osiria
/

deberta-base-italian

Feature Extraction

text-embeddings-inference

Model card Files Files and versions

Model: DeBERTa
Lang: IT

Model description

This is a DeBERTa [1] model for the Italian language, obtained using mDeBERTa (mdeberta-v3-base) as a starting point and focusing it on the Italian language by modifying the embedding layer (as in [2], computing document-level frequencies over the Wikipedia dataset)

The resulting model has 124M parameters, a vocabulary of 50.256 tokens, and a size of ~500 MB.

Quick usage

from transformers import DebertaV2TokenizerFast, DebertaV2Model

tokenizer = DebertaV2TokenizerFast.from_pretrained("osiria/deberta-base-italian")
model = DebertaV2Model.from_pretrained("osiria/deberta-base-italian")

References

[1] https://arxiv.org/abs/2111.09543

[2] https://arxiv.org/abs/2010.05609

License

The model is released under MIT license

Downloads last month: 13

Safetensors

Model size

0.1B params

Tensor type

I64

·

F32

·

Collection including osiria/deberta-base-italian

Osiria "Earth" Series 🌱

This collection is composed of robust and reliable models for common NLP tasks • 10 items • Updated Dec 9, 2023 • 1

Papers for osiria/deberta-base-italian

DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing

Paper • 2111.09543 • Published Nov 18, 2021 • 3

Load What You Need: Smaller Versions of Multilingual BERT

Paper • 2010.05609 • Published Oct 12, 2020 • 1