mms-1b-all-bigc_mini-train-drodat-gdro-1.15e-5-dat-0.207-no-sd-0.00-52
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: 34.5988
- Cer: 0.1198
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 |
|---|---|---|---|---|
| 188.1295 | 0.9579 | 500 | 176.9798 | 0.9908 |
| 42.1821 | 1.9157 | 1000 | 42.4846 | 0.1419 |
| 40.2314 | 2.8736 | 1500 | 41.1659 | 0.1376 |
| 39.5713 | 3.8314 | 2000 | 39.4512 | 0.1354 |
| 37.3857 | 4.7893 | 2500 | 38.5420 | 0.1346 |
| 35.9759 | 5.7471 | 3000 | 40.1159 | 0.1333 |
| 33.705 | 6.7050 | 3500 | 37.6714 | 0.1384 |
| 33.0013 | 7.6628 | 4000 | 37.8264 | 0.1325 |
| 32.5986 | 8.6207 | 4500 | 37.2972 | 0.1434 |
| 35.0784 | 9.5785 | 5000 | 36.0234 | 0.1274 |
| 34.4002 | 10.5364 | 5500 | 37.2481 | 0.1413 |
| 31.1324 | 11.4943 | 6000 | 35.7512 | 0.1240 |
| 34.1548 | 12.4521 | 6500 | 35.7485 | 0.1262 |
| 31.2386 | 13.4100 | 7000 | 35.8516 | 0.1235 |
| 29.911 | 14.3678 | 7500 | 35.0069 | 0.1246 |
| 28.7641 | 15.3257 | 8000 | 34.9898 | 0.1268 |
| 27.6891 | 16.2835 | 8500 | 34.7279 | 0.1205 |
| 30.9093 | 17.2414 | 9000 | 34.6778 | 0.1223 |
| 30.4195 | 18.1992 | 9500 | 35.2303 | 0.1268 |
| 29.0716 | 19.1571 | 10000 | 34.5992 | 0.1199 |
| 28.043 | 20.1149 | 10500 | 34.6607 | 0.1216 |
| 26.8507 | 21.0728 | 11000 | 34.9618 | 0.1236 |
| 31.2203 | 22.0307 | 11500 | 34.8334 | 0.1243 |
Framework versions
- Transformers 4.53.0
- Pytorch 2.10.0+cu128
- Datasets 4.5.0
- Tokenizers 0.21.4