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: 0.0002
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 14
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
(table generated directly from trainer.state.log_history, keyed by metric name -- avoids a transformers model-card column-ordering bug seen in fb_mms_1b.py runs)
Table with columns: Training Loss, Epoch, Step, Ccer, Cer, Validation Loss, Ncer, Wer| Training Loss | Epoch | Step | Ccer | Cer | Validation Loss | Ncer | Wer |
|---|
| 0.5394 | 1.0 | 299 | 0.2622 | 0.2439 | 0.2230 | 0.2115 | 0.5136 |
| 0.0748 | 2.0 | 598 | 0.6463 | 0.6250 |
Framework versions
- Transformers 5.15.1
- Pytorch 2.13.0+cu130
- Datasets 5.0.1
- Tokenizers 0.22.2
