mbart-large-50-English_German_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: 1.2342
- Bleu: 35.5931
- Rouge1: 0.5803386608353808
- Rouge2: 0.3939141514072567
- RougeL: 0.5438629663406402
- RougeLsum: 0.544153348468965
- Meteor: 0.5500546034636025
Model description
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:DE.ipynb
Intended uses & limitations
This model is intended to demonstrate my ability to solve a complex problem using technology.
Training and evaluation data
Here is a the link to the page where I found this dataset: 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: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu | Rouge1 | Rouge2 | RougeL | RougeLsum | Meteor |
|---|---|---|---|---|---|---|---|---|---|
| 1.7738 | 1.0 | 900 | 1.2342 | 35.7436 | 0.5806 | 0.3941 | 0.5442 | 0.5444 | 0.5512 |
- All values in the chart above are rounded to near ten-thousandth.
Framework versions
- Transformers 4.22.2
- Pytorch 1.12.1
- Datasets 2.5.2
- Tokenizers 0.12.1
