metadata
language:
- ko
- ja
base_model: facebook/mbart-large-50-many-to-many-mmt
tags:
- generated_from_trainer
metrics:
- bleu
model-index:
- name: mbartLarge_koja_37p
results: []
mbartLarge_koja_37p
This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.8995
- Bleu: 6.2934
- Gen Len: 17.4862
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: 5e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- total_eval_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1500
- num_epochs: 35
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu | Gen Len |
|---|---|---|---|---|---|
| 1.2712 | 0.48 | 5500 | 1.2204 | 3.0858 | 18.1085 |
| 1.0946 | 0.97 | 11000 | 1.0578 | 3.3162 | 17.8651 |
| 0.9546 | 1.45 | 16500 | 0.9688 | 5.6024 | 17.902 |
| 0.89 | 1.94 | 22000 | 0.9414 | 5.1453 | 17.6144 |
| 0.834 | 2.42 | 27500 | 0.9213 | 5.3985 | 17.6899 |
| 0.7439 | 2.91 | 33000 | 0.8995 | 6.2934 | 17.4862 |
| 0.6803 | 3.39 | 38500 | 0.9016 | 6.3565 | 17.8899 |
| 0.733 | 3.88 | 44000 | 0.9226 | 7.0351 | 17.6112 |
| 0.6601 | 4.36 | 49500 | 0.9807 | 5.3084 | 17.4292 |
| 0.6933 | 4.84 | 55000 | 0.9238 | 6.8389 | 17.5131 |
Framework versions
- Transformers 4.34.1
- Pytorch 2.1.0+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1