metadata
license: apache-2.0
tags:
- translation
- generated_from_trainer
model-index:
- name: nils-nl-to-rx-pt-v7
results: []
nils-nl-to-rx-pt-v7
This model is a fine-tuned version of Helsinki-NLP/opus-mt-de-en on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0224
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: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.4389 | 1.0 | 500 | 0.0470 |
| 0.0533 | 2.0 | 1000 | 0.0286 |
| 0.0346 | 3.0 | 1500 | 0.0224 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.12.1+cu113
- Datasets 2.6.1
- Tokenizers 0.13.1