metadata
license: apache-2.0
datasets:
- panosgriz/COVID-QA-el-small
language:
- el
metrics:
- f1
- exact_match
base_model:
- timpal0l/mdeberta-v3-base-squad2
pipeline_tag: question-answering
tags:
- closed-domain
- extractive-qa
Description
This model is an adapted version of an mDeBERTa model, fine-tuned on the SQuAD v2 dataset for the COVID-19 domain and optimized for the Greek language.
Training Details
- Training Dataset:
COVID-QA-el-small - Batch Size: 8
- Number of Epochs: 3
- Learning Rate: 3e-05
- Gradient Accumulation Steps: 2