Edit model card


This is a model, based on BERT trained on cased Italian, that can be used for Extractive Q&A on Italian texts.

Model description

This model has been trained on squad_it dataset starting from the pre-trained model dbmdz/bert-base-italian-xxl-cased.

These are the metrics computed on evaluation set:

  • EM: 63.95
  • F1: 75.27

How to use

from transformers import pipeline

pipe_qa = pipeline('question-answering', model='luigisaetta/squad_it_xxl_cased_hub1')

pipe_qa(context="Io sono nato a Napoli. Il mare bagna Napoli. Napoli è la più bella città del mondo", 
        question="Qual è la più bella città del mondo?")

Intended uses & limitations

This model can be used for Extractive Q&A on Italian Text

Training and evaluation data


Training procedure

see code in this NoteBook

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 1234
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 2
  • mixed_precision_training: Native AMP

Framework versions

  • Transformers 4.20.0.dev0
  • Pytorch 1.9.0
  • Datasets 1.11.0
  • Tokenizers 0.12.1
Downloads last month
Hosted inference API
Question Answering
This model can be loaded on the Inference API on-demand.

Dataset used to train luigisaetta/squad_it_xxl_cased_hub1