mms-1b-all-bemgen-combined-train-drodat-gdro-1.65e-4-dat-0.019-no-sd-0.00-62
This model is a fine-tuned version of facebook/mms-1b-all on the BEMGEN - BEM dataset. It achieves the following results on the evaluation set:
- Loss: 12.3468
- Cer: 0.0532
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 |
|---|---|---|---|---|
| 269.1012 | 0.6349 | 500 | 266.5935 | 0.9212 |
| 29.2321 | 1.2692 | 1000 | 14.5810 | 0.0647 |
| 30.2389 | 1.9041 | 1500 | 14.0522 | 0.0614 |
| 27.816 | 2.5384 | 2000 | 13.8196 | 0.0609 |
| 24.7595 | 3.1727 | 2500 | 13.4446 | 0.0591 |
| 25.4914 | 3.8076 | 3000 | 13.2415 | 0.0582 |
| 25.0092 | 4.4419 | 3500 | 13.4610 | 0.0576 |
| 25.4794 | 5.0762 | 4000 | 12.8564 | 0.0558 |
| 22.5102 | 5.7111 | 4500 | 12.8421 | 0.0558 |
| 21.6394 | 6.3454 | 5000 | 12.8535 | 0.0563 |
| 22.5194 | 6.9803 | 5500 | 12.7199 | 0.0550 |
| 20.391 | 7.6146 | 6000 | 12.7337 | 0.0548 |
| 21.547 | 8.2489 | 6500 | 12.7232 | 0.0549 |
| 19.5675 | 8.8838 | 7000 | 12.5168 | 0.0542 |
| 19.8442 | 9.5181 | 7500 | 12.3465 | 0.0532 |
| 18.4202 | 10.1524 | 8000 | 12.4335 | 0.0532 |
| 19.5385 | 10.7873 | 8500 | 12.3955 | 0.0529 |
| 18.4112 | 11.4216 | 9000 | 12.6552 | 0.0535 |
Framework versions
- Transformers 4.53.0
- Pytorch 2.10.0+cu128
- Datasets 4.5.0
- Tokenizers 0.21.4