README.md

---
license: apache-2.0
datasets:
- projecte-aina/CA-FR_Parallel_Corpus
language:
- ca
- fr
metrics:
- bleu
library_name: fairseq
---
## Projecte Aina’s Catalan-French machine translation model
 
## Model description

This model was trained from scratch using the [Fairseq toolkit](https://fairseq.readthedocs.io/en/latest/) on a combination of Catalan-French datasets, 
which after filtering and cleaning comprised 18.634.844 sentence pairs. The model is evaluated on the Flores and NTREX evaluation sets.  

## Intended uses and limitations

You can use this model for machine translation from Catalan to French.

## How to use

### Usage
Required libraries:

```bash
pip install ctranslate2 pyonmttok
```

Translate a sentence using python
```python
import ctranslate2
import pyonmttok
from huggingface_hub import snapshot_download
model_dir = snapshot_download(repo_id="projecte-aina/aina-translator-ca-fr", revision="main")

tokenizer=pyonmttok.Tokenizer(mode="none", sp_model_path = model_dir + "/spm.model")
tokenized=tokenizer.tokenize("Benvingut al projecte Aina!!")

translator = ctranslate2.Translator(model_dir)
translated = translator.translate_batch([tokenized[0]])
print(tokenizer.detokenize(translated[0][0]['tokens']))
```

## Limitations and bias
At the time of submission, no measures have been taken to estimate the bias and toxicity embedded in the model. 
However, we are well aware that our models may be biased. We intend to conduct research in these areas in the future, and if completed, this model card will be updated. 

## Training

### Training data

The model was trained on a combination of the following datasets:

| Dataset       	| Sentences  	| Sentences after Cleaning        	|
|-------------------|----------------|-------------------|
| CCMatrix      	|24.386.198  	| 16.305.758   	|
| Multi CCAligned | 1.954.475 | 1.442.584 |
| WikiMatrix | 490.871 | 437.665 |
| GNOME | 12.962 | 1.686 |
|KDE 4 | 163.143 | 111.750 |
| Open Subtitles | 392.159 | 225.786 |

### Training procedure

### Data preparation

All datasets are deduplicated and filtered to remove any sentence pairs with a cosine similarity of less than 0.75. 
This is done using sentence embeddings calculated using LaBSE. The filtered datasets are then concatenated to form the final corpus 
and before training the punctuation is normalized using a modified version of the join-single-file.py script from SoftCatalà


#### Tokenization

All data is tokenized using sentencepiece, with 50 thousand token sentencepiece model learned from the combination of all filtered training data. 
This model is included.  

#### Hyperparameters

The model is based on the Transformer-XLarge proposed by [Subramanian et al.](https://aclanthology.org/2021.wmt-1.18.pdf)
The following hyperparamenters were set on the Fairseq toolkit:

| Hyperparameter                 	| Value                        	|
|------------------------------------|----------------------------------|
| Architecture                   	| transformer_vaswani_wmt_en_de_bi |
| Embedding size                 	| 1024                         	|
| Feedforward size               	| 4096                         	|
| Number of heads                	| 16                           	|
| Encoder layers                 	| 24                           	|
| Decoder layers                 	| 6                            	|
| Normalize before attention     	| True                         	|
| --share-decoder-input-output-embed | True                         	|
| --share-all-embeddings         	| True                         	|
| Effective batch size           	| 96.000                       	|
| Optimizer                      	| adam                         	|
| Adam betas                     	| (0.9, 0.980)                 	|
| Clip norm                      	| 0.0                          	|
| Learning rate                  	| 1e-3                         	|
| Lr. schedurer                  	| inverse sqrt                 	|
| Warmup updates                 	| 4000                         	|
| Dropout                        	| 0.1                          	|
| Label smoothing                	| 0.1                          	|

The model was trained using shards of 10 million sentences, for a total of 11.000 updates. 
Weights were saved every 1000 updates and reported results are the average of the last 4 checkpoints.

## Evaluation

### Variable and metrics

We use the BLEU score for evaluation on test sets: [Flores-101](https://github.com/facebookresearch/flores), [NTREX](https://github.com/MicrosoftTranslator/NTREX) 

### Evaluation results

Below are the evaluation results on the machine translation from Catalan to French compared to [Softcatalà](https://www.softcatala.org/) 
and [Google Translate](https://translate.google.es/?hl=es):

| Test set         	| SoftCatalà | Google Translate | aina-translator-ca-fr |
|----------------------|------------|------------------|---------------|
| Flores 101 dev   	| 34,6    	| **43,4**     	| 38,2      	|
| Flores 101 devtest   | 35,3   	| **43,4**     	| 38,2      	|
| NTREX    	| 25,3   	| **31,5**     	| 27,7      	|
| Average          	| 31,7  	| **39,4**     	| 34,7      	|

## Additional information

### Author
The Language Technologies Unit from Barcelona Supercomputing Center.

### Contact
For further information, please send an email to <langtech@bsc.es>.

### Copyright
Copyright(c) 2023 by Language Technologies Unit, Barcelona Supercomputing Center.

### License
[Apache License, Version 2.0](https://www.apache.org/licenses/LICENSE-2.0)

### Funding
This work has been promoted and financed by the Generalitat de Catalunya through the [Aina project](https://projecteaina.cat/).

### Disclaimer

<details>
<summary>Click to expand</summary>

The model published in this repository is intended for a generalist purpose and is available to third parties under a permissive Apache License, Version 2.0. 

Be aware that the model may have biases and/or any other undesirable distortions.

When third parties deploy or provide systems and/or services to other parties using this model (or any system based on it) 
or become users of the model, they should note that it is their responsibility to mitigate the risks arising from its use and, 
in any event, to comply with applicable regulations, including regulations regarding the use of Artificial Intelligence.

In no event shall the owner and creator of the model (Barcelona Supercomputing Center) 
be liable for any results arising from the use made by third parties.

</details>