mt5-xl-mugec-lora-10best
This model is a fine-tuned version of google/mt5-xl on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0270
- Wer: 0.2414
- Cer: 0.1145
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: 0.0001
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- total_eval_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_ratio: 0.05
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.0373 | 1.0 | 13087 | 0.0304 | 0.2666 | 0.127 |
| 0.0351 | 2.0 | 26174 | 0.0284 | 0.2526 | 0.12 |
| 0.0379 | 3.0 | 39261 | 0.0275 | 0.246 | 0.1169 |
| 0.0269 | 4.0 | 52348 | 0.0271 | 0.2425 | 0.1148 |
| 0.0315 | 4.9997 | 65430 | 0.0270 | 0.2414 | 0.1145 |
Framework versions
- PEFT 0.12.0
- Transformers 4.48.3
- Pytorch 2.7.1+cu126
- Datasets 2.19.1
- Tokenizers 0.21.1