bert-base-uncased for QA

Overview

Language model: bert-base-uncased Language: English Downstream-task: Extractive QA Training data: SQuAD 2.0 Eval data: SQuAD 2.0 Infrastructure : 1x Tesla v100

Hyperparameters

batch_size = 32
n_epochs = 3
base_LM_model = "bert-base-uncased"
max_seq_len = 384
learning_rate = 3e-5
lr_schedule = LinearWarmup
warmup_proportion = 0.2
doc_stride=128
max_query_length=64

Performance

"exact": 73.67977764676156
"f1": 77.87647139308865

Authors

• Timo Möller: timo.moeller [at] deepset.ai
• Julian Risch: julian.risch [at] deepset.ai
• Malte Pietsch: malte.pietsch [at] deepset.ai
• Michel Bartels: michel.bartels [at] deepset.ai

About us

We bring NLP to the industry via open source! Our focus: Industry specific language models & large scale QA systems.

Some of our work:

• German BERT (aka "bert-base-german-cased")
• GermanQuAD and GermanDPR datasets and models (aka "gelectra-base-germanquad", "gbert-base-germandpr")
• FARM
• Haystack

Get in touch: Twitter | LinkedIn | Discord | GitHub Discussions | Website

By the way: we're hiring!