swinv2-tiny-patch4-window8-256-finetuned-galaxy10-decals
This model is a fine-tuned version of microsoft/swinv2-tiny-patch4-window8-256 on the matthieulel/galaxy10_decals dataset. It achieves the following results on the evaluation set:
- Loss: 0.4552
- Accuracy: 0.8551
- Precision: 0.8529
- Recall: 0.8551
- F1: 0.8513
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: 5e-05
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 30
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 |
|---|---|---|---|---|---|---|---|
| 1.7462 | 0.99 | 62 | 1.4592 | 0.4431 | 0.4309 | 0.4431 | 0.3967 |
| 1.1805 | 2.0 | 125 | 1.0335 | 0.6460 | 0.6741 | 0.6460 | 0.6241 |
| 0.9342 | 2.99 | 187 | 0.7051 | 0.7537 | 0.7478 | 0.7537 | 0.7394 |
| 0.786 | 4.0 | 250 | 0.6468 | 0.7745 | 0.7731 | 0.7745 | 0.7637 |
| 0.7062 | 4.99 | 312 | 0.6013 | 0.8038 | 0.8052 | 0.8038 | 0.8008 |
| 0.7011 | 6.0 | 375 | 0.5373 | 0.8123 | 0.8171 | 0.8123 | 0.8041 |
| 0.7014 | 6.99 | 437 | 0.5470 | 0.8044 | 0.8048 | 0.8044 | 0.7995 |
| 0.6447 | 8.0 | 500 | 0.5309 | 0.8083 | 0.8087 | 0.8083 | 0.8025 |
| 0.608 | 8.99 | 562 | 0.4836 | 0.8337 | 0.8323 | 0.8337 | 0.8300 |
| 0.6196 | 10.0 | 625 | 0.4797 | 0.8331 | 0.8293 | 0.8331 | 0.8268 |
| 0.6031 | 10.99 | 687 | 0.4863 | 0.8264 | 0.8274 | 0.8264 | 0.8239 |
| 0.5462 | 12.0 | 750 | 0.4749 | 0.8354 | 0.8341 | 0.8354 | 0.8313 |
| 0.5868 | 12.99 | 812 | 0.5269 | 0.8236 | 0.8268 | 0.8236 | 0.8171 |
| 0.5844 | 14.0 | 875 | 0.4402 | 0.8472 | 0.8447 | 0.8472 | 0.8430 |
| 0.5326 | 14.99 | 937 | 0.4635 | 0.8393 | 0.8359 | 0.8393 | 0.8353 |
| 0.5313 | 16.0 | 1000 | 0.4734 | 0.8365 | 0.8345 | 0.8365 | 0.8300 |
| 0.4893 | 16.99 | 1062 | 0.4675 | 0.8365 | 0.8335 | 0.8365 | 0.8316 |
| 0.4983 | 18.0 | 1125 | 0.4441 | 0.8444 | 0.8431 | 0.8444 | 0.8401 |
| 0.518 | 18.99 | 1187 | 0.4693 | 0.8416 | 0.8441 | 0.8416 | 0.8376 |
| 0.5228 | 20.0 | 1250 | 0.4732 | 0.8410 | 0.8379 | 0.8410 | 0.8358 |
| 0.4761 | 20.99 | 1312 | 0.4567 | 0.8489 | 0.8493 | 0.8489 | 0.8460 |
| 0.5311 | 22.0 | 1375 | 0.4582 | 0.8484 | 0.8469 | 0.8484 | 0.8433 |
| 0.4894 | 22.99 | 1437 | 0.4627 | 0.8467 | 0.8450 | 0.8467 | 0.8433 |
| 0.4791 | 24.0 | 1500 | 0.4580 | 0.8506 | 0.8493 | 0.8506 | 0.8481 |
| 0.479 | 24.99 | 1562 | 0.4625 | 0.8472 | 0.8443 | 0.8472 | 0.8433 |
| 0.4487 | 26.0 | 1625 | 0.4557 | 0.8495 | 0.8469 | 0.8495 | 0.8447 |
| 0.4515 | 26.99 | 1687 | 0.4501 | 0.8534 | 0.8510 | 0.8534 | 0.8500 |
| 0.4862 | 28.0 | 1750 | 0.4552 | 0.8551 | 0.8529 | 0.8551 | 0.8513 |
| 0.4348 | 28.99 | 1812 | 0.4512 | 0.8506 | 0.8486 | 0.8506 | 0.8469 |
| 0.4623 | 29.76 | 1860 | 0.4539 | 0.8551 | 0.8533 | 0.8551 | 0.8516 |
Framework versions
- Transformers 4.37.2
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.15.1