swin-CEMEDE-og
This model is a fine-tuned version of microsoft/swin-base-simmim-window6-192 on the cemede dataset. It achieves the following results on the evaluation set:
- Loss: 0.8707
- Accuracy: 0.8018
- F1: 0.7127
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: 0.0002
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 |
|---|---|---|---|---|---|
| 2.2296 | 0.0840 | 100 | 2.1836 | 0.2570 | 0.0825 |
| 1.8921 | 0.1679 | 200 | 2.2063 | 0.3080 | 0.1300 |
| 1.682 | 0.2519 | 300 | 2.4050 | 0.3862 | 0.1449 |
| 1.8146 | 0.3359 | 400 | 1.9343 | 0.4253 | 0.2199 |
| 1.423 | 0.4198 | 500 | 1.9058 | 0.4579 | 0.2335 |
| 1.5637 | 0.5038 | 600 | 1.6756 | 0.5085 | 0.3331 |
| 1.0535 | 0.5877 | 700 | 1.4023 | 0.5641 | 0.3791 |
| 1.0955 | 0.6717 | 800 | 1.3086 | 0.6069 | 0.4426 |
| 0.8927 | 0.7557 | 900 | 1.2083 | 0.6377 | 0.5108 |
| 0.8035 | 0.8396 | 1000 | 1.3281 | 0.6340 | 0.4921 |
| 0.8517 | 0.9236 | 1100 | 1.2840 | 0.6492 | 0.5175 |
| 0.6035 | 1.0076 | 1200 | 1.2919 | 0.6446 | 0.5013 |
| 0.7727 | 1.0915 | 1300 | 1.0839 | 0.6878 | 0.5742 |
| 0.625 | 1.1755 | 1400 | 1.1132 | 0.7034 | 0.5552 |
| 0.554 | 1.2594 | 1500 | 1.2120 | 0.6492 | 0.5758 |
| 0.4117 | 1.3434 | 1600 | 1.1343 | 0.7030 | 0.5748 |
| 0.7557 | 1.4274 | 1700 | 1.1490 | 0.6975 | 0.5751 |
| 0.4841 | 1.5113 | 1800 | 0.9364 | 0.7756 | 0.6364 |
| 0.4899 | 1.5953 | 1900 | 1.1162 | 0.6929 | 0.5657 |
| 0.6598 | 1.6793 | 2000 | 0.9602 | 0.7402 | 0.6597 |
| 0.2826 | 1.7632 | 2100 | 1.2618 | 0.7044 | 0.6255 |
| 0.4785 | 1.8472 | 2200 | 1.0743 | 0.7269 | 0.6488 |
| 0.4427 | 1.9312 | 2300 | 0.8803 | 0.7641 | 0.6690 |
| 0.5305 | 2.0151 | 2400 | 0.8739 | 0.7830 | 0.6996 |
| 0.3814 | 2.0991 | 2500 | 0.9660 | 0.7789 | 0.6873 |
| 0.2273 | 2.1830 | 2600 | 1.0271 | 0.7789 | 0.7071 |
| 0.232 | 2.2670 | 2700 | 0.9957 | 0.7724 | 0.6961 |
| 0.2101 | 2.3510 | 2800 | 0.9729 | 0.7798 | 0.7196 |
| 0.4029 | 2.4349 | 2900 | 1.0296 | 0.7526 | 0.6911 |
| 0.2645 | 2.5189 | 3000 | 1.0878 | 0.7747 | 0.7058 |
| 0.3111 | 2.6029 | 3100 | 1.0745 | 0.7623 | 0.7072 |
| 0.1767 | 2.6868 | 3200 | 0.8820 | 0.7913 | 0.7424 |
| 0.167 | 2.7708 | 3300 | 0.8707 | 0.8018 | 0.7127 |
| 0.2523 | 2.8547 | 3400 | 1.0131 | 0.8046 | 0.7418 |
| 0.0786 | 2.9387 | 3500 | 1.0026 | 0.7807 | 0.7249 |
| 0.259 | 3.0227 | 3600 | 0.9817 | 0.7922 | 0.7109 |
| 0.3004 | 3.1066 | 3700 | 1.0838 | 0.7977 | 0.7341 |
| 0.1594 | 3.1906 | 3800 | 0.9184 | 0.8078 | 0.7323 |
| 0.1957 | 3.2746 | 3900 | 0.8777 | 0.8248 | 0.7255 |
| 0.1107 | 3.3585 | 4000 | 0.9186 | 0.8216 | 0.7360 |
| 0.1389 | 3.4425 | 4100 | 0.9996 | 0.8032 | 0.7358 |
| 0.1273 | 3.5264 | 4200 | 1.0062 | 0.8147 | 0.7604 |
| 0.2635 | 3.6104 | 4300 | 1.0976 | 0.8041 | 0.7406 |
Framework versions
- Transformers 4.55.4
- Pytorch 2.6.0+cu124
- Datasets 4.0.0
- Tokenizers 0.21.4