Skip to content
This repository has been archived by the owner on Apr 3, 2020. It is now read-only.
/ bert Public archive

A barebones (Distil)BERT pipeline for token classification tasks driven by catalyst

Notifications You must be signed in to change notification settings

catalyst-team/bert

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Build Status Telegram Gitter Slack Donate

Catalyst.Bert

A barebones (Distil)BERT pipeline for token classification tasks driven by catalyst.

Getting started

  • In your virtual environment run
    pip install -e .
  • Check experiment.py for loading train/test data. At the moment the pipeline assumes two JSON lines files containing ['content', 'tagged_attributes'] columns, where tagged_attributes is a list of substrings in content.
  • Possibly modify dataset.py to suit your data preprocessing needs. The pipeline makes assumption that there are two classes of tokens.
  • Start training your model
catalyst-dl run -C bert_ner/config.yml

Monitoring

Run the following command to see metrics in Tensorboard

    CUDA_VISIBLE_DEVICE="" tensorboard --logdir=./logs

About

A barebones (Distil)BERT pipeline for token classification tasks driven by catalyst

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages