metadata
license: apache-2.0
tags:
- generated_from_trainer
metrics:
- bleu
model-index:
- name: MarianMix_en-10
results: []
MarianMix_en-10
This model is a fine-tuned version of Helsinki-NLP/opus-tatoeba-en-ja on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.6857
- Bleu: 27.5143
- Gen Len: 34.3598
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: 32
- eval_batch_size: 32
- seed: 99
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 10
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu | Gen Len |
|---|---|---|---|---|---|
| 1.9771 | 0.44 | 500 | 1.5875 | 1.2678 | 52.398 |
| 0.9175 | 0.89 | 1000 | 1.2906 | 6.1512 | 33.5839 |
| 0.7443 | 1.33 | 1500 | 1.0735 | 11.8884 | 33.0704 |
| 0.6426 | 1.78 | 2000 | 0.9260 | 16.6793 | 36.4523 |
| 0.5637 | 2.22 | 2500 | 0.8367 | 20.7927 | 35.9216 |
| 0.5013 | 2.67 | 3000 | 0.7833 | 22.5786 | 34.401 |
| 0.465 | 3.11 | 3500 | 0.7479 | 24.2307 | 36.3799 |
| 0.4187 | 3.56 | 4000 | 0.7232 | 25.5875 | 33.5236 |
| 0.4108 | 4.0 | 4500 | 0.6995 | 26.672 | 33.9236 |
| 0.3703 | 4.45 | 5000 | 0.6924 | 27.0479 | 34.9759 |
| 0.3701 | 4.89 | 5500 | 0.6857 | 27.5143 | 34.3598 |
Framework versions
- Transformers 4.12.5
- Pytorch 1.9.1
- Datasets 1.17.0
- Tokenizers 0.10.3