|
--- |
|
license: apache-2.0 |
|
datasets: |
|
- giga_fren |
|
language: |
|
- fr |
|
- en |
|
--- |
|
|
|
|
|
# Model Card for fr_en-t5-large |
|
|
|
<!-- Provide a quick summary of what the model is/does. --> |
|
|
|
This model has been optimized for French and English language processing while minimizing overall size. To achieve this, I only retained relevant parameters and tokens specific to these two languages, ensuring that performance remains as good as the original mt5. |
|
|
|
## Model Details |
|
I used a method outlined in a [blog post](https://towardsdatascience.com/how-to-adapt-a-multilingual-t5-model-for-a-single-language-b9f94f3d9c90) by David Dale to downsize the multilingual T5 model for French and English use cases specifically. By utilizing the giga_fren dataset, I was able to successfully reduce the total number of tokens and decrease both the model and tokenizer sizes by 38% and 80% respectively. |
|
|
|
### Model Description |
|
|
|
- **Developed by:** Korventenn |
|
- **Model type:** mt5 |
|
- **Language(s) (NLP):** French and English |
|
- **License:** Apache 2.0 |
|
- **Generated from model:** mt5-large |
|
|
|
### Model Sources [optional] |
|
|
|
<!-- Provide the basic links for the model. --> |
|
|
|
- **Repository:** https://colab.research.google.com/drive/1cDWtO5BqWMm_nxnM7lHmPEKMWMejHdBJ#scrollTo=s6ebzRxA1VGv |
|
|
|
## Uses |
|
|
|
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. --> |
|
You can use the raw model for any sequence to sequence task that is focused on either french, english or both. |
|
|
|
|
|
## How to Get Started with the Model |
|
|
|
Use the code below to get started with the model. |
|
``` |
|
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM |
|
|
|
tokenizer = AutoTokenizer.from_pretrained("Korventenn/fr_en-t5-large") |
|
|
|
model = AutoModelForSeq2SeqLM.from_pretrained("Korventenn/fr_en-t5-large") |
|
``` |
|
|
|
### Training Data |
|
|
|
<!-- This should link to a Data Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. --> |
|
|
|
[giga_fren](https://huggingface.co/datasets/giga_fren) |
