<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
vit-base-patch16-224-in21k-tiny_rvl_cdip-NK1000_hint
This model is a fine-tuned version of WinKawaks/vit-tiny-patch16-224 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 75.3146
- Accuracy: 0.811
- Brier Loss: 0.3395
- Nll: 2.0856
- F1 Micro: 0.811
- F1 Macro: 0.8109
- Ece: 0.1625
- Aurc: 0.0586
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: 0.0001
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 50
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Brier Loss | Nll | F1 Micro | F1 Macro | Ece | Aurc |
|---|---|---|---|---|---|---|---|---|---|---|
| 78.9278 | 1.0 | 1000 | 78.3645 | 0.588 | 0.5476 | 2.4320 | 0.588 | 0.5825 | 0.0552 | 0.1928 |
| 77.6109 | 2.0 | 2000 | 77.4909 | 0.684 | 0.4259 | 2.1686 | 0.684 | 0.6879 | 0.0496 | 0.1167 |
| 77.1849 | 3.0 | 3000 | 77.1828 | 0.7185 | 0.3854 | 2.1032 | 0.7185 | 0.7200 | 0.0495 | 0.0932 |
| 76.8526 | 4.0 | 4000 | 76.9800 | 0.748 | 0.3549 | 2.0716 | 0.748 | 0.7492 | 0.0768 | 0.0781 |
| 76.5928 | 5.0 | 5000 | 76.7544 | 0.743 | 0.3576 | 2.0634 | 0.743 | 0.7461 | 0.0564 | 0.0846 |
| 76.1507 | 6.0 | 6000 | 76.5850 | 0.7688 | 0.3354 | 2.0506 | 0.7688 | 0.7698 | 0.0857 | 0.0701 |
| 75.8107 | 7.0 | 7000 | 76.5816 | 0.75 | 0.3766 | 2.1753 | 0.75 | 0.7542 | 0.1230 | 0.0815 |
| 75.868 | 8.0 | 8000 | 76.5048 | 0.785 | 0.3324 | 2.0865 | 0.785 | 0.7869 | 0.1244 | 0.0623 |
| 75.6016 | 9.0 | 9000 | 76.3919 | 0.7827 | 0.3475 | 2.0858 | 0.7828 | 0.7849 | 0.1340 | 0.0657 |
| 75.4883 | 10.0 | 10000 | 76.5121 | 0.7768 | 0.3644 | 2.1372 | 0.7768 | 0.7747 | 0.1550 | 0.0662 |
| 75.2568 | 11.0 | 11000 | 76.4107 | 0.7857 | 0.3603 | 2.1229 | 0.7857 | 0.7863 | 0.1561 | 0.0619 |
| 75.1623 | 12.0 | 12000 | 76.4517 | 0.771 | 0.3857 | 2.1118 | 0.771 | 0.7721 | 0.1681 | 0.0702 |
| 75.0021 | 13.0 | 13000 | 76.3632 | 0.7885 | 0.3635 | 2.1178 | 0.7885 | 0.7870 | 0.1607 | 0.0621 |
| 74.9056 | 14.0 | 14000 | 76.3074 | 0.7925 | 0.3533 | 2.1361 | 0.7925 | 0.7926 | 0.1626 | 0.0573 |
| 74.9295 | 15.0 | 15000 | 76.3445 | 0.785 | 0.3730 | 2.0515 | 0.785 | 0.7861 | 0.1694 | 0.0661 |
| 74.7288 | 16.0 | 16000 | 76.3441 | 0.7845 | 0.3776 | 2.1216 | 0.7845 | 0.7828 | 0.1731 | 0.0666 |
| 74.5985 | 17.0 | 17000 | 76.1255 | 0.794 | 0.3593 | 2.0759 | 0.7940 | 0.7969 | 0.1640 | 0.0605 |
| 74.471 | 18.0 | 18000 | 76.2140 | 0.7863 | 0.3721 | 2.1872 | 0.7863 | 0.7861 | 0.1705 | 0.0638 |
| 74.4457 | 19.0 | 19000 | 76.1380 | 0.7925 | 0.3650 | 2.1106 | 0.7925 | 0.7940 | 0.1708 | 0.0634 |
| 74.3675 | 20.0 | 20000 | 76.1423 | 0.7897 | 0.3684 | 2.0882 | 0.7897 | 0.7910 | 0.1731 | 0.0642 |
| 74.3618 | 21.0 | 21000 | 76.0578 | 0.7987 | 0.3604 | 2.1007 | 0.7987 | 0.7982 | 0.1676 | 0.0622 |
| 74.1398 | 22.0 | 22000 | 75.9928 | 0.7997 | 0.3578 | 2.0590 | 0.7997 | 0.8008 | 0.1672 | 0.0624 |
| 74.0834 | 23.0 | 23000 | 75.8857 | 0.8013 | 0.3561 | 2.0986 | 0.8013 | 0.8010 | 0.1662 | 0.0602 |
| 74.1467 | 24.0 | 24000 | 75.8767 | 0.8 | 0.3605 | 2.0794 | 0.8000 | 0.8014 | 0.1682 | 0.0608 |
| 73.8823 | 25.0 | 25000 | 75.9471 | 0.799 | 0.3564 | 2.0934 | 0.799 | 0.7997 | 0.1684 | 0.0619 |
| 73.9657 | 26.0 | 26000 | 75.8618 | 0.7987 | 0.3594 | 2.1020 | 0.7987 | 0.7991 | 0.1703 | 0.0599 |
| 73.9721 | 27.0 | 27000 | 75.7331 | 0.8145 | 0.3347 | 2.0514 | 0.8145 | 0.8144 | 0.1569 | 0.0571 |
| 73.8298 | 28.0 | 28000 | 75.8175 | 0.8007 | 0.3582 | 2.0923 | 0.8007 | 0.7999 | 0.1714 | 0.0625 |
| 73.8483 | 29.0 | 29000 | 75.7541 | 0.8023 | 0.3554 | 2.1075 | 0.8023 | 0.8002 | 0.1698 | 0.0603 |
| 73.6726 | 30.0 | 30000 | 75.6642 | 0.8095 | 0.3454 | 2.0600 | 0.8095 | 0.8092 | 0.1638 | 0.0600 |
| 73.7118 | 31.0 | 31000 | 75.5905 | 0.8105 | 0.3398 | 2.1354 | 0.8105 | 0.8106 | 0.1587 | 0.0595 |
| 73.5938 | 32.0 | 32000 | 75.5721 | 0.8087 | 0.3429 | 2.0765 | 0.8087 | 0.8094 | 0.1640 | 0.0616 |
| 73.5563 | 33.0 | 33000 | 75.7021 | 0.8085 | 0.3474 | 2.0825 | 0.8085 | 0.8092 | 0.1656 | 0.0633 |
| 73.6469 | 34.0 | 34000 | 75.5322 | 0.8095 | 0.3406 | 2.0907 | 0.8095 | 0.8079 | 0.1632 | 0.0590 |
| 73.4666 | 35.0 | 35000 | 75.4994 | 0.8105 | 0.3397 | 2.0839 | 0.8105 | 0.8102 | 0.1621 | 0.0590 |
| 73.4144 | 36.0 | 36000 | 75.5095 | 0.8063 | 0.3476 | 2.1055 | 0.8062 | 0.8050 | 0.1666 | 0.0616 |
| 73.2744 | 37.0 | 37000 | 75.4980 | 0.8117 | 0.3403 | 2.0693 | 0.8117 | 0.8123 | 0.1607 | 0.0569 |
| 73.4358 | 38.0 | 38000 | 75.4824 | 0.809 | 0.3434 | 2.0996 | 0.809 | 0.8090 | 0.1645 | 0.0586 |
| 73.2696 | 39.0 | 39000 | 75.5088 | 0.8085 | 0.3468 | 2.0697 | 0.8085 | 0.8089 | 0.1658 | 0.0589 |
| 73.382 | 40.0 | 40000 | 75.4705 | 0.8095 | 0.3437 | 2.0738 | 0.8095 | 0.8104 | 0.1641 | 0.0621 |
| 73.3006 | 41.0 | 41000 | 75.4697 | 0.809 | 0.3440 | 2.1203 | 0.809 | 0.8097 | 0.1624 | 0.0614 |
| 73.4237 | 42.0 | 42000 | 75.3601 | 0.8093 | 0.3434 | 2.0736 | 0.8093 | 0.8094 | 0.1629 | 0.0575 |
| 73.2571 | 43.0 | 43000 | 75.3364 | 0.8103 | 0.3398 | 2.0665 | 0.8103 | 0.8101 | 0.1630 | 0.0599 |
| 73.2241 | 44.0 | 44000 | 75.3369 | 0.8135 | 0.3381 | 2.0609 | 0.8135 | 0.8136 | 0.1600 | 0.0581 |
| 73.2271 | 45.0 | 45000 | 75.2917 | 0.814 | 0.3355 | 2.0906 | 0.8140 | 0.8142 | 0.1576 | 0.0582 |
| 73.1427 | 46.0 | 46000 | 75.3108 | 0.8125 | 0.3377 | 2.0784 | 0.8125 | 0.8127 | 0.1613 | 0.0586 |
| 73.2754 | 47.0 | 47000 | 75.3195 | 0.8127 | 0.3386 | 2.0860 | 0.8128 | 0.8127 | 0.1604 | 0.0586 |
| 73.132 | 48.0 | 48000 | 75.3168 | 0.812 | 0.3391 | 2.0853 | 0.8120 | 0.8118 | 0.1612 | 0.0582 |
| 73.1482 | 49.0 | 49000 | 75.2943 | 0.8117 | 0.3395 | 2.0895 | 0.8117 | 0.8116 | 0.1615 | 0.0586 |
| 73.1849 | 50.0 | 50000 | 75.3146 | 0.811 | 0.3395 | 2.0856 | 0.811 | 0.8109 | 0.1625 | 0.0586 |
Framework versions
- Transformers 4.26.1
- Pytorch 1.13.1.post200
- Datasets 2.9.0
- Tokenizers 0.13.2
