legal_t5_small_trans_cs_en model
Model on translating legal text from Cszech 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_cs_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 Cszech to English.
How to use
Here is how to use this model to translate legal text from Cszech to English in PyTorch:
1from transformers import AutoTokenizer, AutoModelWithLMHead, TranslationPipeline
2
3pipeline = TranslationPipeline(
4model=AutoModelWithLMHead.from_pretrained("SEBIS/legal_t5_small_trans_cs_en"),
5tokenizer=AutoTokenizer.from_pretrained(pretrained_model_name_or_path = "SEBIS/legal_t5_small_trans_cs_en", do_lower_case=False,
6 skip_special_tokens=True),
7 device=0
8)
9
10cs_text = "s ohledem na druhou schůzku států OSN, která se konala 11.–15. června 2005 a měla posoudit provádění akčního programu OSN k prevenci, potírání a vymýcení nezákonného obchodu s ručními a lehkými zbraněmi ve všech jeho aspektech, která se koná jednou za dva roky,"
11
12pipeline([cs_text], max_length=512)
Training data
The legal_t5_small_trans_cs_en model was trained on
JRC-ACQUIS,
EUROPARL, and
DCEP dataset consisting of 5 Million parallel texts.
Training procedure
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
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.
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_cs_en | 56.92 |
BibTeX entry and citation info
