vit-base-patch16-224-finetuned-on-all-affectnet_short
This model is a fine-tuned version of google/vit-base-patch16-224 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.0520
- Accuracy: 0.7259
- Precision: 0.7293
- Recall: 0.7259
- F1: 0.7255
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: 2e-05
- train_batch_size: 64
- eval_batch_size: 16
- 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: 14
- label_smoothing_factor: 0.1
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 |
|---|---|---|---|---|---|---|---|
| 1.5399 | 1.0 | 91 | 1.4698 | 0.5097 | 0.5079 | 0.5097 | 0.4827 |
| 1.284 | 2.0 | 182 | 1.2026 | 0.6409 | 0.6514 | 0.6409 | 0.6226 |
| 1.2259 | 3.0 | 273 | 1.1367 | 0.6722 | 0.6749 | 0.6722 | 0.6694 |
| 1.1663 | 4.0 | 364 | 1.1086 | 0.6838 | 0.6903 | 0.6838 | 0.6814 |
| 1.1401 | 5.0 | 455 | 1.0782 | 0.7055 | 0.7070 | 0.7055 | 0.7042 |
| 1.1229 | 6.0 | 546 | 1.0734 | 0.7055 | 0.7093 | 0.7055 | 0.7036 |
| 1.0929 | 7.0 | 637 | 1.0674 | 0.7120 | 0.7147 | 0.7120 | 0.7099 |
| 1.0826 | 8.0 | 728 | 1.0601 | 0.7210 | 0.7226 | 0.7210 | 0.7191 |
| 1.0414 | 9.0 | 819 | 1.0558 | 0.7203 | 0.7211 | 0.7203 | 0.7196 |
| 1.0649 | 10.0 | 910 | 1.0499 | 0.7179 | 0.7194 | 0.7179 | 0.7175 |
| 1.0554 | 11.0 | 1001 | 1.0520 | 0.7259 | 0.7293 | 0.7259 | 0.7255 |
| 1.0496 | 12.0 | 1092 | 1.0466 | 0.7210 | 0.7212 | 0.7210 | 0.7204 |
| 1.064 | 13.0 | 1183 | 1.0502 | 0.7220 | 0.7235 | 0.7220 | 0.7211 |
| 1.0386 | 14.0 | 1274 | 1.0475 | 0.7206 | 0.7208 | 0.7206 | 0.7200 |
Framework versions
- Transformers 4.29.1
- Pytorch 2.0.0+cu118
- Datasets 2.12.0
- Tokenizers 0.13.3