This project exposes an LSTM model that auto labels Github issues. The goal is to have the full pipeline in continous deployment and rapidly improve the model accuracy. This is an independant project to clear the cloud on how to actually build an ML model and understand what it takes for taking it from 0 - 1 with continues integration.
Following Octoverse, we looked at the "Ten most discussed repositores" and trained on the closed issues.
After looking at Github's default labels and what's being used in the projets above we as a first pass are training. See Labels
bug
question
enhancement
feature
help wanted
doc
We are picking the title and mapping the existing labels in the issues above to our baseline version. We are using LSTM
We are using pytorch v0.3.1
Here you go and the backend API sits here
TODO
- Train more labels
- Output multi labels