| --- |
| library_name: transformers |
| tags: |
| - generated_from_trainer |
| metrics: |
| - wer |
| model-index: |
| - name: iteboshi-small |
| 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-small |
|
|
| This model was trained from scratch on an unknown dataset. |
| It achieves the following results on the evaluation set: |
| - Loss: 0.9635 |
| - Wer: 94.6252 |
| - Cer: 50.3003 |
|
|
| ## 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 | |
| |:-------------:|:-------:|:-----:|:---------------:|:-------:|:-------:| |
| | 1.1451 | 1.1013 | 1000 | 1.3127 | 98.0009 | 52.9954 | |
| | 0.6729 | 2.2026 | 2000 | 0.9211 | 94.9364 | 36.0076 | |
| | 0.4087 | 3.3040 | 3000 | 0.8170 | 91.4757 | 33.0592 | |
| | 0.303 | 4.4053 | 4000 | 0.7996 | 95.9642 | 35.4912 | |
| | 0.211 | 5.5066 | 5000 | 0.7910 | 90.7685 | 39.8708 | |
| | 0.1389 | 6.6079 | 6000 | 0.8133 | 91.2588 | 46.3311 | |
| | 0.0864 | 7.7093 | 7000 | 0.8312 | 92.6638 | 39.9178 | |
| | 0.0729 | 8.8106 | 8000 | 0.8530 | 91.6172 | 50.7434 | |
| | 0.0381 | 9.9119 | 9000 | 0.8698 | 91.5700 | 47.7159 | |
| | 0.028 | 11.0132 | 10000 | 0.8864 | 92.1452 | 54.3756 | |
| | 0.0142 | 12.1145 | 11000 | 0.8988 | 93.2107 | 53.6414 | |
| | 0.0131 | 13.2159 | 12000 | 0.9192 | 92.8053 | 46.2153 | |
| | 0.0088 | 14.3172 | 13000 | 0.9230 | 93.8142 | 54.3103 | |
| | 0.0092 | 15.4185 | 14000 | 0.9310 | 94.0311 | 53.1192 | |
| | 0.0069 | 16.5198 | 15000 | 0.9370 | 93.8802 | 51.9775 | |
| | 0.0023 | 17.6211 | 16000 | 0.9437 | 94.2386 | 49.6876 | |
| | 0.0026 | 18.7225 | 17000 | 0.9495 | 94.0971 | 51.5929 | |
| | 0.0016 | 19.8238 | 18000 | 0.9531 | 94.1631 | 51.0942 | |
| | 0.0015 | 20.9251 | 19000 | 0.9608 | 94.7006 | 50.9125 | |
| | 0.0012 | 22.0264 | 20000 | 0.9635 | 94.6252 | 50.3003 | |
| |
| |
| ### Framework versions |
| |
| - Transformers 4.48.3 |
| - Pytorch 2.7.0+cu128 |
| - Datasets 3.6.0 |
| - Tokenizers 0.21.1 |
| |
