swin-tiny-patch4-window7-224-finetuned-aiornot-step
This model is a fine-tuned version of microsoft/swin-tiny-patch4-window7-224 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1075
- Accuracy: 0.9613
- F1: 0.9606
- Log Loss: 1.3937
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: 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
- lr_scheduler_warmup_steps: 10
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Log Loss |
|---|---|---|---|---|---|---|
| 0.5919 | 0.17 | 10 | 0.3100 | 0.8722 | 0.8663 | 4.6071 |
| 0.3175 | 0.34 | 20 | 0.2757 | 0.8861 | 0.8793 | 4.1038 |
| 0.2445 | 0.52 | 30 | 0.2021 | 0.9130 | 0.9092 | 3.1359 |
| 0.2117 | 0.69 | 40 | 0.1510 | 0.9377 | 0.9360 | 2.2455 |
| 0.2062 | 0.86 | 50 | 0.1535 | 0.9334 | 0.9311 | 2.4003 |
| 0.1815 | 1.03 | 60 | 0.1493 | 0.9382 | 0.9361 | 2.2261 |
| 0.1458 | 1.2 | 70 | 0.2196 | 0.9135 | 0.9089 | 3.1166 |
| 0.1507 | 1.37 | 80 | 0.1826 | 0.9243 | 0.9207 | 2.7294 |
| 0.1389 | 1.55 | 90 | 0.1396 | 0.9377 | 0.9356 | 2.2455 |
| 0.1331 | 1.72 | 100 | 0.2182 | 0.9125 | 0.9077 | 3.1553 |
| 0.1458 | 1.89 | 110 | 0.2661 | 0.9033 | 0.8978 | 3.4843 |
| 0.147 | 2.06 | 120 | 0.1068 | 0.9597 | 0.9588 | 1.4518 |
| 0.119 | 2.23 | 130 | 0.1284 | 0.9501 | 0.9484 | 1.8002 |
| 0.1045 | 2.4 | 140 | 0.1802 | 0.9291 | 0.9259 | 2.5552 |
| 0.1115 | 2.58 | 150 | 0.1434 | 0.9441 | 0.9420 | 2.0132 |
| 0.1108 | 2.75 | 160 | 0.1365 | 0.9441 | 0.9421 | 2.0132 |
| 0.1044 | 2.92 | 170 | 0.1488 | 0.9398 | 0.9375 | 2.1680 |
Framework versions
- Transformers 4.30.2
- Pytorch 2.0.0
- Datasets 2.14.4
- Tokenizers 0.13.3