swinv2-tiny-panorama-IQA
This model is a fine-tuned version of microsoft/swinv2-tiny-patch4-window8-256 on the isiqa-2019-hf dataset. It achieves the following results on the evaluation set:
- Loss: 0.0211
- Srocc: 0.0896
- Lcc: 0.2316
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: 32
- eval_batch_size: 32
- seed: 10
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 50.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Srocc | Lcc |
|---|---|---|---|---|---|
| No log | 0.8571 | 3 | 0.1747 | 0.1087 | 0.1976 |
| No log | 2.0 | 7 | 0.0570 | 0.1042 | 0.1898 |
| 0.1599 | 2.8571 | 10 | 0.0320 | 0.1126 | 0.1686 |
| 0.1599 | 4.0 | 14 | 0.0511 | 0.0265 | 0.1235 |
| 0.1599 | 4.8571 | 17 | 0.0274 | -0.0004 | 0.1029 |
| 0.0602 | 6.0 | 21 | 0.0375 | -0.0406 | 0.0900 |
| 0.0602 | 6.8571 | 24 | 0.0306 | -0.0533 | 0.0830 |
| 0.0602 | 8.0 | 28 | 0.0255 | -0.0726 | 0.0714 |
| 0.029 | 8.8571 | 31 | 0.0247 | -0.0568 | 0.0734 |
| 0.029 | 10.0 | 35 | 0.0293 | -0.0429 | 0.0900 |
| 0.029 | 10.8571 | 38 | 0.0259 | -0.0317 | 0.0982 |
| 0.0199 | 12.0 | 42 | 0.0238 | -0.0073 | 0.1288 |
| 0.0199 | 12.8571 | 45 | 0.0243 | 0.0216 | 0.1594 |
| 0.0199 | 14.0 | 49 | 0.0259 | 0.0454 | 0.1810 |
| 0.0161 | 14.8571 | 52 | 0.0224 | 0.0568 | 0.1954 |
| 0.0161 | 16.0 | 56 | 0.0211 | 0.0896 | 0.2316 |
| 0.0161 | 16.8571 | 59 | 0.0223 | 0.1001 | 0.2544 |
| 0.0132 | 18.0 | 63 | 0.0217 | 0.0981 | 0.2681 |
| 0.0132 | 18.8571 | 66 | 0.0221 | 0.1155 | 0.2746 |
| 0.0103 | 20.0 | 70 | 0.0228 | 0.1230 | 0.2831 |
| 0.0103 | 20.8571 | 73 | 0.0245 | 0.1327 | 0.2944 |
Framework versions
- Transformers 4.42.3
- Pytorch 2.1.2
- Datasets 2.20.0
- Tokenizers 0.19.1