vit-base-patch16-224-in21k-rotated-dungeons-v001
This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the rotated_maps dataset. It achieves the following results on the evaluation set:
- Loss: 1.1914
- Accuracy: 0.8214
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: 8
- eval_batch_size: 8
- seed: 1024
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 25
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| No log | 1.0 | 6 | 1.4924 | 0.5 |
| No log | 2.0 | 12 | 1.4798 | 0.5 |
| No log | 3.0 | 18 | 1.4518 | 0.5357 |
| No log | 4.0 | 24 | 1.4402 | 0.5357 |
| No log | 5.0 | 30 | 1.4052 | 0.5357 |
| No log | 6.0 | 36 | 1.3828 | 0.6786 |
| No log | 7.0 | 42 | 1.3588 | 0.6786 |
| No log | 8.0 | 48 | 1.3295 | 0.6786 |
| No log | 9.0 | 54 | 1.3263 | 0.7143 |
| No log | 10.0 | 60 | 1.3072 | 0.75 |
| No log | 11.0 | 66 | 1.2918 | 0.7143 |
| No log | 12.0 | 72 | 1.2718 | 0.8214 |
| No log | 13.0 | 78 | 1.2728 | 0.7857 |
| No log | 14.0 | 84 | 1.2628 | 0.75 |
| No log | 15.0 | 90 | 1.2333 | 0.7857 |
| No log | 16.0 | 96 | 1.2253 | 0.7857 |
| No log | 17.0 | 102 | 1.2240 | 0.7857 |
| No log | 18.0 | 108 | 1.2249 | 0.7857 |
| No log | 19.0 | 114 | 1.2177 | 0.7857 |
| No log | 20.0 | 120 | 1.2098 | 0.7857 |
| No log | 21.0 | 126 | 1.2029 | 0.8214 |
| No log | 22.0 | 132 | 1.1875 | 0.8571 |
| No log | 23.0 | 138 | 1.1873 | 0.8571 |
| No log | 24.0 | 144 | 1.2051 | 0.7857 |
| No log | 25.0 | 150 | 1.1914 | 0.8214 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.5.0+cu121
- Datasets 3.1.0
- Tokenizers 0.19.1