tapt_ulmfit_reinit_whole_word_5K_no_reinit_classifier-LR_2e-05

This model is a fine-tuned version of microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 3.2526
  • Accuracy: 0.4517

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 3407
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 32
  • optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-06 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.06
  • num_epochs: 100
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Accuracy
10.4971 1.0 21 9.9747 0.0130
9.5453 2.0 42 7.5742 0.1006
7.1693 3.0 63 5.7197 0.2509
5.8951 4.0 84 5.0410 0.3039
5.3672 5.0 105 4.5359 0.3297
5.0224 6.0 126 4.3382 0.3462
4.742 7.0 147 4.1588 0.3589
4.5698 8.0 168 4.0386 0.3707
4.4564 9.0 189 4.0000 0.3725
4.357 10.0 210 3.8163 0.3961
4.2363 11.0 231 3.8212 0.3921
4.1515 12.0 252 3.8091 0.3944
4.0535 13.0 273 3.7407 0.4014
3.9913 14.0 294 3.6963 0.4021
3.9702 15.0 315 3.6459 0.4086
3.9457 16.0 336 3.6291 0.4084
3.8999 17.0 357 3.5868 0.4162
3.8133 18.0 378 3.5709 0.4184
3.7768 19.0 399 3.5593 0.4165
3.7556 20.0 420 3.5804 0.4177
3.6968 21.0 441 3.5838 0.4177
3.7113 22.0 462 3.5509 0.4196
3.6415 23.0 483 3.4941 0.4231
3.6249 24.0 504 3.4634 0.4238
3.6029 25.0 525 3.4496 0.4278
3.5838 26.0 546 3.4745 0.4272
3.5798 27.0 567 3.4518 0.4324
3.5762 28.0 588 3.5176 0.4237
3.5317 29.0 609 3.4472 0.4208
3.5312 30.0 630 3.4399 0.4298
3.484 31.0 651 3.3953 0.4348
3.492 32.0 672 3.4387 0.4287
3.4642 33.0 693 3.4385 0.4301
3.4556 34.0 714 3.4118 0.4352
3.4703 35.0 735 3.4121 0.4306
3.4157 36.0 756 3.3746 0.4381
3.3863 37.0 777 3.3709 0.4413
3.4176 38.0 798 3.3834 0.4364
3.4316 39.0 819 3.3444 0.4396
3.3904 40.0 840 3.3664 0.4377
3.3595 41.0 861 3.2911 0.4489
3.369 42.0 882 3.3726 0.4360
3.3612 43.0 903 3.3574 0.4329
3.3994 44.0 924 3.3158 0.4428
3.3062 45.0 945 3.3189 0.4499
3.2856 46.0 966 3.3371 0.4444
3.314 47.0 987 3.3012 0.4494
3.333 48.0 1008 3.2689 0.4522
3.3043 49.0 1029 3.3684 0.4401
3.2732 50.0 1050 3.3020 0.4408
3.3005 51.0 1071 3.3013 0.4443
3.2666 52.0 1092 3.3370 0.4419
3.2397 53.0 1113 3.3533 0.4405
3.261 54.0 1134 3.3398 0.4422
3.2677 55.0 1155 3.3325 0.4405
3.2501 56.0 1176 3.3436 0.4428
3.2443 57.0 1197 3.3427 0.4451
3.28 58.0 1218 3.3806 0.4389
3.2301 59.0 1239 3.2235 0.4572
3.2526 60.0 1260 3.3543 0.4402
3.2256 61.0 1281 3.3432 0.4451
3.2288 62.0 1302 3.2800 0.4516
3.205 63.0 1323 3.2833 0.4542
3.1952 64.0 1344 3.3317 0.4439
3.2333 65.0 1365 3.3256 0.4453
3.2015 66.0 1386 3.3485 0.4472
3.2215 67.0 1407 3.3037 0.4416
3.1505 68.0 1428 3.3030 0.4458
3.1521 69.0 1449 3.2951 0.4491
3.172 70.0 1470 3.2670 0.4511
3.1717 71.0 1491 3.3420 0.4417
3.1664 72.0 1512 3.3523 0.4456
3.1702 73.0 1533 3.2418 0.4536
3.1604 74.0 1554 3.2931 0.4480
3.1479 75.0 1575 3.2274 0.4525
3.1578 76.0 1596 3.3571 0.4469
3.1335 77.0 1617 3.2580 0.4536
3.155 78.0 1638 3.2328 0.4521
3.1475 79.0 1659 3.2460 0.4529
3.1701 80.0 1680 3.2757 0.4450
3.1547 81.0 1701 3.2507 0.4500
3.1483 82.0 1722 3.2842 0.4538
3.1144 83.0 1743 3.2563 0.4519
3.1554 84.0 1764 3.2697 0.4508
3.1142 85.0 1785 3.3466 0.4347
3.1438 86.0 1806 3.2828 0.4481
3.1466 87.0 1827 3.2621 0.4517
3.1504 88.0 1848 3.2583 0.4456
3.1116 89.0 1869 3.2623 0.4538
3.1362 90.0 1890 3.2623 0.4501
3.1368 91.0 1911 3.2146 0.4557
3.1106 92.0 1932 3.2504 0.4546
3.1341 93.0 1953 3.2531 0.4542
3.1312 94.0 1974 3.3133 0.4451
3.1581 95.0 1995 3.3001 0.4465
3.1451 96.0 2016 3.2344 0.4527
3.1172 97.0 2037 3.2965 0.4490
3.1234 98.0 2058 3.2498 0.4553
3.1064 99.0 2079 3.2326 0.4588
3.1169 100.0 2100 3.2526 0.4517

Framework versions

  • Transformers 4.48.2
  • Pytorch 2.4.1+cu121
  • Datasets 3.0.2
  • Tokenizers 0.21.0
Downloads last month
4
Safetensors
Model size
0.1B params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for Mardiyyah/tapt_ulmfit_reinit_whole_word_5K_no_reinit_classifier-LR_2e-05

Evaluation results