finetuned-beit-limb-seq-t2-8heads-1layers-2.5e-4lr
This model is a fine-tuned version of c14kevincardenas/beit-large-patch16-384-limb-person-crop on the c14kevincardenas/beta_caller_284_person_crop_seq_withlimb_2 dataset. It achieves the following results on the evaluation set:
- Loss: 0.5353
- Accuracy: 0.8847
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: 0.00025
- train_batch_size: 16
- eval_batch_size: 16
- seed: 2014
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 250
- num_epochs: 25.0
- mixed_precision_training: Native AMP
- label_smoothing_factor: 0.05
Training results
Training Loss | Epoch | Step | Validation Loss | Accuracy |
---|---|---|---|---|
0.8925 | 1.0 | 362 | 0.7904 | 0.7449 |
0.858 | 2.0 | 724 | 0.8538 | 0.7175 |
0.8091 | 3.0 | 1086 | 0.9683 | 0.7019 |
0.7739 | 4.0 | 1448 | 0.7409 | 0.7957 |
0.7614 | 5.0 | 1810 | 0.7168 | 0.8035 |
0.8067 | 6.0 | 2172 | 0.7570 | 0.8006 |
0.749 | 7.0 | 2534 | 0.8142 | 0.7634 |
0.7555 | 8.0 | 2896 | 0.6704 | 0.8319 |
0.6637 | 9.0 | 3258 | 0.6815 | 0.8231 |
0.6502 | 10.0 | 3620 | 0.6419 | 0.8543 |
0.6432 | 11.0 | 3982 | 0.6516 | 0.8504 |
0.6431 | 12.0 | 4344 | 0.6119 | 0.8631 |
0.6504 | 13.0 | 4706 | 0.5947 | 0.8631 |
0.5558 | 14.0 | 5068 | 0.5988 | 0.8602 |
0.53 | 15.0 | 5430 | 0.5833 | 0.8700 |
0.6418 | 16.0 | 5792 | 0.6134 | 0.8710 |
0.5703 | 17.0 | 6154 | 0.6183 | 0.8700 |
0.5244 | 18.0 | 6516 | 0.5692 | 0.8680 |
0.5311 | 19.0 | 6878 | 0.5639 | 0.8661 |
0.5504 | 20.0 | 7240 | 0.5518 | 0.8837 |
0.4922 | 21.0 | 7602 | 0.5777 | 0.8671 |
0.4801 | 22.0 | 7964 | 0.5549 | 0.8778 |
0.5085 | 23.0 | 8326 | 0.5502 | 0.8768 |
0.5002 | 24.0 | 8688 | 0.5445 | 0.8856 |
0.4404 | 25.0 | 9050 | 0.5353 | 0.8847 |
Framework versions
- Transformers 4.43.0.dev0
- Pytorch 2.0.1+cu117
- Datasets 2.19.1
- Tokenizers 0.19.1