| --- |
| library_name: transformers |
| tags: |
| - generated_from_trainer |
| metrics: |
| - wer |
| model-index: |
| - name: iteboshi-tiny |
| 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-tiny |
|
|
| This model was trained from scratch on an unknown dataset. |
| It achieves the following results on the evaluation set: |
| - Loss: 0.9326 |
| - Wer: 115.2570 |
| - Cer: 50.1238 |
|
|
| ## 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 | |
| |:-------------:|:-------:|:-----:|:---------------:|:--------:|:--------:| |
| | 0.5915 | 1.1013 | 1000 | 0.7334 | 158.2838 | 65.0516 | |
| | 0.4754 | 2.2026 | 2000 | 0.6800 | 176.7751 | 66.8021 | |
| | 0.3484 | 3.3040 | 3000 | 0.6674 | 250.1933 | 86.5160 | |
| | 0.3012 | 4.4053 | 4000 | 0.6733 | 390.6648 | 143.7552 | |
| | 0.2416 | 5.5066 | 5000 | 0.6857 | 259.8491 | 89.0706 | |
| | 0.194 | 6.6079 | 6000 | 0.7101 | 197.0769 | 75.6325 | |
| | 0.1436 | 7.7093 | 7000 | 0.7327 | 235.4833 | 103.3691 | |
| | 0.135 | 8.8106 | 8000 | 0.7635 | 223.1306 | 96.6303 | |
| | 0.0854 | 9.9119 | 9000 | 0.7848 | 235.6624 | 96.6693 | |
| | 0.062 | 11.0132 | 10000 | 0.8102 | 199.8114 | 83.9929 | |
| | 0.0299 | 12.1145 | 11000 | 0.8364 | 177.0486 | 102.8057 | |
| | 0.0254 | 13.2159 | 12000 | 0.8552 | 176.0868 | 85.5468 | |
| | 0.0196 | 14.3172 | 13000 | 0.8671 | 126.2801 | 60.4427 | |
| | 0.0136 | 15.4185 | 14000 | 0.8813 | 177.9727 | 73.2561 | |
| | 0.0102 | 16.5198 | 15000 | 0.8930 | 142.6968 | 57.3544 | |
| | 0.0079 | 17.6211 | 16000 | 0.9064 | 132.6167 | 59.8736 | |
| | 0.0074 | 18.7225 | 17000 | 0.9160 | 125.6011 | 55.9026 | |
| | 0.0053 | 19.8238 | 18000 | 0.9245 | 116.0113 | 50.1628 | |
| | 0.0052 | 20.9251 | 19000 | 0.9299 | 115.0872 | 47.7766 | |
| | 0.0043 | 22.0264 | 20000 | 0.9326 | 115.2570 | 50.1238 | |
| |
| |
| ### Framework versions |
| |
| - Transformers 4.48.3 |
| - Pytorch 2.7.0+cu128 |
| - Datasets 3.6.0 |
| - Tokenizers 0.21.1 |
| |
