whoiswillma/nlp-experiments

nlp-experiments

To get set up,

1. Clone this repo

2. pip install -r requirements.txt

3. [Optional] Clone https://github.com/whoiswillma/fewnerdparse, https://github.com/whoiswillma/wifineparse and set them up as well. With all the repos, your directory structure might look something like:

   fewnerdparse
   fewnerdparse/inter
   fewnerdparse/intra
   fewnerdparse/supervised
   nlp-experiments
   wifineparse
   wifineparse/Documents
   wifineparse/FineNE
   

Style stuff:

• 80 columns is a soft guide.
• Typically a good idea to add type annotations, especially for methods which take in common data structures such as dict, list, etc.
  • Less necessary for tokenizers and models because those types are often clear from the file you're working in.
• Tests for "pure" methods are nice.
• Even though there is close collaboration, it's still a good idea to branch and PR because
  • It let's you document the changes you made e.g. for weekly reports
  • Batch all your changes into one place for others to see
• util.py contains code for loading indicators, logging, and checkpointing

Linting

black

This project uses black as code formatter.

black .

flake8

We also use flake8 as an aid. Use it as follows.

flake8 --config=.github/linters/.flake8 .