metadata
license: mit
tags:
- translation
- generated_from_trainer
metrics:
- bleu
- rouge
model-index:
- name: mbart-large-50-English_Spanish_Translation
results: []
language:
- en
- es
mbart-large-50-English_Spanish_Translation
This model is a fine-tuned version of facebook/mbart-large-50 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.0290
- Bleu: 41.4437
- Rouge: {'rouge1': 0.6751402780531002, 'rouge2': 0.49769602014143044, 'rougeL': 0.6371513427059108, 'rougeLsum': 0.6376403149816605}
- Meteor: {'meteor': 0.6479226630466496}
Model description
Translating English inputs to Spanish.
Here is the link to the script I created to train this model: https://github.com/DunnBC22/NLP_Projects/blob/main/Machine%20Translation/NLP%20Translation%20Project-EN:ES.ipynb
Intended uses & limitations
This model is intended to demonstrate my ability to solve a complex problem using technology.
Training and evaluation data
Dataset Source: https://www.kaggle.com/datasets/hgultekin/paralel-translation-corpus-in-22-languages
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu | Rouge | Meteor |
|---|---|---|---|---|---|---|
| 1.5608 | 1.0 | 900 | 1.0899 | 39.9184 | {'rouge1': 0.6645461901016299, 'rouge2': 0.48457734138815345, 'rougeL': 0.6254335531454508, 'rougeLsum': 0.6258737583448748} | {'meteor': 0.6376166612731494} |
| 0.9734 | 2.0 | 1800 | 1.0290 | 41.4436 | {'rouge1': 0.6751348620702116, 'rouge2': 0.4976855704059807, 'rougeL': 0.6371345376462452, 'rougeLsum': 0.6376186633843448} | {'meteor': 0.6479188510808377} |
Framework versions
- Transformers 4.22.2
- Pytorch 1.12.1
- Datasets 2.5.2
- Tokenizers 0.12.1