Upload RegTech-4B-Instruct
Browse files- model-00001-of-00002.safetensors +1 -1
- model-00002-of-00002.safetensors +1 -1
- training_report.json +56 -533
model-00001-of-00002.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4967215360
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f59ea02ad2810389028c4270786bd5b0c1b87435f666acd50bb71fc2e438273f
|
| 3 |
size 4967215360
|
model-00002-of-00002.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3077766632
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f348a1ba7d66204a1b5de57ff3b79a834a88aefbf020fb45b37f1f6299989645
|
| 3 |
size 3077766632
|
training_report.json
CHANGED
|
@@ -3,18 +3,18 @@
|
|
| 3 |
"model_name": "RegTech-4B-Instruct",
|
| 4 |
"dataset": "./train.jsonl",
|
| 5 |
"env_file": "/home/ubuntu/sophia-core-server/.tuning/.env.4B",
|
| 6 |
-
"train_samples":
|
| 7 |
-
"eval_samples":
|
| 8 |
"params": {
|
| 9 |
-
"rank":
|
| 10 |
-
"alpha":
|
| 11 |
-
"dropout": 0.
|
| 12 |
-
"lr":
|
| 13 |
"scheduler": "cosine",
|
| 14 |
-
"epochs":
|
| 15 |
-
"effective_batch":
|
| 16 |
"max_seq_length": 4096,
|
| 17 |
-
"neftune_alpha":
|
| 18 |
"target_modules": [
|
| 19 |
"q_proj",
|
| 20 |
"k_proj",
|
|
@@ -26,635 +26,158 @@
|
|
| 26 |
]
|
| 27 |
},
|
| 28 |
"results": {
|
| 29 |
-
"total_steps":
|
| 30 |
-
"final_train_loss": 1.
|
| 31 |
-
"best_eval_loss": 1.
|
| 32 |
-
"best_eval_step":
|
| 33 |
-
"
|
|
|
|
| 34 |
},
|
| 35 |
"loss_history": {
|
| 36 |
"train": [
|
| 37 |
-
[
|
| 38 |
-
5,
|
| 39 |
-
1.9266
|
| 40 |
-
],
|
| 41 |
[
|
| 42 |
10,
|
| 43 |
-
2.
|
| 44 |
-
],
|
| 45 |
-
[
|
| 46 |
-
15,
|
| 47 |
-
2.377
|
| 48 |
],
|
| 49 |
[
|
| 50 |
20,
|
| 51 |
-
2.
|
| 52 |
-
],
|
| 53 |
-
[
|
| 54 |
-
25,
|
| 55 |
-
2.3294
|
| 56 |
],
|
| 57 |
[
|
| 58 |
30,
|
| 59 |
-
2.
|
| 60 |
-
],
|
| 61 |
-
[
|
| 62 |
-
35,
|
| 63 |
-
2.1198
|
| 64 |
],
|
| 65 |
[
|
| 66 |
40,
|
| 67 |
-
2.
|
| 68 |
-
],
|
| 69 |
-
[
|
| 70 |
-
45,
|
| 71 |
-
2.0804
|
| 72 |
],
|
| 73 |
[
|
| 74 |
50,
|
| 75 |
-
1.
|
| 76 |
-
],
|
| 77 |
-
[
|
| 78 |
-
55,
|
| 79 |
-
1.6711
|
| 80 |
],
|
| 81 |
[
|
| 82 |
60,
|
| 83 |
-
|
| 84 |
-
],
|
| 85 |
-
[
|
| 86 |
-
65,
|
| 87 |
-
1.7905
|
| 88 |
],
|
| 89 |
[
|
| 90 |
70,
|
| 91 |
-
1.
|
| 92 |
-
],
|
| 93 |
-
[
|
| 94 |
-
75,
|
| 95 |
-
1.8905
|
| 96 |
],
|
| 97 |
[
|
| 98 |
80,
|
| 99 |
-
1.
|
| 100 |
-
],
|
| 101 |
-
[
|
| 102 |
-
85,
|
| 103 |
-
1.6943
|
| 104 |
],
|
| 105 |
[
|
| 106 |
90,
|
| 107 |
-
1.
|
| 108 |
-
],
|
| 109 |
-
[
|
| 110 |
-
95,
|
| 111 |
-
1.71
|
| 112 |
],
|
| 113 |
[
|
| 114 |
100,
|
| 115 |
-
1.
|
| 116 |
-
],
|
| 117 |
-
[
|
| 118 |
-
105,
|
| 119 |
-
1.6177
|
| 120 |
],
|
| 121 |
[
|
| 122 |
110,
|
| 123 |
-
1.
|
| 124 |
-
],
|
| 125 |
-
[
|
| 126 |
-
115,
|
| 127 |
-
1.6453
|
| 128 |
],
|
| 129 |
[
|
| 130 |
120,
|
| 131 |
-
1.
|
| 132 |
-
],
|
| 133 |
-
[
|
| 134 |
-
125,
|
| 135 |
-
1.3471
|
| 136 |
],
|
| 137 |
[
|
| 138 |
130,
|
| 139 |
-
1.
|
| 140 |
-
],
|
| 141 |
-
[
|
| 142 |
-
135,
|
| 143 |
-
1.307
|
| 144 |
],
|
| 145 |
[
|
| 146 |
140,
|
| 147 |
-
1.
|
| 148 |
-
],
|
| 149 |
-
[
|
| 150 |
-
145,
|
| 151 |
-
1.4156
|
| 152 |
],
|
| 153 |
[
|
| 154 |
150,
|
| 155 |
-
1.
|
| 156 |
-
],
|
| 157 |
-
[
|
| 158 |
-
155,
|
| 159 |
-
1.5545
|
| 160 |
],
|
| 161 |
[
|
| 162 |
160,
|
| 163 |
-
1.
|
| 164 |
-
],
|
| 165 |
-
[
|
| 166 |
-
165,
|
| 167 |
-
1.5327
|
| 168 |
],
|
| 169 |
[
|
| 170 |
170,
|
| 171 |
-
1.
|
| 172 |
-
],
|
| 173 |
-
[
|
| 174 |
-
175,
|
| 175 |
-
1.1613
|
| 176 |
],
|
| 177 |
[
|
| 178 |
180,
|
| 179 |
-
1.
|
| 180 |
-
],
|
| 181 |
-
[
|
| 182 |
-
185,
|
| 183 |
-
1.443
|
| 184 |
],
|
| 185 |
[
|
| 186 |
190,
|
| 187 |
-
1.
|
| 188 |
-
],
|
| 189 |
-
[
|
| 190 |
-
195,
|
| 191 |
-
1.3793
|
| 192 |
],
|
| 193 |
[
|
| 194 |
200,
|
| 195 |
-
1.
|
| 196 |
-
],
|
| 197 |
-
[
|
| 198 |
-
205,
|
| 199 |
-
1.3441
|
| 200 |
],
|
| 201 |
[
|
| 202 |
210,
|
| 203 |
-
1.
|
| 204 |
-
],
|
| 205 |
-
[
|
| 206 |
-
215,
|
| 207 |
-
1.296
|
| 208 |
],
|
| 209 |
[
|
| 210 |
220,
|
| 211 |
-
1.
|
| 212 |
-
],
|
| 213 |
-
[
|
| 214 |
-
225,
|
| 215 |
-
1.3165
|
| 216 |
],
|
| 217 |
[
|
| 218 |
230,
|
| 219 |
-
1.
|
| 220 |
-
],
|
| 221 |
-
[
|
| 222 |
-
235,
|
| 223 |
-
1.5172
|
| 224 |
],
|
| 225 |
[
|
| 226 |
240,
|
| 227 |
-
1.
|
| 228 |
-
],
|
| 229 |
-
[
|
| 230 |
-
245,
|
| 231 |
-
1.2391
|
| 232 |
],
|
| 233 |
[
|
| 234 |
250,
|
| 235 |
-
1.
|
| 236 |
-
],
|
| 237 |
-
[
|
| 238 |
-
255,
|
| 239 |
-
1.181
|
| 240 |
],
|
| 241 |
[
|
| 242 |
260,
|
| 243 |
-
1.
|
| 244 |
-
],
|
| 245 |
-
[
|
| 246 |
-
265,
|
| 247 |
-
1.4351
|
| 248 |
],
|
| 249 |
[
|
| 250 |
270,
|
| 251 |
-
1.
|
| 252 |
-
],
|
| 253 |
-
[
|
| 254 |
-
275,
|
| 255 |
-
1.2195
|
| 256 |
],
|
| 257 |
[
|
| 258 |
280,
|
| 259 |
-
1.
|
| 260 |
-
],
|
| 261 |
-
[
|
| 262 |
-
285,
|
| 263 |
-
1.4254
|
| 264 |
],
|
| 265 |
[
|
| 266 |
290,
|
| 267 |
-
1.
|
| 268 |
-
],
|
| 269 |
-
[
|
| 270 |
-
295,
|
| 271 |
-
1.1506
|
| 272 |
-
],
|
| 273 |
-
[
|
| 274 |
-
300,
|
| 275 |
-
1.138
|
| 276 |
-
],
|
| 277 |
-
[
|
| 278 |
-
305,
|
| 279 |
-
1.3738
|
| 280 |
-
],
|
| 281 |
-
[
|
| 282 |
-
310,
|
| 283 |
-
1.2216
|
| 284 |
-
],
|
| 285 |
-
[
|
| 286 |
-
315,
|
| 287 |
-
1.2873
|
| 288 |
-
],
|
| 289 |
-
[
|
| 290 |
-
320,
|
| 291 |
-
1.1959
|
| 292 |
-
],
|
| 293 |
-
[
|
| 294 |
-
325,
|
| 295 |
-
1.2746
|
| 296 |
-
],
|
| 297 |
-
[
|
| 298 |
-
330,
|
| 299 |
-
1.2089
|
| 300 |
-
],
|
| 301 |
-
[
|
| 302 |
-
335,
|
| 303 |
-
1.2466
|
| 304 |
-
],
|
| 305 |
-
[
|
| 306 |
-
340,
|
| 307 |
-
1.1855
|
| 308 |
-
],
|
| 309 |
-
[
|
| 310 |
-
345,
|
| 311 |
-
1.3954
|
| 312 |
-
],
|
| 313 |
-
[
|
| 314 |
-
350,
|
| 315 |
-
1.0918
|
| 316 |
-
],
|
| 317 |
-
[
|
| 318 |
-
355,
|
| 319 |
-
1.2062
|
| 320 |
-
],
|
| 321 |
-
[
|
| 322 |
-
360,
|
| 323 |
-
1.2809
|
| 324 |
-
],
|
| 325 |
-
[
|
| 326 |
-
365,
|
| 327 |
-
1.3698
|
| 328 |
-
],
|
| 329 |
-
[
|
| 330 |
-
370,
|
| 331 |
-
1.3147
|
| 332 |
-
],
|
| 333 |
-
[
|
| 334 |
-
375,
|
| 335 |
-
1.1451
|
| 336 |
-
],
|
| 337 |
-
[
|
| 338 |
-
380,
|
| 339 |
-
1.0194
|
| 340 |
-
],
|
| 341 |
-
[
|
| 342 |
-
385,
|
| 343 |
-
1.1412
|
| 344 |
-
],
|
| 345 |
-
[
|
| 346 |
-
390,
|
| 347 |
-
1.1606
|
| 348 |
-
],
|
| 349 |
-
[
|
| 350 |
-
395,
|
| 351 |
-
1.1469
|
| 352 |
-
],
|
| 353 |
-
[
|
| 354 |
-
400,
|
| 355 |
-
1.2781
|
| 356 |
-
],
|
| 357 |
-
[
|
| 358 |
-
405,
|
| 359 |
-
1.2447
|
| 360 |
-
],
|
| 361 |
-
[
|
| 362 |
-
410,
|
| 363 |
-
1.3158
|
| 364 |
-
],
|
| 365 |
-
[
|
| 366 |
-
415,
|
| 367 |
-
1.1832
|
| 368 |
-
],
|
| 369 |
-
[
|
| 370 |
-
420,
|
| 371 |
-
1.2036
|
| 372 |
-
],
|
| 373 |
-
[
|
| 374 |
-
425,
|
| 375 |
-
1.1602
|
| 376 |
-
],
|
| 377 |
-
[
|
| 378 |
-
430,
|
| 379 |
-
1.4217
|
| 380 |
-
],
|
| 381 |
-
[
|
| 382 |
-
435,
|
| 383 |
-
1.1954
|
| 384 |
-
],
|
| 385 |
-
[
|
| 386 |
-
440,
|
| 387 |
-
1.1913
|
| 388 |
-
],
|
| 389 |
-
[
|
| 390 |
-
445,
|
| 391 |
-
1.0779
|
| 392 |
-
],
|
| 393 |
-
[
|
| 394 |
-
450,
|
| 395 |
-
1.1522
|
| 396 |
-
],
|
| 397 |
-
[
|
| 398 |
-
455,
|
| 399 |
-
1.2223
|
| 400 |
-
],
|
| 401 |
-
[
|
| 402 |
-
460,
|
| 403 |
-
1.0818
|
| 404 |
-
],
|
| 405 |
-
[
|
| 406 |
-
465,
|
| 407 |
-
1.429
|
| 408 |
-
],
|
| 409 |
-
[
|
| 410 |
-
470,
|
| 411 |
-
1.145
|
| 412 |
-
],
|
| 413 |
-
[
|
| 414 |
-
475,
|
| 415 |
-
1.1874
|
| 416 |
-
],
|
| 417 |
-
[
|
| 418 |
-
480,
|
| 419 |
-
1.0495
|
| 420 |
-
],
|
| 421 |
-
[
|
| 422 |
-
485,
|
| 423 |
-
1.0664
|
| 424 |
-
],
|
| 425 |
-
[
|
| 426 |
-
490,
|
| 427 |
-
1.2633
|
| 428 |
-
],
|
| 429 |
-
[
|
| 430 |
-
495,
|
| 431 |
-
1.3179
|
| 432 |
-
],
|
| 433 |
-
[
|
| 434 |
-
500,
|
| 435 |
-
1.2056
|
| 436 |
-
],
|
| 437 |
-
[
|
| 438 |
-
505,
|
| 439 |
-
1.1259
|
| 440 |
-
],
|
| 441 |
-
[
|
| 442 |
-
510,
|
| 443 |
-
1.2683
|
| 444 |
-
],
|
| 445 |
-
[
|
| 446 |
-
515,
|
| 447 |
-
1.0229
|
| 448 |
-
],
|
| 449 |
-
[
|
| 450 |
-
520,
|
| 451 |
-
1.2201
|
| 452 |
-
],
|
| 453 |
-
[
|
| 454 |
-
525,
|
| 455 |
-
1.1291
|
| 456 |
-
],
|
| 457 |
-
[
|
| 458 |
-
530,
|
| 459 |
-
1.1688
|
| 460 |
-
],
|
| 461 |
-
[
|
| 462 |
-
535,
|
| 463 |
-
1.2019
|
| 464 |
-
],
|
| 465 |
-
[
|
| 466 |
-
540,
|
| 467 |
-
1.1733
|
| 468 |
-
],
|
| 469 |
-
[
|
| 470 |
-
545,
|
| 471 |
-
1.0913
|
| 472 |
-
],
|
| 473 |
-
[
|
| 474 |
-
550,
|
| 475 |
-
1.2309
|
| 476 |
-
],
|
| 477 |
-
[
|
| 478 |
-
555,
|
| 479 |
-
1.315
|
| 480 |
-
],
|
| 481 |
-
[
|
| 482 |
-
560,
|
| 483 |
-
1.1578
|
| 484 |
-
],
|
| 485 |
-
[
|
| 486 |
-
565,
|
| 487 |
-
1.2228
|
| 488 |
-
],
|
| 489 |
-
[
|
| 490 |
-
570,
|
| 491 |
-
0.9872
|
| 492 |
-
],
|
| 493 |
-
[
|
| 494 |
-
575,
|
| 495 |
-
1.1328
|
| 496 |
-
],
|
| 497 |
-
[
|
| 498 |
-
580,
|
| 499 |
-
1.1709
|
| 500 |
-
],
|
| 501 |
-
[
|
| 502 |
-
585,
|
| 503 |
-
1.1675
|
| 504 |
-
],
|
| 505 |
-
[
|
| 506 |
-
590,
|
| 507 |
-
1.3192
|
| 508 |
-
],
|
| 509 |
-
[
|
| 510 |
-
595,
|
| 511 |
-
1.0701
|
| 512 |
-
],
|
| 513 |
-
[
|
| 514 |
-
600,
|
| 515 |
-
1.1484
|
| 516 |
-
],
|
| 517 |
-
[
|
| 518 |
-
605,
|
| 519 |
-
0.9753
|
| 520 |
-
],
|
| 521 |
-
[
|
| 522 |
-
610,
|
| 523 |
-
1.2143
|
| 524 |
-
],
|
| 525 |
-
[
|
| 526 |
-
615,
|
| 527 |
-
1.238
|
| 528 |
-
],
|
| 529 |
-
[
|
| 530 |
-
620,
|
| 531 |
-
1.149
|
| 532 |
-
],
|
| 533 |
-
[
|
| 534 |
-
625,
|
| 535 |
-
1.1859
|
| 536 |
-
],
|
| 537 |
-
[
|
| 538 |
-
630,
|
| 539 |
-
1.225
|
| 540 |
-
],
|
| 541 |
-
[
|
| 542 |
-
635,
|
| 543 |
-
1.1635
|
| 544 |
-
],
|
| 545 |
-
[
|
| 546 |
-
640,
|
| 547 |
-
1.2214
|
| 548 |
-
],
|
| 549 |
-
[
|
| 550 |
-
645,
|
| 551 |
-
1.1366
|
| 552 |
-
],
|
| 553 |
-
[
|
| 554 |
-
650,
|
| 555 |
-
1.3328
|
| 556 |
-
],
|
| 557 |
-
[
|
| 558 |
-
655,
|
| 559 |
-
1.1442
|
| 560 |
-
],
|
| 561 |
-
[
|
| 562 |
-
660,
|
| 563 |
-
1.0937
|
| 564 |
-
],
|
| 565 |
-
[
|
| 566 |
-
665,
|
| 567 |
-
1.1968
|
| 568 |
-
],
|
| 569 |
-
[
|
| 570 |
-
670,
|
| 571 |
-
1.2908
|
| 572 |
-
],
|
| 573 |
-
[
|
| 574 |
-
675,
|
| 575 |
-
1.2003
|
| 576 |
-
],
|
| 577 |
-
[
|
| 578 |
-
680,
|
| 579 |
-
1.0809
|
| 580 |
-
],
|
| 581 |
-
[
|
| 582 |
-
685,
|
| 583 |
-
1.2908
|
| 584 |
-
],
|
| 585 |
-
[
|
| 586 |
-
690,
|
| 587 |
-
1.241
|
| 588 |
]
|
| 589 |
],
|
| 590 |
"eval": [
|
| 591 |
-
[
|
| 592 |
-
40,
|
| 593 |
-
2.1869590282440186
|
| 594 |
-
],
|
| 595 |
[
|
| 596 |
80,
|
| 597 |
-
|
| 598 |
-
],
|
| 599 |
-
[
|
| 600 |
-
120,
|
| 601 |
-
1.5918081998825073
|
| 602 |
],
|
| 603 |
[
|
| 604 |
160,
|
| 605 |
-
1.
|
| 606 |
-
],
|
| 607 |
-
[
|
| 608 |
-
200,
|
| 609 |
-
1.3858165740966797
|
| 610 |
],
|
| 611 |
[
|
| 612 |
240,
|
| 613 |
-
1.
|
| 614 |
-
]
|
| 615 |
-
|
| 616 |
-
|
| 617 |
-
1.3039580583572388
|
| 618 |
-
],
|
| 619 |
-
[
|
| 620 |
-
320,
|
| 621 |
-
1.2727303504943848
|
| 622 |
-
],
|
| 623 |
-
[
|
| 624 |
-
360,
|
| 625 |
-
1.245557188987732
|
| 626 |
-
],
|
| 627 |
-
[
|
| 628 |
-
400,
|
| 629 |
-
1.2276334762573242
|
| 630 |
-
],
|
| 631 |
-
[
|
| 632 |
-
440,
|
| 633 |
-
1.213688850402832
|
| 634 |
-
],
|
| 635 |
-
[
|
| 636 |
-
480,
|
| 637 |
-
1.2049111127853394
|
| 638 |
-
],
|
| 639 |
-
[
|
| 640 |
-
520,
|
| 641 |
-
1.1985464096069336
|
| 642 |
-
],
|
| 643 |
-
[
|
| 644 |
-
560,
|
| 645 |
-
1.1946879625320435
|
| 646 |
-
],
|
| 647 |
[
|
| 648 |
-
|
| 649 |
-
|
| 650 |
],
|
| 651 |
[
|
| 652 |
-
|
| 653 |
-
|
| 654 |
],
|
| 655 |
[
|
| 656 |
-
|
| 657 |
-
|
| 658 |
]
|
| 659 |
]
|
| 660 |
}
|
|
|
|
| 3 |
"model_name": "RegTech-4B-Instruct",
|
| 4 |
"dataset": "./train.jsonl",
|
| 5 |
"env_file": "/home/ubuntu/sophia-core-server/.tuning/.env.4B",
|
| 6 |
+
"train_samples": 2330,
|
| 7 |
+
"eval_samples": 258,
|
| 8 |
"params": {
|
| 9 |
+
"rank": 4,
|
| 10 |
+
"alpha": 8,
|
| 11 |
+
"dropout": 0.05,
|
| 12 |
+
"lr": 1e-05,
|
| 13 |
"scheduler": "cosine",
|
| 14 |
+
"epochs": 1,
|
| 15 |
+
"effective_batch": 8,
|
| 16 |
"max_seq_length": 4096,
|
| 17 |
+
"neftune_alpha": 0.0,
|
| 18 |
"target_modules": [
|
| 19 |
"q_proj",
|
| 20 |
"k_proj",
|
|
|
|
| 26 |
]
|
| 27 |
},
|
| 28 |
"results": {
|
| 29 |
+
"total_steps": 292,
|
| 30 |
+
"final_train_loss": 1.5045,
|
| 31 |
+
"best_eval_loss": 1.601854681968689,
|
| 32 |
+
"best_eval_step": 240,
|
| 33 |
+
"best_token_accuracy": 0.6812,
|
| 34 |
+
"elapsed_minutes": 8.6
|
| 35 |
},
|
| 36 |
"loss_history": {
|
| 37 |
"train": [
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
[
|
| 39 |
10,
|
| 40 |
+
2.1906
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
],
|
| 42 |
[
|
| 43 |
20,
|
| 44 |
+
2.0417
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
],
|
| 46 |
[
|
| 47 |
30,
|
| 48 |
+
2.1217
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
],
|
| 50 |
[
|
| 51 |
40,
|
| 52 |
+
2.0513
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
],
|
| 54 |
[
|
| 55 |
50,
|
| 56 |
+
1.9839
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
],
|
| 58 |
[
|
| 59 |
60,
|
| 60 |
+
1.9423
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
],
|
| 62 |
[
|
| 63 |
70,
|
| 64 |
+
1.9321
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
],
|
| 66 |
[
|
| 67 |
80,
|
| 68 |
+
1.8047
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
],
|
| 70 |
[
|
| 71 |
90,
|
| 72 |
+
1.7045
|
|
|
|
|
|
|
|
|
|
|
|
|
| 73 |
],
|
| 74 |
[
|
| 75 |
100,
|
| 76 |
+
1.8603
|
|
|
|
|
|
|
|
|
|
|
|
|
| 77 |
],
|
| 78 |
[
|
| 79 |
110,
|
| 80 |
+
1.721
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
],
|
| 82 |
[
|
| 83 |
120,
|
| 84 |
+
1.6419
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
],
|
| 86 |
[
|
| 87 |
130,
|
| 88 |
+
1.5821
|
|
|
|
|
|
|
|
|
|
|
|
|
| 89 |
],
|
| 90 |
[
|
| 91 |
140,
|
| 92 |
+
1.5593
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
],
|
| 94 |
[
|
| 95 |
150,
|
| 96 |
+
1.4756
|
|
|
|
|
|
|
|
|
|
|
|
|
| 97 |
],
|
| 98 |
[
|
| 99 |
160,
|
| 100 |
+
1.4945
|
|
|
|
|
|
|
|
|
|
|
|
|
| 101 |
],
|
| 102 |
[
|
| 103 |
170,
|
| 104 |
+
1.5168
|
|
|
|
|
|
|
|
|
|
|
|
|
| 105 |
],
|
| 106 |
[
|
| 107 |
180,
|
| 108 |
+
1.5689
|
|
|
|
|
|
|
|
|
|
|
|
|
| 109 |
],
|
| 110 |
[
|
| 111 |
190,
|
| 112 |
+
1.3763
|
|
|
|
|
|
|
|
|
|
|
|
|
| 113 |
],
|
| 114 |
[
|
| 115 |
200,
|
| 116 |
+
1.5759
|
|
|
|
|
|
|
|
|
|
|
|
|
| 117 |
],
|
| 118 |
[
|
| 119 |
210,
|
| 120 |
+
1.477
|
|
|
|
|
|
|
|
|
|
|
|
|
| 121 |
],
|
| 122 |
[
|
| 123 |
220,
|
| 124 |
+
1.4889
|
|
|
|
|
|
|
|
|
|
|
|
|
| 125 |
],
|
| 126 |
[
|
| 127 |
230,
|
| 128 |
+
1.4514
|
|
|
|
|
|
|
|
|
|
|
|
|
| 129 |
],
|
| 130 |
[
|
| 131 |
240,
|
| 132 |
+
1.441
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
],
|
| 134 |
[
|
| 135 |
250,
|
| 136 |
+
1.427
|
|
|
|
|
|
|
|
|
|
|
|
|
| 137 |
],
|
| 138 |
[
|
| 139 |
260,
|
| 140 |
+
1.4423
|
|
|
|
|
|
|
|
|
|
|
|
|
| 141 |
],
|
| 142 |
[
|
| 143 |
270,
|
| 144 |
+
1.4199
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
],
|
| 146 |
[
|
| 147 |
280,
|
| 148 |
+
1.457
|
|
|
|
|
|
|
|
|
|
|
|
|
| 149 |
],
|
| 150 |
[
|
| 151 |
290,
|
| 152 |
+
1.5045
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 153 |
]
|
| 154 |
],
|
| 155 |
"eval": [
|
|
|
|
|
|
|
|
|
|
|
|
|
| 156 |
[
|
| 157 |
80,
|
| 158 |
+
2.036996841430664
|
|
|
|
|
|
|
|
|
|
|
|
|
| 159 |
],
|
| 160 |
[
|
| 161 |
160,
|
| 162 |
+
1.6603444814682007
|
|
|
|
|
|
|
|
|
|
|
|
|
| 163 |
],
|
| 164 |
[
|
| 165 |
240,
|
| 166 |
+
1.601854681968689
|
| 167 |
+
]
|
| 168 |
+
],
|
| 169 |
+
"token_accuracy": [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 170 |
[
|
| 171 |
+
80,
|
| 172 |
+
0.661
|
| 173 |
],
|
| 174 |
[
|
| 175 |
+
160,
|
| 176 |
+
0.6759
|
| 177 |
],
|
| 178 |
[
|
| 179 |
+
240,
|
| 180 |
+
0.6812
|
| 181 |
]
|
| 182 |
]
|
| 183 |
}
|