metadata
base_model: microsoft/git-base
library_name: transformers
license: mit
tags:
- generated_from_trainer
model-index:
- name: git-base-bdd100k
results: []
git-base-bdd100k
This model is a fine-tuned version of microsoft/git-base on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.4317
- Wer Score: 0.7406
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: 0.0003
- train_batch_size: 10
- eval_batch_size: 10
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 20
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 15
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Score |
|---|---|---|---|---|
| 10.8372 | 1.0 | 3 | 10.0676 | 12.2594 |
| 9.8981 | 2.0 | 6 | 8.6733 | 9.4194 |
| 8.1063 | 3.0 | 9 | 6.8677 | 0.8955 |
| 6.5282 | 4.0 | 12 | 5.5489 | 3.5342 |
| 5.2838 | 5.0 | 15 | 4.3715 | 2.2077 |
| 4.16 | 6.0 | 18 | 3.3203 | 3.32 |
| 3.1554 | 7.0 | 21 | 2.3917 | 1.5897 |
| 2.2691 | 8.0 | 24 | 1.6361 | 0.7832 |
| 1.553 | 9.0 | 27 | 1.1034 | 0.7703 |
| 1.0453 | 10.0 | 30 | 0.7820 | 0.7781 |
| 0.7256 | 11.0 | 33 | 0.6073 | 0.7703 |
| 0.5425 | 12.0 | 36 | 0.5168 | 0.7548 |
| 0.4393 | 13.0 | 39 | 0.4689 | 0.7419 |
| 0.3801 | 14.0 | 42 | 0.4449 | 0.7445 |
| 0.3404 | 15.0 | 45 | 0.4317 | 0.7406 |
Framework versions
- Transformers 4.46.0.dev0
- Pytorch 2.0.1+cu117
- Datasets 3.0.1
- Tokenizers 0.20.1