| --- |
| library_name: transformers |
| tags: |
| - generated_from_trainer |
| metrics: |
| - wer |
| model-index: |
| - name: iteboshi |
| results: [] |
| --- |
| |
| <!-- This model card has been generated automatically according to the information the Trainer had access to. You |
| should probably proofread and complete it, then remove this comment. --> |
|
|
| # iteboshi |
|
|
| This model was trained from scratch on an unknown dataset. |
| It achieves the following results on the evaluation set: |
| - Loss: 0.8947 |
| - Wer: 82.3479 |
| - Cer: 22.6268 |
|
|
| ## 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: 4 |
| - eval_batch_size: 4 |
| - seed: 42 |
| - distributed_type: multi-GPU |
| - gradient_accumulation_steps: 8 |
| - total_train_batch_size: 32 |
| - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments |
| - lr_scheduler_type: linear |
| - lr_scheduler_warmup_steps: 500 |
| - training_steps: 20000 |
| |
| ### Training results |
| |
| | Training Loss | Epoch | Step | Validation Loss | Wer | Cer | |
| |:-------------:|:-------:|:-----:|:---------------:|:-------:|:-------:| |
| | 1.0854 | 1.1013 | 1000 | 1.2534 | 97.5672 | 52.7088 | |
| | 0.5859 | 2.2026 | 2000 | 0.8996 | 90.9477 | 48.1097 | |
| | 0.3373 | 3.3040 | 3000 | 0.7766 | 87.7699 | 29.9950 | |
| | 0.2445 | 4.4053 | 4000 | 0.7662 | 86.6761 | 28.1264 | |
| | 0.1548 | 5.5066 | 5000 | 0.7709 | 86.6007 | 27.8748 | |
| | 0.1102 | 6.6079 | 6000 | 0.7889 | 86.3178 | 26.2934 | |
| | 0.0682 | 7.7093 | 7000 | 0.7991 | 84.4507 | 27.3578 | |
| | 0.0647 | 8.8106 | 8000 | 0.8132 | 84.6488 | 25.6262 | |
| | 0.0343 | 9.9119 | 9000 | 0.8282 | 84.8279 | 24.6948 | |
| | 0.0181 | 11.0132 | 10000 | 0.8396 | 83.8001 | 24.3618 | |
| | 0.0117 | 12.1145 | 11000 | 0.8592 | 84.1584 | 24.0030 | |
| | 0.0111 | 13.2159 | 12000 | 0.8610 | 83.8378 | 24.3537 | |
| | 0.0088 | 14.3172 | 13000 | 0.8743 | 84.0924 | 24.6323 | |
| | 0.0112 | 15.4185 | 14000 | 0.8769 | 84.1867 | 24.9344 | |
| | 0.0109 | 16.5198 | 15000 | 0.8774 | 84.6770 | 24.6214 | |
| | 0.0032 | 17.6211 | 16000 | 0.8810 | 82.6591 | 23.3174 | |
| | 0.0017 | 18.7225 | 17000 | 0.8870 | 82.9986 | 22.8532 | |
| | 0.0019 | 19.8238 | 18000 | 0.8900 | 82.5083 | 22.6634 | |
| | 0.0008 | 20.9251 | 19000 | 0.8924 | 82.4800 | 22.5878 | |
| | 0.0006 | 22.0264 | 20000 | 0.8947 | 82.3479 | 22.6268 | |
| |
| |
| ### Framework versions |
| |
| - Transformers 4.48.3 |
| - Pytorch 2.6.0+cu124 |
| - Datasets 3.5.0 |
| - Tokenizers 0.21.1 |
| |
