<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
mbart-large-50-many-to-many-mmt-finetuned-ar-to-en
This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.5204
- Bleu: 34.8213
- Gen Len: 33.544
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: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- training_steps: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu | Gen Len |
|---|---|---|---|---|---|
| 1.4657 | 1.0 | 816 | 1.1739 | 30.1212 | 32.868 |
| 0.8541 | 2.0 | 1632 | 1.1190 | 33.0098 | 32.808 |
| 0.6176 | 3.0 | 2448 | 1.1681 | 33.3634 | 32.756 |
| 0.3397 | 4.0 | 3264 | 1.3327 | 33.2941 | 33.6 |
| 0.2227 | 5.0 | 4080 | 1.4211 | 33.9298 | 33.128 |
| 0.1597 | 6.0 | 4896 | 1.5157 | 34.7405 | 33.496 |
| 0.1426 | 6.13 | 5000 | 1.5204 | 34.8213 | 33.544 |
Framework versions
- Transformers 4.21.1
- Pytorch 1.12.1+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1
