jem-whisper-yi-2026-09-03-v3
This model is a fine-tuned version of ivrit-ai/yi-whisper-large-v3 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.3852
- Wer Ortho: 0.3108
- Wer: 0.2399
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: 1e-05
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- 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: constant_with_warmup
- lr_scheduler_warmup_ratio: 0.1
- lr_scheduler_warmup_steps: 100
- num_epochs: 4
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.4733 | 0.2866 | 250 | 0.4428 | 0.4496 | 0.4048 |
| 0.3791 | 0.5732 | 500 | 0.3999 | 0.3502 | 0.2960 |
| 0.3077 | 0.8598 | 750 | 0.3765 | 0.3514 | 0.2856 |
| 0.2957 | 1.1456 | 1000 | 0.3720 | 0.3739 | 0.3119 |
| 0.2956 | 1.4322 | 1250 | 0.3660 | 0.3348 | 0.2792 |
| 0.2979 | 1.7188 | 1500 | 0.3602 | 0.3275 | 0.2709 |
| 0.212 | 2.0046 | 1750 | 0.3604 | 0.3620 | 0.2787 |
| 0.201 | 2.2912 | 2000 | 0.3682 | 0.3743 | 0.2825 |
| 0.2502 | 2.5778 | 2250 | 0.3623 | 0.3231 | 0.2583 |
| 0.2068 | 2.8644 | 2500 | 0.3596 | 0.3178 | 0.2493 |
| 0.1734 | 3.1502 | 2750 | 0.3869 | 0.3219 | 0.2601 |
| 0.1432 | 3.4368 | 3000 | 0.3873 | 0.3480 | 0.2977 |
| 0.185 | 3.7234 | 3250 | 0.3852 | 0.3108 | 0.2399 |
Framework versions
- Transformers 4.48.1
- Pytorch 2.4.0
- Datasets 3.6.0
- Tokenizers 0.21.4