364416845579e-06,
|
| 2822 |
-
"loss": 0.6506,
|
| 2823 |
-
"step": 2005
|
| 2824 |
-
},
|
| 2825 |
-
{
|
| 2826 |
-
"epoch": 2.12,
|
| 2827 |
-
"grad_norm": 1.06994559921509,
|
| 2828 |
-
"learning_rate": 1.9949706195924235e-06,
|
| 2829 |
-
"loss": 0.6743,
|
| 2830 |
-
"step": 2010
|
| 2831 |
-
},
|
| 2832 |
-
{
|
| 2833 |
-
"epoch": 2.13,
|
| 2834 |
-
"grad_norm": 0.9930487524595831,
|
| 2835 |
-
"learning_rate": 1.97267079117942e-06,
|
| 2836 |
-
"loss": 0.6596,
|
| 2837 |
-
"step": 2015
|
| 2838 |
-
},
|
| 2839 |
-
{
|
| 2840 |
-
"epoch": 2.13,
|
| 2841 |
-
"grad_norm": 1.0334858708046972,
|
| 2842 |
-
"learning_rate": 1.950465628933863e-06,
|
| 2843 |
-
"loss": 0.6679,
|
| 2844 |
-
"step": 2020
|
| 2845 |
-
},
|
| 2846 |
-
{
|
| 2847 |
-
"epoch": 2.14,
|
| 2848 |
-
"grad_norm": 1.060064879245556,
|
| 2849 |
-
"learning_rate": 1.9283558272227866e-06,
|
| 2850 |
-
"loss": 0.6749,
|
| 2851 |
-
"step": 2025
|
| 2852 |
-
},
|
| 2853 |
-
{
|
| 2854 |
-
"epoch": 2.15,
|
| 2855 |
-
"grad_norm": 1.0171368650427,
|
| 2856 |
-
"learning_rate": 1.9063420774312509e-06,
|
| 2857 |
-
"loss": 0.6703,
|
| 2858 |
-
"step": 2030
|
| 2859 |
-
},
|
| 2860 |
-
{
|
| 2861 |
-
"epoch": 2.15,
|
| 2862 |
-
"grad_norm": 0.9646165360014197,
|
| 2863 |
-
"learning_rate": 1.8844250679407272e-06,
|
| 2864 |
-
"loss": 0.6878,
|
| 2865 |
-
"step": 2035
|
| 2866 |
-
},
|
| 2867 |
-
{
|
| 2868 |
-
"epoch": 2.16,
|
| 2869 |
-
"grad_norm": 1.0209055430674492,
|
| 2870 |
-
"learning_rate": 1.862605484107562e-06,
|
| 2871 |
-
"loss": 0.7052,
|
| 2872 |
-
"step": 2040
|
| 2873 |
-
},
|
| 2874 |
-
{
|
| 2875 |
-
"epoch": 2.16,
|
| 2876 |
-
"grad_norm": 1.0216869737250995,
|
| 2877 |
-
"learning_rate": 1.840884008241549e-06,
|
| 2878 |
-
"loss": 0.6778,
|
| 2879 |
-
"step": 2045
|
| 2880 |
-
},
|
| 2881 |
-
{
|
| 2882 |
-
"epoch": 2.17,
|
| 2883 |
-
"grad_norm": 0.990030094537176,
|
| 2884 |
-
"learning_rate": 1.819261319584602e-06,
|
| 2885 |
-
"loss": 0.675,
|
| 2886 |
-
"step": 2050
|
| 2887 |
-
},
|
| 2888 |
-
{
|
| 2889 |
-
"epoch": 2.17,
|
| 2890 |
-
"grad_norm": 0.9972968188321764,
|
| 2891 |
-
"learning_rate": 1.7977380942895007e-06,
|
| 2892 |
-
"loss": 0.6832,
|
| 2893 |
-
"step": 2055
|
| 2894 |
-
},
|
| 2895 |
-
{
|
| 2896 |
-
"epoch": 2.18,
|
| 2897 |
-
"grad_norm": 1.002919858574642,
|
| 2898 |
-
"learning_rate": 1.7763150053987532e-06,
|
| 2899 |
-
"loss": 0.6669,
|
| 2900 |
-
"step": 2060
|
| 2901 |
-
},
|
| 2902 |
-
{
|
| 2903 |
-
"epoch": 2.18,
|
| 2904 |
-
"grad_norm": 1.040641077805689,
|
| 2905 |
-
"learning_rate": 1.7549927228235547e-06,
|
| 2906 |
-
"loss": 0.6874,
|
| 2907 |
-
"step": 2065
|
| 2908 |
-
},
|
| 2909 |
-
{
|
| 2910 |
-
"epoch": 2.19,
|
| 2911 |
-
"grad_norm": 1.0136593089712416,
|
| 2912 |
-
"learning_rate": 1.7337719133228308e-06,
|
| 2913 |
-
"loss": 0.6662,
|
| 2914 |
-
"step": 2070
|
| 2915 |
-
},
|
| 2916 |
-
{
|
| 2917 |
-
"epoch": 2.19,
|
| 2918 |
-
"grad_norm": 1.0032381970613455,
|
| 2919 |
-
"learning_rate": 1.7126532404823898e-06,
|
| 2920 |
-
"loss": 0.657,
|
| 2921 |
-
"step": 2075
|
| 2922 |
-
},
|
| 2923 |
-
{
|
| 2924 |
-
"epoch": 2.2,
|
| 2925 |
-
"grad_norm": 1.0107311218156156,
|
| 2926 |
-
"learning_rate": 1.6916373646941774e-06,
|
| 2927 |
-
"loss": 0.6706,
|
| 2928 |
-
"step": 2080
|
| 2929 |
-
},
|
| 2930 |
-
{
|
| 2931 |
-
"epoch": 2.2,
|
| 2932 |
-
"grad_norm": 1.0313882769598175,
|
| 2933 |
-
"learning_rate": 1.6707249431356188e-06,
|
| 2934 |
-
"loss": 0.6803,
|
| 2935 |
-
"step": 2085
|
| 2936 |
-
},
|
| 2937 |
-
{
|
| 2938 |
-
"epoch": 2.21,
|
| 2939 |
-
"grad_norm": 1.0013867402651844,
|
| 2940 |
-
"learning_rate": 1.6499166297490716e-06,
|
| 2941 |
-
"loss": 0.6896,
|
| 2942 |
-
"step": 2090
|
| 2943 |
-
},
|
| 2944 |
-
{
|
| 2945 |
-
"epoch": 2.21,
|
| 2946 |
-
"grad_norm": 0.9974367112606389,
|
| 2947 |
-
"learning_rate": 1.6292130752213747e-06,
|
| 2948 |
-
"loss": 0.6773,
|
| 2949 |
-
"step": 2095
|
| 2950 |
-
},
|
| 2951 |
-
{
|
| 2952 |
-
"epoch": 2.22,
|
| 2953 |
-
"grad_norm": 1.0457782650116,
|
| 2954 |
-
"learning_rate": 1.6086149269635081e-06,
|
| 2955 |
-
"loss": 0.668,
|
| 2956 |
-
"step": 2100
|
| 2957 |
-
},
|
| 2958 |
-
{
|
| 2959 |
-
"epoch": 2.22,
|
| 2960 |
-
"grad_norm": 0.9930241935385495,
|
| 2961 |
-
"learning_rate": 1.5881228290903367e-06,
|
| 2962 |
-
"loss": 0.6508,
|
| 2963 |
-
"step": 2105
|
| 2964 |
-
},
|
| 2965 |
-
{
|
| 2966 |
-
"epoch": 2.23,
|
| 2967 |
-
"grad_norm": 1.0059354322817335,
|
| 2968 |
-
"learning_rate": 1.5677374224004793e-06,
|
| 2969 |
-
"loss": 0.6529,
|
| 2970 |
-
"step": 2110
|
| 2971 |
-
},
|
| 2972 |
-
{
|
| 2973 |
-
"epoch": 2.24,
|
| 2974 |
-
"grad_norm": 1.0338579100235163,
|
| 2975 |
-
"learning_rate": 1.547459344356262e-06,
|
| 2976 |
-
"loss": 0.6614,
|
| 2977 |
-
"step": 2115
|
| 2978 |
-
},
|
| 2979 |
-
{
|
| 2980 |
-
"epoch": 2.24,
|
| 2981 |
-
"grad_norm": 1.0203126239591027,
|
| 2982 |
-
"learning_rate": 1.5272892290637892e-06,
|
| 2983 |
-
"loss": 0.6749,
|
| 2984 |
-
"step": 2120
|
| 2985 |
-
},
|
| 2986 |
-
{
|
| 2987 |
-
"epoch": 2.25,
|
| 2988 |
-
"grad_norm": 0.983643586611109,
|
| 2989 |
-
"learning_rate": 1.5072277072531127e-06,
|
| 2990 |
-
"loss": 0.6517,
|
| 2991 |
-
"step": 2125
|
| 2992 |
-
},
|
| 2993 |
-
{
|
| 2994 |
-
"epoch": 2.25,
|
| 2995 |
-
"grad_norm": 1.0203957676102433,
|
| 2996 |
-
"learning_rate": 1.4872754062585126e-06,
|
| 2997 |
-
"loss": 0.6716,
|
| 2998 |
-
"step": 2130
|
| 2999 |
-
},
|
| 3000 |
-
{
|
| 3001 |
-
"epoch": 2.26,
|
| 3002 |
-
"grad_norm": 1.036201909144992,
|
| 3003 |
-
"learning_rate": 1.4674329499988737e-06,
|
| 3004 |
-
"loss": 0.6574,
|
| 3005 |
-
"step": 2135
|
| 3006 |
-
},
|
| 3007 |
-
{
|
| 3008 |
-
"epoch": 2.26,
|
| 3009 |
-
"grad_norm": 1.0277085537623492,
|
| 3010 |
-
"learning_rate": 1.4477009589581787e-06,
|
| 3011 |
-
"loss": 0.6593,
|
| 3012 |
-
"step": 2140
|
| 3013 |
-
},
|
| 3014 |
-
{
|
| 3015 |
-
"epoch": 2.27,
|
| 3016 |
-
"grad_norm": 0.9713425669443266,
|
| 3017 |
-
"learning_rate": 1.4280800501661057e-06,
|
| 3018 |
-
"loss": 0.6621,
|
| 3019 |
-
"step": 2145
|
| 3020 |
-
},
|
| 3021 |
-
{
|
| 3022 |
-
"epoch": 2.27,
|
| 3023 |
-
"grad_norm": 1.028497947768737,
|
| 3024 |
-
"learning_rate": 1.408570837178735e-06,
|
| 3025 |
-
"loss": 0.6656,
|
| 3026 |
-
"step": 2150
|
| 3027 |
-
},
|
| 3028 |
-
{
|
| 3029 |
-
"epoch": 2.28,
|
| 3030 |
-
"grad_norm": 1.0565632370972053,
|
| 3031 |
-
"learning_rate": 1.3891739300593559e-06,
|
| 3032 |
-
"loss": 0.6644,
|
| 3033 |
-
"step": 2155
|
| 3034 |
-
},
|
| 3035 |
-
{
|
| 3036 |
-
"epoch": 2.28,
|
| 3037 |
-
"grad_norm": 1.0043346444991121,
|
| 3038 |
-
"learning_rate": 1.369889935359402e-06,
|
| 3039 |
-
"loss": 0.6539,
|
| 3040 |
-
"step": 2160
|
| 3041 |
-
},
|
| 3042 |
-
{
|
| 3043 |
-
"epoch": 2.29,
|
| 3044 |
-
"grad_norm": 1.0294689299797029,
|
| 3045 |
-
"learning_rate": 1.3507194560994657e-06,
|
| 3046 |
-
"loss": 0.6666,
|
| 3047 |
-
"step": 2165
|
| 3048 |
-
},
|
| 3049 |
-
{
|
| 3050 |
-
"epoch": 2.29,
|
| 3051 |
-
"grad_norm": 1.0123495429792864,
|
| 3052 |
-
"learning_rate": 1.331663091750463e-06,
|
| 3053 |
-
"loss": 0.6928,
|
| 3054 |
-
"step": 2170
|
| 3055 |
-
},
|
| 3056 |
-
{
|
| 3057 |
-
"epoch": 2.3,
|
| 3058 |
-
"grad_norm": 0.9951164224382856,
|
| 3059 |
-
"learning_rate": 1.312721438214869e-06,
|
| 3060 |
-
"loss": 0.6501,
|
| 3061 |
-
"step": 2175
|
| 3062 |
-
},
|
| 3063 |
-
{
|
| 3064 |
-
"epoch": 2.3,
|
| 3065 |
-
"grad_norm": 1.025832661356824,
|
| 3066 |
-
"learning_rate": 1.293895087808098e-06,
|
| 3067 |
-
"loss": 0.6658,
|
| 3068 |
-
"step": 2180
|
| 3069 |
-
},
|
| 3070 |
-
{
|
| 3071 |
-
"epoch": 2.31,
|
| 3072 |
-
"grad_norm": 0.9888366700648139,
|
| 3073 |
-
"learning_rate": 1.2751846292399705e-06,
|
| 3074 |
-
"loss": 0.6592,
|
| 3075 |
-
"step": 2185
|
| 3076 |
-
},
|
| 3077 |
-
{
|
| 3078 |
-
"epoch": 2.31,
|
| 3079 |
-
"grad_norm": 1.0208359350524125,
|
| 3080 |
-
"learning_rate": 1.2565906475963102e-06,
|
| 3081 |
-
"loss": 0.6483,
|
| 3082 |
-
"step": 2190
|
| 3083 |
-
},
|
| 3084 |
-
{
|
| 3085 |
-
"epoch": 2.32,
|
| 3086 |
-
"grad_norm": 1.0568986951058392,
|
| 3087 |
-
"learning_rate": 1.2381137243206455e-06,
|
| 3088 |
-
"loss": 0.6557,
|
| 3089 |
-
"step": 2195
|
| 3090 |
-
},
|
| 3091 |
-
{
|
| 3092 |
-
"epoch": 2.32,
|
| 3093 |
-
"grad_norm": 0.9849389521844061,
|
| 3094 |
-
"learning_rate": 1.2197544371960317e-06,
|
| 3095 |
-
"loss": 0.6488,
|
| 3096 |
-
"step": 2200
|
| 3097 |
-
},
|
| 3098 |
-
{
|
| 3099 |
-
"epoch": 2.33,
|
| 3100 |
-
"grad_norm": 1.0466426799607875,
|
| 3101 |
-
"learning_rate": 1.2015133603269753e-06,
|
| 3102 |
-
"loss": 0.6596,
|
| 3103 |
-
"step": 2205
|
| 3104 |
-
},
|
| 3105 |
-
{
|
| 3106 |
-
"epoch": 2.34,
|
| 3107 |
-
"grad_norm": 0.9985742048846067,
|
| 3108 |
-
"learning_rate": 1.183391064121493e-06,
|
| 3109 |
-
"loss": 0.6572,
|
| 3110 |
-
"step": 2210
|
| 3111 |
-
},
|
| 3112 |
-
{
|
| 3113 |
-
"epoch": 2.34,
|
| 3114 |
-
"grad_norm": 0.9661312369342807,
|
| 3115 |
-
"learning_rate": 1.1653881152732582e-06,
|
| 3116 |
-
"loss": 0.6439,
|
| 3117 |
-
"step": 2215
|
| 3118 |
-
},
|
| 3119 |
-
{
|
| 3120 |
-
"epoch": 2.35,
|
| 3121 |
-
"grad_norm": 1.0327058718249167,
|
| 3122 |
-
"learning_rate": 1.1475050767439e-06,
|
| 3123 |
-
"loss": 0.6811,
|
| 3124 |
-
"step": 2220
|
| 3125 |
-
},
|
| 3126 |
-
{
|
| 3127 |
-
"epoch": 2.35,
|
| 3128 |
-
"grad_norm": 1.0365200638536969,
|
| 3129 |
-
"learning_rate": 1.129742507745382e-06,
|
| 3130 |
-
"loss": 0.6588,
|
| 3131 |
-
"step": 2225
|
| 3132 |
-
},
|
| 3133 |
-
{
|
| 3134 |
-
"epoch": 2.36,
|
| 3135 |
-
"grad_norm": 0.9804079029045045,
|
| 3136 |
-
"learning_rate": 1.1121009637225283e-06,
|
| 3137 |
-
"loss": 0.6783,
|
| 3138 |
-
"step": 2230
|
| 3139 |
-
},
|
| 3140 |
-
{
|
| 3141 |
-
"epoch": 2.36,
|
| 3142 |
-
"grad_norm": 1.0326866018136251,
|
| 3143 |
-
"learning_rate": 1.0945809963356442e-06,
|
| 3144 |
-
"loss": 0.6705,
|
| 3145 |
-
"step": 2235
|
| 3146 |
-
},
|
| 3147 |
-
{
|
| 3148 |
-
"epoch": 2.37,
|
| 3149 |
-
"grad_norm": 1.0314679157662048,
|
| 3150 |
-
"learning_rate": 1.0771831534432714e-06,
|
| 3151 |
-
"loss": 0.6353,
|
| 3152 |
-
"step": 2240
|
| 3153 |
-
},
|
| 3154 |
-
{
|
| 3155 |
-
"epoch": 2.37,
|
| 3156 |
-
"grad_norm": 0.9589889108924486,
|
| 3157 |
-
"learning_rate": 1.0599079790850542e-06,
|
| 3158 |
-
"loss": 0.655,
|
| 3159 |
-
"step": 2245
|
| 3160 |
-
},
|
| 3161 |
-
{
|
| 3162 |
-
"epoch": 2.38,
|
| 3163 |
-
"grad_norm": 0.9894914192305704,
|
| 3164 |
-
"learning_rate": 1.0427560134647308e-06,
|
| 3165 |
-
"loss": 0.643,
|
| 3166 |
-
"step": 2250
|
| 3167 |
-
},
|
| 3168 |
-
{
|
| 3169 |
-
"epoch": 2.38,
|
| 3170 |
-
"grad_norm": 1.0693419775513076,
|
| 3171 |
-
"learning_rate": 1.0257277929332332e-06,
|
| 3172 |
-
"loss": 0.6611,
|
| 3173 |
-
"step": 2255
|
| 3174 |
-
},
|
| 3175 |
-
{
|
| 3176 |
-
"epoch": 2.39,
|
| 3177 |
-
"grad_norm": 0.9951590219864285,
|
| 3178 |
-
"learning_rate": 1.0088238499719254e-06,
|
| 3179 |
-
"loss": 0.6403,
|
| 3180 |
-
"step": 2260
|
| 3181 |
-
},
|
| 3182 |
-
{
|
| 3183 |
-
"epoch": 2.39,
|
| 3184 |
-
"grad_norm": 1.0105626202971048,
|
| 3185 |
-
"learning_rate": 9.920447131759392e-07,
|
| 3186 |
-
"loss": 0.6707,
|
| 3187 |
-
"step": 2265
|
| 3188 |
-
},
|
| 3189 |
-
{
|
| 3190 |
-
"epoch": 2.4,
|
| 3191 |
-
"grad_norm": 1.0186289750333066,
|
| 3192 |
-
"learning_rate": 9.753909072376594e-07,
|
| 3193 |
-
"loss": 0.6809,
|
| 3194 |
-
"step": 2270
|
| 3195 |
-
},
|
| 3196 |
-
{
|
| 3197 |
-
"epoch": 2.4,
|
| 3198 |
-
"grad_norm": 1.0267980845318398,
|
| 3199 |
-
"learning_rate": 9.58862952930304e-07,
|
| 3200 |
-
"loss": 0.6642,
|
| 3201 |
-
"step": 2275
|
| 3202 |
-
},
|
| 3203 |
-
{
|
| 3204 |
-
"epoch": 2.41,
|
| 3205 |
-
"grad_norm": 1.0314667402705489,
|
| 3206 |
-
"learning_rate": 9.424613670916499e-07,
|
| 3207 |
-
"loss": 0.6815,
|
| 3208 |
-
"step": 2280
|
| 3209 |
-
},
|
| 3210 |
-
{
|
| 3211 |
-
"epoch": 2.41,
|
| 3212 |
-
"grad_norm": 0.9818510396592551,
|
| 3213 |
-
"learning_rate": 9.261866626078625e-07,
|
| 3214 |
-
"loss": 0.6579,
|
| 3215 |
-
"step": 2285
|
| 3216 |
-
},
|
| 3217 |
-
{
|
| 3218 |
-
"epoch": 2.42,
|
| 3219 |
-
"grad_norm": 0.998040916561116,
|
| 3220 |
-
"learning_rate": 9.100393483974612e-07,
|
| 3221 |
-
"loss": 0.6815,
|
| 3222 |
-
"step": 2290
|
| 3223 |
-
},
|
| 3224 |
-
{
|
| 3225 |
-
"epoch": 2.43,
|
| 3226 |
-
"grad_norm": 1.007529165875462,
|
| 3227 |
-
"learning_rate": 8.940199293954033e-07,
|
| 3228 |
-
"loss": 0.6609,
|
| 3229 |
-
"step": 2295
|
| 3230 |
-
},
|
| 3231 |
-
{
|
| 3232 |
-
"epoch": 2.43,
|
| 3233 |
-
"grad_norm": 1.0489165413908048,
|
| 3234 |
-
"learning_rate": 8.781289065373016e-07,
|
| 3235 |
-
"loss": 0.6661,
|
| 3236 |
-
"step": 2300
|
| 3237 |
-
},
|
| 3238 |
-
{
|
| 3239 |
-
"epoch": 2.44,
|
| 3240 |
-
"grad_norm": 1.0586483881635766,
|
| 3241 |
-
"learning_rate": 8.623667767437483e-07,
|
| 3242 |
-
"loss": 0.6494,
|
| 3243 |
-
"step": 2305
|
| 3244 |
-
},
|
| 3245 |
-
{
|
| 3246 |
-
"epoch": 2.44,
|
| 3247 |
-
"grad_norm": 0.970861929985865,
|
| 3248 |
-
"learning_rate": 8.467340329047874e-07,
|
| 3249 |
-
"loss": 0.6403,
|
| 3250 |
-
"step": 2310
|
| 3251 |
-
},
|
| 3252 |
-
{
|
| 3253 |
-
"epoch": 2.45,
|
| 3254 |
-
"grad_norm": 1.0315170437890622,
|
| 3255 |
-
"learning_rate": 8.312311638644888e-07,
|
| 3256 |
-
"loss": 0.6802,
|
| 3257 |
-
"step": 2315
|
| 3258 |
-
},
|
| 3259 |
-
{
|
| 3260 |
-
"epoch": 2.45,
|
| 3261 |
-
"grad_norm": 1.018615901485097,
|
| 3262 |
-
"learning_rate": 8.158586544056791e-07,
|
| 3263 |
-
"loss": 0.6813,
|
| 3264 |
-
"step": 2320
|
| 3265 |
-
},
|
| 3266 |
-
{
|
| 3267 |
-
"epoch": 2.46,
|
| 3268 |
-
"grad_norm": 0.9991739019084611,
|
| 3269 |
-
"learning_rate": 8.00616985234764e-07,
|
| 3270 |
-
"loss": 0.6757,
|
| 3271 |
-
"step": 2325
|
| 3272 |
-
},
|
| 3273 |
-
{
|
| 3274 |
-
"epoch": 2.46,
|
| 3275 |
-
"grad_norm": 1.039226698329409,
|
| 3276 |
-
"learning_rate": 7.855066329667121e-07,
|
| 3277 |
-
"loss": 0.6421,
|
| 3278 |
-
"step": 2330
|
| 3279 |
-
},
|
| 3280 |
-
{
|
| 3281 |
-
"epoch": 2.47,
|
| 3282 |
-
"grad_norm": 1.0505394427255816,
|
| 3283 |
-
"learning_rate": 7.705280701101392e-07,
|
| 3284 |
-
"loss": 0.6655,
|
| 3285 |
-
"step": 2335
|
| 3286 |
-
},
|
| 3287 |
-
{
|
| 3288 |
-
"epoch": 2.47,
|
| 3289 |
-
"grad_norm": 0.9750027460632938,
|
| 3290 |
-
"learning_rate": 7.556817650525383e-07,
|
| 3291 |
-
"loss": 0.6526,
|
| 3292 |
-
"step": 2340
|
| 3293 |
-
},
|
| 3294 |
-
{
|
| 3295 |
-
"epoch": 2.48,
|
| 3296 |
-
"grad_norm": 0.989246982143368,
|
| 3297 |
-
"learning_rate": 7.409681820456315e-07,
|
| 3298 |
-
"loss": 0.667,
|
| 3299 |
-
"step": 2345
|
| 3300 |
-
},
|
| 3301 |
-
{
|
| 3302 |
-
"epoch": 2.48,
|
| 3303 |
-
"grad_norm": 0.9977414734019189,
|
| 3304 |
-
"learning_rate": 7.263877811908553e-07,
|
| 3305 |
-
"loss": 0.6647,
|
| 3306 |
-
"step": 2350
|
| 3307 |
-
},
|
| 3308 |
-
{
|
| 3309 |
-
"epoch": 2.49,
|
| 3310 |
-
"grad_norm": 0.9875292562685886,
|
| 3311 |
-
"learning_rate": 7.11941018424967e-07,
|
| 3312 |
-
"loss": 0.667,
|
| 3313 |
-
"step": 2355
|
| 3314 |
-
},
|
| 3315 |
-
{
|
| 3316 |
-
"epoch": 2.49,
|
| 3317 |
-
"grad_norm": 0.9932801930288735,
|
| 3318 |
-
"learning_rate": 6.97628345505797e-07,
|
| 3319 |
-
"loss": 0.6511,
|
| 3320 |
-
"step": 2360
|
| 3321 |
-
},
|
| 3322 |
-
{
|
| 3323 |
-
"epoch": 2.5,
|
| 3324 |
-
"grad_norm": 1.0199295886729471,
|
| 3325 |
-
"learning_rate": 6.83450209998106e-07,
|
| 3326 |
-
"loss": 0.6556,
|
| 3327 |
-
"step": 2365
|
| 3328 |
-
},
|
| 3329 |
-
{
|
| 3330 |
-
"epoch": 2.5,
|
| 3331 |
-
"grad_norm": 1.0279710885988984,
|
| 3332 |
-
"learning_rate": 6.694070552596105e-07,
|
| 3333 |
-
"loss": 0.6676,
|
| 3334 |
-
"step": 2370
|
| 3335 |
-
},
|
| 3336 |
-
{
|
| 3337 |
-
"epoch": 2.51,
|
| 3338 |
-
"grad_norm": 1.0221845787587531,
|
| 3339 |
-
"learning_rate": 6.554993204270993e-07,
|
| 3340 |
-
"loss": 0.6512,
|
| 3341 |
-
"step": 2375
|
| 3342 |
-
},
|
| 3343 |
-
{
|
| 3344 |
-
"epoch": 2.52,
|
| 3345 |
-
"grad_norm": 0.9597530531552908,
|
| 3346 |
-
"learning_rate": 6.417274404027163e-07,
|
| 3347 |
-
"loss": 0.6482,
|
| 3348 |
-
"step": 2380
|
| 3349 |
-
},
|
| 3350 |
-
{
|
| 3351 |
-
"epoch": 2.52,
|
| 3352 |
-
"grad_norm": 1.0201542647464452,
|
| 3353 |
-
"learning_rate": 6.280918458403506e-07,
|
| 3354 |
-
"loss": 0.6623,
|
| 3355 |
-
"step": 2385
|
| 3356 |
-
},
|
| 3357 |
-
{
|
| 3358 |
-
"epoch": 2.53,
|
| 3359 |
-
"grad_norm": 0.9818765108255797,
|
| 3360 |
-
"learning_rate": 6.14592963132174e-07,
|
| 3361 |
-
"loss": 0.6599,
|
| 3362 |
-
"step": 2390
|
| 3363 |
-
},
|
| 3364 |
-
{
|
| 3365 |
-
"epoch": 2.53,
|
| 3366 |
-
"grad_norm": 1.0020031777534095,
|
| 3367 |
-
"learning_rate": 6.012312143953075e-07,
|
| 3368 |
-
"loss": 0.6818,
|
| 3369 |
-
"step": 2395
|
| 3370 |
-
},
|
| 3371 |
-
{
|
| 3372 |
-
"epoch": 2.54,
|
| 3373 |
-
"grad_norm": 1.020601700800406,
|
| 3374 |
-
"learning_rate": 5.880070174586228e-07,
|
| 3375 |
-
"loss": 0.6794,
|
| 3376 |
-
"step": 2400
|
| 3377 |
-
},
|
| 3378 |
-
{
|
| 3379 |
-
"epoch": 2.54,
|
| 3380 |
-
"grad_norm": 0.9781529112263975,
|
| 3381 |
-
"learning_rate": 5.74920785849673e-07,
|
| 3382 |
-
"loss": 0.6612,
|
| 3383 |
-
"step": 2405
|
| 3384 |
-
},
|
| 3385 |
-
{
|
| 3386 |
-
"epoch": 2.55,
|
| 3387 |
-
"grad_norm": 1.020456830272749,
|
| 3388 |
-
"learning_rate": 5.619729287817621e-07,
|
| 3389 |
-
"loss": 0.6638,
|
| 3390 |
-
"step": 2410
|
| 3391 |
-
},
|
| 3392 |
-
{
|
| 3393 |
-
"epoch": 2.55,
|
| 3394 |
-
"grad_norm": 1.0134058298180835,
|
| 3395 |
-
"learning_rate": 5.49163851141154e-07,
|
| 3396 |
-
"loss": 0.6468,
|
| 3397 |
-
"step": 2415
|
| 3398 |
-
},
|
| 3399 |
-
{
|
| 3400 |
-
"epoch": 2.56,
|
| 3401 |
-
"grad_norm": 1.0051724307379968,
|
| 3402 |
-
"learning_rate": 5.36493953474404e-07,
|
| 3403 |
-
"loss": 0.6411,
|
| 3404 |
-
"step": 2420
|
| 3405 |
-
},
|
| 3406 |
-
{
|
| 3407 |
-
"epoch": 2.56,
|
| 3408 |
-
"grad_norm": 0.9963926377815217,
|
| 3409 |
-
"learning_rate": 5.239636319758356e-07,
|
| 3410 |
-
"loss": 0.668,
|
| 3411 |
-
"step": 2425
|
| 3412 |
-
},
|
| 3413 |
-
{
|
| 3414 |
-
"epoch": 2.57,
|
| 3415 |
-
"grad_norm": 0.9731428272925532,
|
| 3416 |
-
"learning_rate": 5.115732784751576e-07,
|
| 3417 |
-
"loss": 0.6444,
|
| 3418 |
-
"step": 2430
|
| 3419 |
-
},
|
| 3420 |
-
{
|
| 3421 |
-
"epoch": 2.57,
|
| 3422 |
-
"grad_norm": 1.0185774017291327,
|
| 3423 |
-
"learning_rate": 4.993232804252018e-07,
|
| 3424 |
-
"loss": 0.6529,
|
| 3425 |
-
"step": 2435
|
| 3426 |
-
},
|
| 3427 |
-
{
|
| 3428 |
-
"epoch": 2.58,
|
| 3429 |
-
"grad_norm": 1.00711656230006,
|
| 3430 |
-
"learning_rate": 4.872140208898118e-07,
|
| 3431 |
-
"loss": 0.6539,
|
| 3432 |
-
"step": 2440
|
| 3433 |
-
},
|
| 3434 |
-
{
|
| 3435 |
-
"epoch": 2.58,
|
| 3436 |
-
"grad_norm": 1.0045164786035452,
|
| 3437 |
-
"learning_rate": 4.7524587853186866e-07,
|
| 3438 |
-
"loss": 0.6629,
|
| 3439 |
-
"step": 2445
|
| 3440 |
-
},
|
| 3441 |
-
{
|
| 3442 |
-
"epoch": 2.59,
|
| 3443 |
-
"grad_norm": 0.9961645157673277,
|
| 3444 |
-
"learning_rate": 4.634192276014399e-07,
|
| 3445 |
-
"loss": 0.6738,
|
| 3446 |
-
"step": 2450
|
| 3447 |
-
},
|
| 3448 |
-
{
|
| 3449 |
-
"epoch": 2.59,
|
| 3450 |
-
"grad_norm": 1.0214318273829783,
|
| 3451 |
-
"learning_rate": 4.5173443792408625e-07,
|
| 3452 |
-
"loss": 0.6552,
|
| 3453 |
-
"step": 2455
|
| 3454 |
-
},
|
| 3455 |
-
{
|
| 3456 |
-
"epoch": 2.6,
|
| 3457 |
-
"grad_norm": 1.0163355618069994,
|
| 3458 |
-
"learning_rate": 4.4019187488928914e-07,
|
| 3459 |
-
"loss": 0.6638,
|
| 3460 |
-
"step": 2460
|
| 3461 |
-
},
|
| 3462 |
-
{
|
| 3463 |
-
"epoch": 2.61,
|
| 3464 |
-
"grad_norm": 1.032574771687925,
|
| 3465 |
-
"learning_rate": 4.2879189943903335e-07,
|
| 3466 |
-
"loss": 0.6877,
|
| 3467 |
-
"step": 2465
|
| 3468 |
-
},
|
| 3469 |
-
{
|
| 3470 |
-
"epoch": 2.61,
|
| 3471 |
-
"grad_norm": 0.9930486578442914,
|
| 3472 |
-
"learning_rate": 4.1753486805651e-07,
|
| 3473 |
-
"loss": 0.6832,
|
| 3474 |
-
"step": 2470
|
| 3475 |
-
},
|
| 3476 |
-
{
|
| 3477 |
-
"epoch": 2.62,
|
| 3478 |
-
"grad_norm": 0.969259241462703,
|
| 3479 |
-
"learning_rate": 4.064211327549794e-07,
|
| 3480 |
-
"loss": 0.6738,
|
| 3481 |
-
"step": 2475
|
| 3482 |
-
},
|
| 3483 |
-
{
|
| 3484 |
-
"epoch": 2.62,
|
| 3485 |
-
"grad_norm": 1.018380412495952,
|
| 3486 |
-
"learning_rate": 3.95451041066755e-07,
|
| 3487 |
-
"loss": 0.671,
|
| 3488 |
-
"step": 2480
|
| 3489 |
-
},
|
| 3490 |
-
{
|
| 3491 |
-
"epoch": 2.63,
|
| 3492 |
-
"grad_norm": 0.9735720562840744,
|
| 3493 |
-
"learning_rate": 3.8462493603234064e-07,
|
| 3494 |
-
"loss": 0.6433,
|
| 3495 |
-
"step": 2485
|
| 3496 |
-
},
|
| 3497 |
-
{
|
| 3498 |
-
"epoch": 2.63,
|
| 3499 |
-
"grad_norm": 1.023935871901339,
|
| 3500 |
-
"learning_rate": 3.739431561897011e-07,
|
| 3501 |
-
"loss": 0.6593,
|
| 3502 |
-
"step": 2490
|
| 3503 |
-
},
|
| 3504 |
-
{
|
| 3505 |
-
"epoch": 2.64,
|
| 3506 |
-
"grad_norm": 0.9931869209408388,
|
| 3507 |
-
"learning_rate": 3.634060355636798e-07,
|
| 3508 |
-
"loss": 0.6647,
|
| 3509 |
-
"step": 2495
|
| 3510 |
-
},
|
| 3511 |
-
{
|
| 3512 |
-
"epoch": 2.64,
|
| 3513 |
-
"grad_norm": 1.0007736035504975,
|
| 3514 |
-
"learning_rate": 3.53013903655548e-07,
|
| 3515 |
-
"loss": 0.6683,
|
| 3516 |
-
"step": 2500
|
| 3517 |
-
},
|
| 3518 |
-
{
|
| 3519 |
-
"epoch": 2.65,
|
| 3520 |
-
"grad_norm": 0.9926593135266999,
|
| 3521 |
-
"learning_rate": 3.427670854327042e-07,
|
| 3522 |
-
"loss": 0.6668,
|
| 3523 |
-
"step": 2505
|
| 3524 |
-
},
|
| 3525 |
-
{
|
| 3526 |
-
"epoch": 2.65,
|
| 3527 |
-
"grad_norm": 0.9870259704326787,
|
| 3528 |
-
"learning_rate": 3.3266590131851296e-07,
|
| 3529 |
-
"loss": 0.6583,
|
| 3530 |
-
"step": 2510
|
| 3531 |
-
},
|
| 3532 |
-
{
|
| 3533 |
-
"epoch": 2.66,
|
| 3534 |
-
"grad_norm": 1.0298553599069395,
|
| 3535 |
-
"learning_rate": 3.227106671822849e-07,
|
| 3536 |
-
"loss": 0.6835,
|
| 3537 |
-
"step": 2515
|
| 3538 |
-
},
|
| 3539 |
-
{
|
| 3540 |
-
"epoch": 2.66,
|
| 3541 |
-
"grad_norm": 0.9915918166378904,
|
| 3542 |
-
"learning_rate": 3.1290169432939556e-07,
|
| 3543 |
-
"loss": 0.6428,
|
| 3544 |
-
"step": 2520
|
| 3545 |
-
},
|
| 3546 |
-
{
|
| 3547 |
-
"epoch": 2.67,
|
| 3548 |
-
"grad_norm": 1.060474012796049,
|
| 3549 |
-
"learning_rate": 3.03239289491557e-07,
|
| 3550 |
-
"loss": 0.6571,
|
| 3551 |
-
"step": 2525
|
| 3552 |
-
},
|
| 3553 |
-
{
|
| 3554 |
-
"epoch": 2.67,
|
| 3555 |
-
"grad_norm": 1.0203183687136719,
|
| 3556 |
-
"learning_rate": 2.937237548172206e-07,
|
| 3557 |
-
"loss": 0.6511,
|
| 3558 |
-
"step": 2530
|
| 3559 |
-
},
|
| 3560 |
-
{
|
| 3561 |
-
"epoch": 2.68,
|
| 3562 |
-
"grad_norm": 0.989507237700814,
|
| 3563 |
-
"learning_rate": 2.8435538786213134e-07,
|
| 3564 |
-
"loss": 0.6746,
|
| 3565 |
-
"step": 2535
|
| 3566 |
-
},
|
| 3567 |
-
{
|
| 3568 |
-
"epoch": 2.68,
|
| 3569 |
-
"grad_norm": 0.9853274639882493,
|
| 3570 |
-
"learning_rate": 2.7513448158002334e-07,
|
| 3571 |
-
"loss": 0.6657,
|
| 3572 |
-
"step": 2540
|
| 3573 |
-
},
|
| 3574 |
-
{
|
| 3575 |
-
"epoch": 2.69,
|
| 3576 |
-
"grad_norm": 0.9957797339050202,
|
| 3577 |
-
"learning_rate": 2.66061324313458e-07,
|
| 3578 |
-
"loss": 0.6496,
|
| 3579 |
-
"step": 2545
|
| 3580 |
-
},
|
| 3581 |
-
{
|
| 3582 |
-
"epoch": 2.69,
|
| 3583 |
-
"grad_norm": 1.0073836211394178,
|
| 3584 |
-
"learning_rate": 2.5713619978480653e-07,
|
| 3585 |
-
"loss": 0.6596,
|
| 3586 |
-
"step": 2550
|
| 3587 |
-
},
|
| 3588 |
-
{
|
| 3589 |
-
"epoch": 2.7,
|
| 3590 |
-
"grad_norm": 0.9798969178233458,
|
| 3591 |
-
"learning_rate": 2.483593870873829e-07,
|
| 3592 |
-
"loss": 0.654,
|
| 3593 |
-
"step": 2555
|
| 3594 |
-
},
|
| 3595 |
-
{
|
| 3596 |
-
"epoch": 2.71,
|
| 3597 |
-
"grad_norm": 0.9936847658098146,
|
| 3598 |
-
"learning_rate": 2.3973116067670665e-07,
|
| 3599 |
-
"loss": 0.6457,
|
| 3600 |
-
"step": 2560
|
| 3601 |
-
},
|
| 3602 |
-
{
|
| 3603 |
-
"epoch": 2.71,
|
| 3604 |
-
"grad_norm": 1.0224466038654803,
|
| 3605 |
-
"learning_rate": 2.3125179036193214e-07,
|
| 3606 |
-
"loss": 0.6572,
|
| 3607 |
-
"step": 2565
|
| 3608 |
-
},
|
| 3609 |
-
{
|
| 3610 |
-
"epoch": 2.72,
|
| 3611 |
-
"grad_norm": 1.0378183041017084,
|
| 3612 |
-
"learning_rate": 2.2292154129740117e-07,
|
| 3613 |
-
"loss": 0.6554,
|
| 3614 |
-
"step": 2570
|
| 3615 |
-
},
|
| 3616 |
-
{
|
| 3617 |
-
"epoch": 2.72,
|
| 3618 |
-
"grad_norm": 0.9787357607930246,
|
| 3619 |
-
"learning_rate": 2.147406739743596e-07,
|
| 3620 |
-
"loss": 0.6689,
|
| 3621 |
-
"step": 2575
|
| 3622 |
-
},
|
| 3623 |
-
{
|
| 3624 |
-
"epoch": 2.73,
|
| 3625 |
-
"grad_norm": 1.003947207260689,
|
| 3626 |
-
"learning_rate": 2.0670944421280646e-07,
|
| 3627 |
-
"loss": 0.6458,
|
| 3628 |
-
"step": 2580
|
| 3629 |
-
},
|
| 3630 |
-
{
|
| 3631 |
-
"epoch": 2.73,
|
| 3632 |
-
"grad_norm": 1.0063190015667964,
|
| 3633 |
-
"learning_rate": 1.9882810315349554e-07,
|
| 3634 |
-
"loss": 0.6648,
|
| 3635 |
-
"step": 2585
|
| 3636 |
-
},
|
| 3637 |
-
{
|
| 3638 |
-
"epoch": 2.74,
|
| 3639 |
-
"grad_norm": 1.0148103533053272,
|
| 3640 |
-
"learning_rate": 1.9109689725008317e-07,
|
| 3641 |
-
"loss": 0.6738,
|
| 3642 |
-
"step": 2590
|
| 3643 |
-
},
|
| 3644 |
-
{
|
| 3645 |
-
"epoch": 2.74,
|
| 3646 |
-
"grad_norm": 1.0122729219524842,
|
| 3647 |
-
"learning_rate": 1.8351606826142176e-07,
|
| 3648 |
-
"loss": 0.6796,
|
| 3649 |
-
"step": 2595
|
| 3650 |
-
},
|
| 3651 |
-
{
|
| 3652 |
-
"epoch": 2.75,
|
| 3653 |
-
"grad_norm": 1.0170129872933447,
|
| 3654 |
-
"learning_rate": 1.7608585324399684e-07,
|
| 3655 |
-
"loss": 0.6798,
|
| 3656 |
-
"step": 2600
|
| 3657 |
-
},
|
| 3658 |
-
{
|
| 3659 |
-
"epoch": 2.75,
|
| 3660 |
-
"grad_norm": 0.992464215850126,
|
| 3661 |
-
"learning_rate": 1.688064845445192e-07,
|
| 3662 |
-
"loss": 0.6695,
|
| 3663 |
-
"step": 2605
|
| 3664 |
-
},
|
| 3665 |
-
{
|
| 3666 |
-
"epoch": 2.76,
|
| 3667 |
-
"grad_norm": 0.9778375876093532,
|
| 3668 |
-
"learning_rate": 1.6167818979265282e-07,
|
| 3669 |
-
"loss": 0.6563,
|
| 3670 |
-
"step": 2610
|
| 3671 |
-
},
|
| 3672 |
-
{
|
| 3673 |
-
"epoch": 2.76,
|
| 3674 |
-
"grad_norm": 1.0165595693382412,
|
| 3675 |
-
"learning_rate": 1.5470119189390342e-07,
|
| 3676 |
-
"loss": 0.6709,
|
| 3677 |
-
"step": 2615
|
| 3678 |
-
},
|
| 3679 |
-
{
|
| 3680 |
-
"epoch": 2.77,
|
| 3681 |
-
"grad_norm": 0.9846854115443192,
|
| 3682 |
-
"learning_rate": 1.4787570902264293e-07,
|
| 3683 |
-
"loss": 0.6468,
|
| 3684 |
-
"step": 2620
|
| 3685 |
-
},
|
| 3686 |
-
{
|
| 3687 |
-
"epoch": 2.77,
|
| 3688 |
-
"grad_norm": 1.0226129803358943,
|
| 3689 |
-
"learning_rate": 1.4120195461529097e-07,
|
| 3690 |
-
"loss": 0.6699,
|
| 3691 |
-
"step": 2625
|
| 3692 |
-
},
|
| 3693 |
-
{
|
| 3694 |
-
"epoch": 2.78,
|
| 3695 |
-
"grad_norm": 1.0082916511837874,
|
| 3696 |
-
"learning_rate": 1.3468013736363694e-07,
|
| 3697 |
-
"loss": 0.6516,
|
| 3698 |
-
"step": 2630
|
| 3699 |
-
},
|
| 3700 |
-
{
|
| 3701 |
-
"epoch": 2.78,
|
| 3702 |
-
"grad_norm": 1.0086534086914538,
|
| 3703 |
-
"learning_rate": 1.2831046120831692e-07,
|
| 3704 |
-
"loss": 0.6483,
|
| 3705 |
-
"step": 2635
|
| 3706 |
-
},
|
| 3707 |
-
{
|
| 3708 |
-
"epoch": 2.79,
|
| 3709 |
-
"grad_norm": 0.9957571698657345,
|
| 3710 |
-
"learning_rate": 1.2209312533243535e-07,
|
| 3711 |
-
"loss": 0.6632,
|
| 3712 |
-
"step": 2640
|
| 3713 |
-
},
|
| 3714 |
-
{
|
| 3715 |
-
"epoch": 2.8,
|
| 3716 |
-
"grad_norm": 1.0298383480420663,
|
| 3717 |
-
"learning_rate": 1.1602832415533616e-07,
|
| 3718 |
-
"loss": 0.6645,
|
| 3719 |
-
"step": 2645
|
| 3720 |
-
},
|
| 3721 |
-
{
|
| 3722 |
-
"epoch": 2.8,
|
| 3723 |
-
"grad_norm": 1.0188314052602203,
|
| 3724 |
-
"learning_rate": 1.1011624732652437e-07,
|
| 3725 |
-
"loss": 0.6752,
|
| 3726 |
-
"step": 2650
|
| 3727 |
-
},
|
| 3728 |
-
{
|
| 3729 |
-
"epoch": 2.81,
|
| 3730 |
-
"grad_norm": 1.0019681746822835,
|
| 3731 |
-
"learning_rate": 1.0435707971973297e-07,
|
| 3732 |
-
"loss": 0.6573,
|
| 3733 |
-
"step": 2655
|
| 3734 |
-
},
|
| 3735 |
-
{
|
| 3736 |
-
"epoch": 2.81,
|
| 3737 |
-
"grad_norm": 0.9926022445477827,
|
| 3738 |
-
"learning_rate": 9.875100142714478e-08,
|
| 3739 |
-
"loss": 0.6396,
|
| 3740 |
-
"step": 2660
|
| 3741 |
-
},
|
| 3742 |
-
{
|
| 3743 |
-
"epoch": 2.82,
|
| 3744 |
-
"grad_norm": 0.9847567872289796,
|
| 3745 |
-
"learning_rate": 9.329818775376088e-08,
|
| 3746 |
-
"loss": 0.672,
|
| 3747 |
-
"step": 2665
|
| 3748 |
-
},
|
| 3749 |
-
{
|
| 3750 |
-
"epoch": 2.82,
|
| 3751 |
-
"grad_norm": 1.0103069579844817,
|
| 3752 |
-
"learning_rate": 8.79988092119144e-08,
|
| 3753 |
-
"loss": 0.678,
|
| 3754 |
-
"step": 2670
|
| 3755 |
-
},
|
| 3756 |
-
{
|
| 3757 |
-
"epoch": 2.83,
|
| 3758 |
-
"grad_norm": 1.0092463732513441,
|
| 3759 |
-
"learning_rate": 8.285303151594537e-08,
|
| 3760 |
-
"loss": 0.6837,
|
| 3761 |
-
"step": 2675
|
| 3762 |
-
},
|
| 3763 |
-
{
|
| 3764 |
-
"epoch": 2.83,
|
| 3765 |
-
"grad_norm": 1.0032753352403014,
|
| 3766 |
-
"learning_rate": 7.786101557701209e-08,
|
| 3767 |
-
"loss": 0.6494,
|
| 3768 |
-
"step": 2680
|
| 3769 |
-
},
|
| 3770 |
-
{
|
| 3771 |
-
"epoch": 2.84,
|
| 3772 |
-
"grad_norm": 1.0278927407365124,
|
| 3773 |
-
"learning_rate": 7.302291749806345e-08,
|
| 3774 |
-
"loss": 0.6597,
|
| 3775 |
-
"step": 2685
|
| 3776 |
-
},
|
| 3777 |
-
{
|
| 3778 |
-
"epoch": 2.84,
|
| 3779 |
-
"grad_norm": 0.9985234255556347,
|
| 3780 |
-
"learning_rate": 6.833888856895676e-08,
|
| 3781 |
-
"loss": 0.6672,
|
| 3782 |
-
"step": 2690
|
| 3783 |
-
},
|
| 3784 |
-
{
|
| 3785 |
-
"epoch": 2.85,
|
| 3786 |
-
"grad_norm": 1.0086435046290338,
|
| 3787 |
-
"learning_rate": 6.380907526172597e-08,
|
| 3788 |
-
"loss": 0.6768,
|
| 3789 |
-
"step": 2695
|
| 3790 |
-
},
|
| 3791 |
-
{
|
| 3792 |
-
"epoch": 2.85,
|
| 3793 |
-
"grad_norm": 0.9639413787477988,
|
| 3794 |
-
"learning_rate": 5.943361922600255e-08,
|
| 3795 |
-
"loss": 0.6346,
|
| 3796 |
-
"step": 2700
|
| 3797 |
-
},
|
| 3798 |
-
{
|
| 3799 |
-
"epoch": 2.86,
|
| 3800 |
-
"grad_norm": 0.9898392259409212,
|
| 3801 |
-
"learning_rate": 5.521265728458347e-08,
|
| 3802 |
-
"loss": 0.6655,
|
| 3803 |
-
"step": 2705
|
| 3804 |
-
},
|
| 3805 |
-
{
|
| 3806 |
-
"epoch": 2.86,
|
| 3807 |
-
"grad_norm": 1.0000733408715612,
|
| 3808 |
-
"learning_rate": 5.114632142915687e-08,
|
| 3809 |
-
"loss": 0.638,
|
| 3810 |
-
"step": 2710
|
| 3811 |
-
},
|
| 3812 |
-
{
|
| 3813 |
-
"epoch": 2.87,
|
| 3814 |
-
"grad_norm": 0.990452054352071,
|
| 3815 |
-
"learning_rate": 4.723473881617147e-08,
|
| 3816 |
-
"loss": 0.6583,
|
| 3817 |
-
"step": 2715
|
| 3818 |
-
},
|
| 3819 |
-
{
|
| 3820 |
-
"epoch": 2.87,
|
| 3821 |
-
"grad_norm": 0.988717000145255,
|
| 3822 |
-
"learning_rate": 4.347803176286025e-08,
|
| 3823 |
-
"loss": 0.6708,
|
| 3824 |
-
"step": 2720
|
| 3825 |
-
},
|
| 3826 |
-
{
|
| 3827 |
-
"epoch": 2.88,
|
| 3828 |
-
"grad_norm": 0.9868081897157113,
|
| 3829 |
-
"learning_rate": 3.98763177434186e-08,
|
| 3830 |
-
"loss": 0.6583,
|
| 3831 |
-
"step": 2725
|
| 3832 |
-
},
|
| 3833 |
-
{
|
| 3834 |
-
"epoch": 2.89,
|
| 3835 |
-
"grad_norm": 1.001603936622736,
|
| 3836 |
-
"learning_rate": 3.642970938532553e-08,
|
| 3837 |
-
"loss": 0.6754,
|
| 3838 |
-
"step": 2730
|
| 3839 |
-
},
|
| 3840 |
-
{
|
| 3841 |
-
"epoch": 2.89,
|
| 3842 |
-
"grad_norm": 1.0028854813842756,
|
| 3843 |
-
"learning_rate": 3.313831446582816e-08,
|
| 3844 |
-
"loss": 0.6784,
|
| 3845 |
-
"step": 2735
|
| 3846 |
-
},
|
| 3847 |
-
{
|
| 3848 |
-
"epoch": 2.9,
|
| 3849 |
-
"grad_norm": 0.9840591494137083,
|
| 3850 |
-
"learning_rate": 3.000223590856666e-08,
|
| 3851 |
-
"loss": 0.6651,
|
| 3852 |
-
"step": 2740
|
| 3853 |
-
},
|
| 3854 |
-
{
|
| 3855 |
-
"epoch": 2.9,
|
| 3856 |
-
"grad_norm": 1.0425902900408417,
|
| 3857 |
-
"learning_rate": 2.7021571780356804e-08,
|
| 3858 |
-
"loss": 0.6489,
|
| 3859 |
-
"step": 2745
|
| 3860 |
-
},
|
| 3861 |
-
{
|
| 3862 |
-
"epoch": 2.91,
|
| 3863 |
-
"grad_norm": 1.0016271763738829,
|
| 3864 |
-
"learning_rate": 2.419641528812522e-08,
|
| 3865 |
-
"loss": 0.6501,
|
| 3866 |
-
"step": 2750
|
| 3867 |
-
},
|
| 3868 |
-
{
|
| 3869 |
-
"epoch": 2.91,
|
| 3870 |
-
"grad_norm": 0.9875844742537229,
|
| 3871 |
-
"learning_rate": 2.1526854775992255e-08,
|
| 3872 |
-
"loss": 0.667,
|
| 3873 |
-
"step": 2755
|
| 3874 |
-
},
|
| 3875 |
-
{
|
| 3876 |
-
"epoch": 2.92,
|
| 3877 |
-
"grad_norm": 0.9909068409835267,
|
| 3878 |
-
"learning_rate": 1.901297372251143e-08,
|
| 3879 |
-
"loss": 0.6649,
|
| 3880 |
-
"step": 2760
|
| 3881 |
-
},
|
| 3882 |
-
{
|
| 3883 |
-
"epoch": 2.92,
|
| 3884 |
-
"grad_norm": 1.0200770120528766,
|
| 3885 |
-
"learning_rate": 1.665485073805817e-08,
|
| 3886 |
-
"loss": 0.6542,
|
| 3887 |
-
"step": 2765
|
| 3888 |
-
},
|
| 3889 |
-
{
|
| 3890 |
-
"epoch": 2.93,
|
| 3891 |
-
"grad_norm": 0.9699214260408161,
|
| 3892 |
-
"learning_rate": 1.4452559562370683e-08,
|
| 3893 |
-
"loss": 0.6644,
|
| 3894 |
-
"step": 2770
|
| 3895 |
-
},
|
| 3896 |
-
{
|
| 3897 |
-
"epoch": 2.93,
|
| 3898 |
-
"grad_norm": 0.9962874170809767,
|
| 3899 |
-
"learning_rate": 1.2406169062246232e-08,
|
| 3900 |
-
"loss": 0.6502,
|
| 3901 |
-
"step": 2775
|
| 3902 |
-
},
|
| 3903 |
-
{
|
| 3904 |
-
"epoch": 2.94,
|
| 3905 |
-
"grad_norm": 1.0264867036759864,
|
| 3906 |
-
"learning_rate": 1.0515743229385645e-08,
|
| 3907 |
-
"loss": 0.6698,
|
| 3908 |
-
"step": 2780
|
| 3909 |
-
},
|
| 3910 |
-
{
|
| 3911 |
-
"epoch": 2.94,
|
| 3912 |
-
"grad_norm": 1.0133222133442825,
|
| 3913 |
-
"learning_rate": 8.781341178393244e-09,
|
| 3914 |
-
"loss": 0.6723,
|
| 3915 |
-
"step": 2785
|
| 3916 |
-
},
|
| 3917 |
-
{
|
| 3918 |
-
"epoch": 2.95,
|
| 3919 |
-
"grad_norm": 1.0159129157737807,
|
| 3920 |
-
"learning_rate": 7.203017144927771e-09,
|
| 3921 |
-
"loss": 0.6561,
|
| 3922 |
-
"step": 2790
|
| 3923 |
-
},
|
| 3924 |
-
{
|
| 3925 |
-
"epoch": 2.95,
|
| 3926 |
-
"grad_norm": 0.9931795490054022,
|
| 3927 |
-
"learning_rate": 5.780820484007632e-09,
|
| 3928 |
-
"loss": 0.6563,
|
| 3929 |
-
"step": 2795
|
| 3930 |
-
},
|
| 3931 |
-
{
|
| 3932 |
-
"epoch": 2.96,
|
| 3933 |
-
"grad_norm": 1.0195254872888724,
|
| 3934 |
-
"learning_rate": 4.514795668466576e-09,
|
| 3935 |
-
"loss": 0.6808,
|
| 3936 |
-
"step": 2800
|
| 3937 |
-
},
|
| 3938 |
-
{
|
| 3939 |
-
"epoch": 2.96,
|
| 3940 |
-
"grad_norm": 1.0210108366337896,
|
| 3941 |
-
"learning_rate": 3.4049822875614757e-09,
|
| 3942 |
-
"loss": 0.6723,
|
| 3943 |
-
"step": 2805
|
| 3944 |
-
},
|
| 3945 |
-
{
|
| 3946 |
-
"epoch": 2.97,
|
| 3947 |
-
"grad_norm": 0.9891130306027911,
|
| 3948 |
-
"learning_rate": 2.4514150457377594e-09,
|
| 3949 |
-
"loss": 0.6763,
|
| 3950 |
-
"step": 2810
|
| 3951 |
-
},
|
| 3952 |
-
{
|
| 3953 |
-
"epoch": 2.97,
|
| 3954 |
-
"grad_norm": 0.9876265686294937,
|
| 3955 |
-
"learning_rate": 1.654123761541393e-09,
|
| 3956 |
-
"loss": 0.6652,
|
| 3957 |
-
"step": 2815
|
| 3958 |
-
},
|
| 3959 |
-
{
|
| 3960 |
-
"epoch": 2.98,
|
| 3961 |
-
"grad_norm": 0.9719073327336301,
|
| 3962 |
-
"learning_rate": 1.0131333666885124e-09,
|
| 3963 |
-
"loss": 0.6793,
|
| 3964 |
-
"step": 2820
|
| 3965 |
-
},
|
| 3966 |
-
{
|
| 3967 |
-
"epoch": 2.99,
|
| 3968 |
-
"grad_norm": 1.004648101535836,
|
| 3969 |
-
"learning_rate": 5.284639052832718e-10,
|
| 3970 |
-
"loss": 0.6643,
|
| 3971 |
-
"step": 2825
|
| 3972 |
-
},
|
| 3973 |
-
{
|
| 3974 |
-
"epoch": 2.99,
|
| 3975 |
-
"grad_norm": 1.0172517540637482,
|
| 3976 |
-
"learning_rate": 2.0013053319334341e-10,
|
| 3977 |
-
"loss": 0.6768,
|
| 3978 |
-
"step": 2830
|
| 3979 |
-
},
|
| 3980 |
-
{
|
| 3981 |
-
"epoch": 3.0,
|
| 3982 |
-
"grad_norm": 0.9650966122076953,
|
| 3983 |
-
"learning_rate": 2.814351757529643e-11,
|
| 3984 |
-
"loss": 0.6356,
|
| 3985 |
-
"step": 2835
|
| 3986 |
-
},
|
| 3987 |
-
{
|
| 3988 |
-
"epoch": 3.0,
|
| 3989 |
-
"step": 2838,
|
| 3990 |
-
"total_flos": 1471706245890048.0,
|
| 3991 |
-
"train_loss": 0.8058284866381398,
|
| 3992 |
-
"train_runtime": 31310.8966,
|
| 3993 |
-
"train_samples_per_second": 5.802,
|
| 3994 |
-
"train_steps_per_second": 0.091
|
| 3995 |
-
}
|
| 3996 |
-
],
|
| 3997 |
-
"logging_steps": 5,
|
| 3998 |
-
"max_steps": 2838,
|
| 3999 |
-
"num_input_tokens_seen": 0,
|
| 4000 |
-
"num_train_epochs": 3,
|
| 4001 |
-
"save_steps": 500,
|
| 4002 |
-
"total_flos": 1471706245890048.0,
|
| 4003 |
-
"train_batch_size": 4,
|
| 4004 |
-
"trial_name": null,
|
| 4005 |
-
"trial_params": null
|
| 4006 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|