resnet101_rvl-cdip
This model is a fine-tuned version of microsoft/resnet-101 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.6158
- Accuracy: 0.8210
- Brier Loss: 0.2556
- Nll: 1.7696
- F1 Micro: 0.8210
- F1 Macro: 0.8209
- Ece: 0.0176
- Aurc: 0.0418
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: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 10
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Brier Loss | Nll | F1 Micro | F1 Macro | Ece | Aurc |
|---|---|---|---|---|---|---|---|---|---|---|
| 1.3521 | 1.0 | 5000 | 1.2626 | 0.6133 | 0.5108 | 2.7262 | 0.6133 | 0.6042 | 0.0455 | 0.1644 |
| 0.942 | 2.0 | 10000 | 0.9005 | 0.7318 | 0.3723 | 2.2139 | 0.7318 | 0.7293 | 0.0174 | 0.0862 |
| 0.7983 | 3.0 | 15000 | 0.7691 | 0.7723 | 0.3198 | 2.0444 | 0.7723 | 0.7714 | 0.0139 | 0.0641 |
| 0.7167 | 4.0 | 20000 | 0.7048 | 0.7924 | 0.2931 | 1.9414 | 0.7924 | 0.7931 | 0.0135 | 0.0541 |
| 0.6656 | 5.0 | 25000 | 0.6658 | 0.8052 | 0.2770 | 1.8581 | 0.8052 | 0.8056 | 0.0108 | 0.0486 |
| 0.6252 | 6.0 | 30000 | 0.6415 | 0.8117 | 0.2670 | 1.8157 | 0.8117 | 0.8112 | 0.0128 | 0.0455 |
| 0.6038 | 7.0 | 35000 | 0.6269 | 0.8176 | 0.2607 | 1.7833 | 0.8176 | 0.8180 | 0.0144 | 0.0432 |
| 0.5784 | 8.0 | 40000 | 0.6217 | 0.8195 | 0.2583 | 1.7723 | 0.8195 | 0.8195 | 0.0151 | 0.0425 |
| 0.5583 | 9.0 | 45000 | 0.6150 | 0.8214 | 0.2553 | 1.7719 | 0.8214 | 0.8214 | 0.0164 | 0.0415 |
| 0.5519 | 10.0 | 50000 | 0.6158 | 0.8210 | 0.2556 | 1.7696 | 0.8210 | 0.8209 | 0.0176 | 0.0418 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.2.0.dev20231002
- Datasets 2.7.1
- Tokenizers 0.13.3