mms-1b-all-bigc_mini-drodat-gdro-eta-3.15e-4-dat-0.1321-sd-4.21e-5-42
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: 34.6637
- Cer: 0.1229
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 |
|---|---|---|---|---|
| 190.5987 | 0.9579 | 500 | 179.2786 | 0.9964 |
| 42.372 | 1.9157 | 1000 | 42.5097 | 0.1420 |
| 40.1895 | 2.8736 | 1500 | 40.9952 | 0.1376 |
| 39.5348 | 3.8314 | 2000 | 40.2046 | 0.1378 |
| 37.3002 | 4.7893 | 2500 | 38.8989 | 0.1372 |
| 36.1214 | 5.7471 | 3000 | 37.9122 | 0.1304 |
| 33.7925 | 6.7050 | 3500 | 39.0563 | 0.1521 |
| 32.851 | 7.6628 | 4000 | 38.6533 | 0.1393 |
| 32.2445 | 8.6207 | 4500 | 37.7899 | 0.1465 |
| 34.7986 | 9.5785 | 5000 | 37.0572 | 0.1349 |
| 34.645 | 10.5364 | 5500 | 36.6212 | 0.1312 |
| 31.3772 | 11.4943 | 6000 | 35.9364 | 0.1236 |
| 34.8718 | 12.4521 | 6500 | 36.6453 | 0.1344 |
| 31.5628 | 13.4100 | 7000 | 35.9305 | 0.1278 |
| 30.3614 | 14.3678 | 7500 | 35.5707 | 0.1300 |
| 29.465 | 15.3257 | 8000 | 35.0367 | 0.1240 |
| 28.1474 | 16.2835 | 8500 | 34.8245 | 0.1211 |
| 31.0972 | 17.2414 | 9000 | 34.7756 | 0.1217 |
| 30.9154 | 18.1992 | 9500 | 34.7894 | 0.1216 |
| 29.1404 | 19.1571 | 10000 | 34.9429 | 0.1199 |
| 28.4509 | 20.1149 | 10500 | 34.7676 | 0.1225 |
| 27.3747 | 21.0728 | 11000 | 35.3914 | 0.1277 |
| 31.5625 | 22.0307 | 11500 | 35.1986 | 0.1272 |
| 29.368 | 22.9885 | 12000 | 34.6641 | 0.1229 |
| 28.6572 | 23.9464 | 12500 | 34.9567 | 0.1236 |
| 29.408 | 24.9042 | 13000 | 34.9854 | 0.1229 |
| 26.2958 | 25.8621 | 13500 | 35.1067 | 0.1214 |
Framework versions
- Transformers 4.53.0
- Pytorch 2.10.0+cu128
- Datasets 4.5.0
- Tokenizers 0.21.4