| --- |
| library_name: transformers |
| tags: |
| - generated_from_trainer |
| metrics: |
| - wer |
| model-index: |
| - name: iteboshi-medium |
| 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-medium |
|
|
| This model was trained from scratch on an unknown dataset. |
| It achieves the following results on the evaluation set: |
| - Loss: 0.8963 |
| - Wer: 81.7162 |
| - Cer: 22.2135 |
|
|
| ## 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 | |
| |:-------------:|:-------:|:-----:|:---------------:|:-------:|:-------:| |
| | 1.0877 | 1.1013 | 1000 | 1.2933 | 98.3781 | 54.0409 | |
| | 0.6047 | 2.2026 | 2000 | 0.8662 | 91.6643 | 34.1946 | |
| | 0.3394 | 3.3040 | 3000 | 0.7763 | 88.5997 | 34.5534 | |
| | 0.243 | 4.4053 | 4000 | 0.7650 | 86.5912 | 26.8098 | |
| | 0.1633 | 5.5066 | 5000 | 0.7654 | 88.1848 | 27.4116 | |
| | 0.1113 | 6.6079 | 6000 | 0.7906 | 86.4215 | 26.7640 | |
| | 0.0673 | 7.7093 | 7000 | 0.7989 | 84.6110 | 27.1371 | |
| | 0.0624 | 8.8106 | 8000 | 0.8190 | 84.6582 | 24.9533 | |
| | 0.0355 | 9.9119 | 9000 | 0.8439 | 84.0924 | 24.8639 | |
| | 0.0262 | 11.0132 | 10000 | 0.8546 | 84.7525 | 25.5190 | |
| | 0.0146 | 12.1145 | 11000 | 0.8571 | 83.5832 | 23.8711 | |
| | 0.0089 | 13.2159 | 12000 | 0.8546 | 82.8666 | 23.3656 | |
| | 0.0098 | 14.3172 | 13000 | 0.8761 | 83.8850 | 23.7055 | |
| | 0.0076 | 15.4185 | 14000 | 0.8775 | 83.1872 | 23.6654 | |
| | 0.0044 | 16.5198 | 15000 | 0.8781 | 83.0740 | 23.5868 | |
| | 0.005 | 17.6211 | 16000 | 0.8774 | 82.3102 | 22.7488 | |
| | 0.0026 | 18.7225 | 17000 | 0.8914 | 82.1499 | 22.5316 | |
| | 0.0015 | 19.8238 | 18000 | 0.8890 | 81.9896 | 22.3132 | |
| | 0.0011 | 20.9251 | 19000 | 0.8928 | 81.6219 | 22.2513 | |
| | 0.0006 | 22.0264 | 20000 | 0.8963 | 81.7162 | 22.2135 | |
| |
| |
| ### Framework versions |
| |
| - Transformers 4.48.3 |
| - Pytorch 2.7.0+cu128 |
| - Datasets 3.6.0 |
| - Tokenizers 0.21.1 |
| |
