mms-1b-all-nyagen-combined-train-drodat-gdro-1.24e-5-dat-0.0272-sd-0.00016-62
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: 37.6380
- Cer: 0.0881
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 |
|---|---|---|---|---|
| 394.5577 | 1.2566 | 500 | 391.5746 | 0.8785 |
| 53.7799 | 2.5132 | 1000 | 54.3380 | 0.1125 |
| 24.6677 | 3.7698 | 1500 | 45.1341 | 0.0950 |
| 20.7673 | 5.0252 | 2000 | 40.8595 | 0.0891 |
| 19.429 | 6.2818 | 2500 | 37.6373 | 0.0881 |
| 16.5107 | 7.5384 | 3000 | 39.7112 | 0.0851 |
| 18.6249 | 8.7950 | 3500 | 37.7436 | 0.0836 |
| 15.7618 | 10.0503 | 4000 | 38.7291 | 0.0833 |
Framework versions
- Transformers 4.53.0
- Pytorch 2.10.0+cu128
- Datasets 4.5.0
- Tokenizers 0.21.4