mms-1b-all-nyagen-combined-train-drodat-gdro-2.5e-3-dat-0.0012-no-sd-0.00-42
This model is a fine-tuned version of facebook/mms-1b-all on the NYAGEN - NYA dataset. It achieves the following results on the evaluation set:
- Loss: 36.0140
- Cer: 0.0822
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: 0.0003
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- 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
- num_epochs: 30.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 428.3506 | 1.2566 | 500 | 419.0719 | 0.8577 |
| 68.987 | 2.5132 | 1000 | 53.9598 | 0.1133 |
| 25.5585 | 3.7698 | 1500 | 45.2925 | 0.0948 |
| 21.578 | 5.0252 | 2000 | 40.8636 | 0.0898 |
| 20.9484 | 6.2818 | 2500 | 39.4182 | 0.0864 |
| 17.3114 | 7.5384 | 3000 | 40.2077 | 0.0842 |
| 19.0107 | 8.7950 | 3500 | 38.0877 | 0.0837 |
| 16.4556 | 10.0503 | 4000 | 38.3639 | 0.0823 |
| 18.1135 | 11.3069 | 4500 | 36.0151 | 0.0822 |
| 18.0688 | 12.5635 | 5000 | 39.2569 | 0.0809 |
| 13.1975 | 13.8201 | 5500 | 38.7911 | 0.0808 |
| 16.4294 | 15.0755 | 6000 | 39.9338 | 0.0794 |
Framework versions
- Transformers 4.53.0
- Pytorch 2.10.0+cu128
- Datasets 4.5.0
- Tokenizers 0.21.4