language: French Italian
tags:
- translation French Italian model
datasets:
- dcep europarl jrc-acquis
widget:
- text: >-
invite le Conseil et la Commission à faire en sorte que les informations
en ce qui concerne les processus et les exigences liés à la production des
rapports annuels, les normes de base applicables aux codes de conduite et
les rapports annuels et codes de conduite modèles soient publiés,
notamment sur un site Web du Forum de l'UE pour la RSE;
legal_t5_small_trans_fr_it model
Model on translating legal text from French to Italian. 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_it 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 Italian.
How to use
Here is how to use this model to translate legal text from French to Italian in PyTorch:
from transformers import AutoTokenizer, AutoModelWithLMHead, TranslationPipeline
pipeline = TranslationPipeline(
model=AutoModelWithLMHead.from_pretrained("SEBIS/legal_t5_small_trans_fr_it"),
tokenizer=AutoTokenizer.from_pretrained(pretrained_model_name_or_path = "SEBIS/legal_t5_small_trans_fr_it", do_lower_case=False,
skip_special_tokens=True),
device=0
)
fr_text = "invite le Conseil et la Commission à faire en sorte que les informations en ce qui concerne les processus et les exigences liés à la production des rapports annuels, les normes de base applicables aux codes de conduite et les rapports annuels et codes de conduite modèles soient publiés, notamment sur un site Web du Forum de l'UE pour la RSE;"
pipeline([fr_text], max_length=512)
Training data
The legal_t5_small_trans_fr_it 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_it | 46.45 |