vit-lr-inverse-sqrt
This model is a fine-tuned version of google/vit-base-patch16-224 on the skin-cancer dataset. It achieves the following results on the evaluation set:
- Loss: 0.4469
- Accuracy: 0.8499
- Precision: 0.8565
- Recall: 0.8499
- F1: 0.8516
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: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: inverse_sqrt
- lr_scheduler_warmup_steps: 80
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 |
---|---|---|---|---|---|---|---|
0.6694 | 0.31 | 100 | 0.6511 | 0.7632 | 0.7558 | 0.7632 | 0.7499 |
0.5468 | 0.62 | 200 | 0.6537 | 0.7618 | 0.7713 | 0.7618 | 0.7109 |
0.6132 | 0.93 | 300 | 0.5132 | 0.8145 | 0.8262 | 0.8145 | 0.8138 |
0.3319 | 1.25 | 400 | 0.4706 | 0.8308 | 0.8327 | 0.8308 | 0.8293 |
0.2286 | 1.56 | 500 | 0.4952 | 0.8353 | 0.8447 | 0.8353 | 0.8226 |
0.2299 | 1.87 | 600 | 0.4696 | 0.8367 | 0.8517 | 0.8367 | 0.8358 |
0.0542 | 2.18 | 700 | 0.4469 | 0.8499 | 0.8565 | 0.8499 | 0.8516 |
0.198 | 2.49 | 800 | 0.5285 | 0.8225 | 0.8616 | 0.8225 | 0.8323 |
0.0311 | 2.8 | 900 | 0.4724 | 0.8651 | 0.8687 | 0.8651 | 0.8662 |
0.0543 | 3.12 | 1000 | 0.4949 | 0.8665 | 0.8612 | 0.8665 | 0.8611 |
0.0242 | 3.43 | 1100 | 0.6283 | 0.8623 | 0.8661 | 0.8623 | 0.8510 |
0.0179 | 3.74 | 1200 | 0.5766 | 0.8724 | 0.8681 | 0.8724 | 0.8675 |
0.01 | 4.05 | 1300 | 0.6232 | 0.8596 | 0.8523 | 0.8596 | 0.8535 |
0.0018 | 4.36 | 1400 | 0.6013 | 0.8741 | 0.8707 | 0.8741 | 0.8710 |
0.0019 | 4.67 | 1500 | 0.6554 | 0.8682 | 0.8689 | 0.8682 | 0.8643 |
0.0024 | 4.98 | 1600 | 0.6107 | 0.8714 | 0.8730 | 0.8714 | 0.8719 |
0.0006 | 5.3 | 1700 | 0.6353 | 0.8755 | 0.8751 | 0.8755 | 0.8725 |
Framework versions
- Transformers 4.39.0.dev0
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 1
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.