Downloads · 30 days
30
2% of all-time downloads
mT5-base_ar
ArabicNLP/mT5-base_ar
mT5-base_ar is a machine learning model from ArabicNLP. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as mit.
An Arabic LLM derived from Google's mT5 multi-lingual model
All-time downloads
1.5K
Public
Repo size
2 GB
Likes
6
Trending 1
Repo makeup
Click a slice to open those files.
.bin977 MB · 100%
At a glance
- Library
- transformers
- License
- mit
- Model type
- t5
- Access
- Public
- Created
- May 14, 2023
- Updated
- May 14, 2023
- SHA
- a4664f70
Model card
From the Hugging Face model README
Model Card
An Arabic LLM derived from Google's mT5 multi-lingual model
Model Details
Model Description
This is a smaller version of the google/mt5-base model with only Arabic and some English embeddings left.
The original model has 582M parameters, with 384M of them being input and output embeddings. After shrinking the sentencepiece vocabulary from 250K to 30K (top 10K English and top 20K Arabic tokens) the number of model parameters reduced to 244M parameters, and model size reduced from 2.2GB to 0.9GB - 42% of the original one.
The creation of this model was inspired from David Dales'article "<a href="https://towardsdatascience.com/how-to-adapt-a-multilingual-t5-model-for-a-single-language-b9f94f3d9c90">How to adapt a multilingual T5 model for a single language</a>" in which mT5 was compressed to support Russian and English languages along with the source code.
- Developed by: Moustafa Banbouk
- Model type: Unsupervised LLM
- Language(s) (NLP): Arabic, English
- License: MIT
Downstream Uses
Quesion Answering, Summarization, Classification ...
- Library
- transformers
- Type
- t5
- License
- mit
- Languages
- ar, en
- Created
- May 14, 2023
- Updated
- May 14, 2023