mms-1b-all-nyagen-combined-train-drodat-gdro-1.24e-5-dat-0.0272-sd-0.00016-52
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: 35.5455
- Cer: 0.0833
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 |
|---|---|---|---|---|
| 395.0277 | 1.2566 | 500 | 391.6377 | 0.8616 |
| 28.4733 | 2.5132 | 1000 | 49.1616 | 0.1026 |
| 23.7521 | 3.7698 | 1500 | 45.0713 | 0.0944 |
| 20.2069 | 5.0252 | 2000 | 41.4736 | 0.0896 |
| 18.9334 | 6.2818 | 2500 | 38.8340 | 0.0863 |
| 16.3297 | 7.5384 | 3000 | 38.8923 | 0.0846 |
| 18.1028 | 8.7950 | 3500 | 36.9617 | 0.0831 |
| 15.5528 | 10.0503 | 4000 | 38.3747 | 0.0831 |
| 17.0504 | 11.3069 | 4500 | 35.5464 | 0.0833 |
| 15.9957 | 12.5635 | 5000 | 39.0163 | 0.0817 |
| 12.825 | 13.8201 | 5500 | 38.0489 | 0.0825 |
| 15.3868 | 15.0755 | 6000 | 37.4834 | 0.0810 |
Framework versions
- Transformers 4.53.0
- Pytorch 2.10.0+cu128
- Datasets 4.5.0
- Tokenizers 0.21.4