pybrain based gpt-2/3, version 0.1.0, Stable release:

in this release:

• scipy & pybrain3 modules
• github_import.py for importing python code to use from github by branch, repo, & username, plus file on git in that repo
• main.py housing the logic for the pybrain3 version of GPT2/3
• string.py from python 3.6.9 to be used by scipy module.

Not in this release:

• Training data (you'll need to find the data or make the data needed for training the model)
• transformers, huggingfaces, or the like modules
• matplotlib
• numpy

Train and use how you will, however, paperspace-gradient has a great setup for using code like this to train your very own GPT2/3 model or test alternative implementations. Before using this code, this release comes in 2 formats.

The first format is for testing on platforms like docker or paperspace-gradient or your own local machine.
The second format is configured to work with replit containers.

I uploaded both so you could customize further if needed by platform or duplicate a repltit container from a zip file.