deit_fold_2_v3
This model is a fine-tuned version of facebook/deit-small-patch16-224 on the imagefolder dataset. It achieves the following results on the evaluation set:
- Loss: 1.1298
- Accuracy: 0.9519
- F1 Score: 0.9537
- Recall: 0.9579
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: 1e-05
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 150
- num_epochs: 100
- label_smoothing_factor: 0.15
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 Score | Recall |
|---|---|---|---|---|---|---|
| 2.8252 | 1.0 | 20 | 2.8294 | 0.2532 | 0.1757 | 0.2799 |
| 2.6170 | 2.0 | 40 | 2.6009 | 0.4231 | 0.3643 | 0.4176 |
| 2.3206 | 3.0 | 60 | 2.3003 | 0.6282 | 0.6023 | 0.6150 |
| 1.9316 | 4.0 | 80 | 1.9340 | 0.7468 | 0.7536 | 0.7418 |
| 1.6434 | 5.0 | 100 | 1.6591 | 0.8301 | 0.8350 | 0.8313 |
| 1.4153 | 6.0 | 120 | 1.4709 | 0.8718 | 0.8757 | 0.8768 |
| 1.2545 | 7.0 | 140 | 1.3896 | 0.8846 | 0.8842 | 0.8805 |
| 1.2233 | 8.0 | 160 | 1.3060 | 0.9006 | 0.9020 | 0.8985 |
| 1.1294 | 9.0 | 180 | 1.2383 | 0.9295 | 0.9305 | 0.9327 |
| 1.1183 | 10.0 | 200 | 1.2321 | 0.9167 | 0.9188 | 0.9187 |
| 1.1030 | 11.0 | 220 | 1.2141 | 0.9263 | 0.9277 | 0.9273 |
| 1.0461 | 12.0 | 240 | 1.1977 | 0.9327 | 0.9341 | 0.9346 |
| 1.0356 | 13.0 | 260 | 1.1742 | 0.9327 | 0.9345 | 0.9359 |
| 1.0387 | 14.0 | 280 | 1.1660 | 0.9295 | 0.9322 | 0.9366 |
| 1.0136 | 15.0 | 300 | 1.1681 | 0.9295 | 0.9318 | 0.9359 |
| 1.0312 | 16.0 | 320 | 1.1628 | 0.9295 | 0.9316 | 0.9346 |
| 1.0127 | 17.0 | 340 | 1.1365 | 0.9391 | 0.9409 | 0.9420 |
| 1.0043 | 18.0 | 360 | 1.1322 | 0.9423 | 0.9439 | 0.9444 |
| 0.9970 | 19.0 | 380 | 1.1653 | 0.9455 | 0.9473 | 0.9532 |
| 1.0022 | 20.0 | 400 | 1.1536 | 0.9359 | 0.9380 | 0.9415 |
| 1.0002 | 21.0 | 420 | 1.1603 | 0.9391 | 0.9412 | 0.9469 |
| 1.0062 | 22.0 | 440 | 1.1409 | 0.9455 | 0.9470 | 0.9518 |
| 0.9826 | 23.0 | 460 | 1.1523 | 0.9423 | 0.9437 | 0.9415 |
| 0.9858 | 24.0 | 480 | 1.1611 | 0.9455 | 0.9472 | 0.9513 |
| 0.9753 | 25.0 | 500 | 1.1412 | 0.9455 | 0.9472 | 0.9506 |
| 0.9730 | 26.0 | 520 | 1.1605 | 0.9423 | 0.9441 | 0.9427 |
| 0.9833 | 27.0 | 540 | 1.1281 | 0.9519 | 0.9531 | 0.9555 |
| 0.9765 | 28.0 | 560 | 1.1411 | 0.9487 | 0.9503 | 0.9501 |
| 0.9791 | 29.0 | 580 | 1.1365 | 0.9487 | 0.9511 | 0.9555 |
| 0.9748 | 30.0 | 600 | 1.1481 | 0.9391 | 0.9415 | 0.9450 |
| 0.9725 | 31.0 | 620 | 1.1567 | 0.9359 | 0.9370 | 0.9366 |
| 0.9645 | 32.0 | 640 | 1.1298 | 0.9519 | 0.9537 | 0.9579 |
| 0.9599 | 33.0 | 660 | 1.1275 | 0.9487 | 0.9500 | 0.9518 |
| 0.9701 | 34.0 | 680 | 1.1281 | 0.9519 | 0.9530 | 0.9543 |
| 0.9785 | 35.0 | 700 | 1.1293 | 0.9455 | 0.9472 | 0.9506 |
| 0.9714 | 36.0 | 720 | 1.1466 | 0.9455 | 0.9473 | 0.9513 |
| 0.9687 | 37.0 | 740 | 1.1502 | 0.9423 | 0.9447 | 0.9501 |
| 0.9581 | 38.0 | 760 | 1.1606 | 0.9391 | 0.9415 | 0.9464 |
| 0.9675 | 39.0 | 780 | 1.1502 | 0.9423 | 0.9442 | 0.9462 |
| 0.9697 | 40.0 | 800 | 1.1641 | 0.9391 | 0.9414 | 0.9476 |
| 0.9664 | 41.0 | 820 | 1.1442 | 0.9455 | 0.9475 | 0.9523 |
| 0.9567 | 42.0 | 840 | 1.1395 | 0.9423 | 0.9441 | 0.9462 |
| 0.9716 | 43.0 | 860 | 1.1537 | 0.9359 | 0.9380 | 0.9415 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 5.0.0
- Tokenizers 0.22.2