language: French English
tags:
- translation French English model
datasets:
- dcep europarl jrc-acquis
widget:
- text: >-
(16) Pour assurer un niveau élevé de protection des enfants et de
l'environnement contre les risques , il convient d'accorder une attention
particulière aux substances dangereuses, en particulier les substances et
éléments CMR et allergènes.
legal_t5_small_trans_fr_en model
Model on translating legal text from French to English. It was first released in this repository. This model is trained on three parallel corpus from jrc-acquis, europarl and dcep.
Model description
legal_t5_small_trans_fr_en is based on the t5-small model and was trained on a large corpus of parallel text. This is a smaller model, which scales the baseline model of t5 down by using dmodel = 512, dff = 2,048, 8-headed attention, and only 6 layers each in the encoder and decoder. This variant has about 60 million parameters.
Intended uses & limitations
The model could be used for translation of legal texts from French to English.
How to use
Here is how to use this model to translate legal text from French to English in PyTorch:
from transformers import AutoTokenizer, AutoModelWithLMHead, TranslationPipeline
pipeline = TranslationPipeline(
model=AutoModelWithLMHead.from_pretrained("SEBIS/legal_t5_small_trans_fr_en"),
tokenizer=AutoTokenizer.from_pretrained(pretrained_model_name_or_path = "SEBIS/legal_t5_small_trans_fr_en", do_lower_case=False,
skip_special_tokens=True),
device=0
)
fr_text = "(16) Pour assurer un niveau élevé de protection des enfants et de l'environnement contre les risques , il convient d'accorder une attention particulière aux substances dangereuses, en particulier les substances et éléments CMR et allergènes."
pipeline([fr_text], max_length=512)
Training data
The legal_t5_small_trans_fr_en model was trained on JRC-ACQUIS, EUROPARL, and DCEP dataset consisting of 5 Million parallel texts.
Training procedure
An unigram model trained with 88M lines of text from the parallel corpus (of all possible language pairs) to get the vocabulary (with byte pair encoding), which is used with this model.
The model was trained on a single TPU Pod V3-8 for 250K steps in total, using sequence length 512 (batch size 4096). It has a total of approximately 220M parameters and was trained using the encoder-decoder architecture. The optimizer used is AdaFactor with inverse square root learning rate schedule for pre-training.
Preprocessing
Pretraining
Evaluation results
When the model is used for translation test dataset, achieves the following results:
Test results :
| Model | BLEU score |
|---|---|
| legal_t5_small_trans_fr_en | 51.44 |