dlicari's picture
Update README.md
01ea850
|
raw
history blame
1.08 kB
---
language: it
license: afl-3.0
widget:
- text: Il <mask> ha chiesto revocarsi l'obbligo di pagamento
---
<img src="https://huggingface.co/dlicari/Italian-Legal-BERT-SC/resolve/main/ITALIAN_LEGAL_BERT-SC.jpg" width="600"/>
# ITALIAN-LEGAL-BERT-SC
It is the [ITALIAN-LEGAL-BERT](https://huggingface.co/dlicari/Italian-Legal-BERT) variant pre-trained from scratch on Italian legal documents (ITA-LEGAL-BERT-SC) based on the CamemBERT architecture
## Training procedure
It was trained from scratch using a larger training dataset, 6.6GB of civil and criminal cases.
We used [CamemBERT](https://huggingface.co/docs/transformers/main/en/model_doc/camembert) architecture with a language modeling head on top, AdamW Optimizer, initial learning rate 2e-5 (with linear learning rate decay), sequence length 512, batch size 18, 1 million training steps,
device 8*NVIDIA A100 40GB using distributed data parallel (each step performs 8 batches). It uses SentencePiece tokenization trained from scratch on a subset of training set (5 milions sentences)
and vocabulary size of 32000