VN_ja-en_helsinki
This model is a fine-tuned version of Helsinki-NLP/opus-mt-ja-en on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 2.2409
- BLEU: 15.28
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: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
2.6165 | 0.19 | 2000 | 2.6734 |
2.3805 | 0.39 | 4000 | 2.6047 |
2.2793 | 0.58 | 6000 | 2.5461 |
2.2028 | 0.78 | 8000 | 2.5127 |
2.1361 | 0.97 | 10000 | 2.4511 |
1.9653 | 1.17 | 12000 | 2.4331 |
1.934 | 1.36 | 14000 | 2.3840 |
1.9002 | 1.56 | 16000 | 2.3901 |
1.87 | 1.75 | 18000 | 2.3508 |
1.8408 | 1.95 | 20000 | 2.3082 |
1.6937 | 2.14 | 22000 | 2.3279 |
1.6371 | 2.34 | 24000 | 2.3052 |
1.6264 | 2.53 | 26000 | 2.3071 |
1.6029 | 2.72 | 28000 | 2.2685 |
1.5847 | 2.92 | 30000 | 2.2409 |
Framework versions
- Transformers 4.19.2
- Pytorch 1.11.0+cu113
- Datasets 2.2.2
- Tokenizers 0.12.1
- Downloads last month
- 5