malayalam-moe-sft
This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1225
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: 2e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 16
- total_train_batch_size: 64
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 0.03
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 10.0989 | 0.1148 | 200 | 0.6207 |
| 6.3285 | 0.2296 | 400 | 0.3892 |
| 4.7164 | 0.3444 | 600 | 0.2813 |
| 3.3328 | 0.4593 | 800 | 0.2134 |
| 2.9602 | 0.5741 | 1000 | 0.1766 |
| 2.4643 | 0.6889 | 1200 | 0.1565 |
| 2.5094 | 0.8037 | 1400 | 0.1447 |
| 2.2154 | 0.9185 | 1600 | 0.1375 |
| 1.8535 | 1.0333 | 1800 | 0.1340 |
| 1.9010 | 1.1481 | 2000 | 0.1307 |
| 1.5678 | 1.2629 | 2200 | 0.1284 |
| 1.8938 | 1.3777 | 2400 | 0.1264 |
| 1.7750 | 1.4926 | 2600 | 0.1241 |
| 1.6168 | 1.6074 | 2800 | 0.1225 |
Framework versions
- Transformers 5.17.0
- Pytorch 2.10.0+cu128
- Datasets 5.0.1
- Tokenizers 0.23.2
- Downloads last month
- 3