I've trained a neural net on the text messages of my families WhatsApp group. This framework can be used for similar input data.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Markus Becker 182f96393d Mention dependencies and goals 11 months ago
README.md Mention dependencies and goals 11 months ago
clean.sh Initial Commit 11 months ago
generator.py Initial Commit 11 months ago



This is a framework built on top of textgenrnn which simplyfies training a network on simple text data using a command line interface. The input data still needs to be cleaned, but training and generation of samples is done through command line arguments.


  1. Get text file with messages (one-per-line)
  2. Clean them of symbols you don’t want to train on
    • You can use clean.sh as a reference
  3. The generator.py expects the cleaned text file to be at cleaned.txt
  4. Run the generator.py file with python3, check it’s help text for further help