mms-1b-all-bigc_mini-drodat-gdro-eta-3.15e-4-dat-0.1454-no-sd-0.00-42
This model is a fine-tuned version of facebook/mms-1b-all on the BIGC_MINI - BEM dataset. It achieves the following results on the evaluation set:
- Loss: 35.3530
- Cer: 0.1220
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 |
|---|---|---|---|---|
| 194.2849 | 0.9579 | 500 | 173.5985 | 0.9431 |
| 42.36 | 1.9157 | 1000 | 42.2178 | 0.1410 |
| 40.014 | 2.8736 | 1500 | 40.7657 | 0.1364 |
| 39.8469 | 3.8314 | 2000 | 39.7961 | 0.1351 |
| 37.188 | 4.7893 | 2500 | 37.7050 | 0.1314 |
| 35.8835 | 5.7471 | 3000 | 37.2826 | 0.1298 |
| 33.578 | 6.7050 | 3500 | 37.3107 | 0.1387 |
| 32.8475 | 7.6628 | 4000 | 38.3177 | 0.1459 |
| 32.2659 | 8.6207 | 4500 | 36.5074 | 0.1349 |
| 35.0408 | 9.5785 | 5000 | 35.9097 | 0.1265 |
| 34.258 | 10.5364 | 5500 | 36.0921 | 0.1286 |
| 31.1104 | 11.4943 | 6000 | 35.3524 | 0.1220 |
| 34.1773 | 12.4521 | 6500 | 36.0286 | 0.1324 |
| 31.0132 | 13.4100 | 7000 | 35.6938 | 0.1306 |
| 29.7286 | 14.3678 | 7500 | 36.1869 | 0.1395 |
Framework versions
- Transformers 4.53.0
- Pytorch 2.10.0+cu128
- Datasets 4.5.0
- Tokenizers 0.21.4