metadata

language:
  - de
tags:
  - qa
widget:
  - text: ''
    context: ''
    example_title: Extractive QA

GELECTRA-large-LegalQuAD

Overview

Language model: GELECTRA-large Language: German Downstream-task: Extractive QA
Training data: German-legal-SQuAD
Eval data: German-legal-SQuAD testset

Hyperparameters

batch_size = 10
n_epochs = 2
max_seq_len=256,
learning_rate=1e-5,

## Eval results
Evaluated on German-legal-SQuAD testset
  "exact": 40.912
  "f1": 72.368