bio2

TODO: Actually flesh this README file out

A gross little discord bot based in Python for private use with various entertainment functions, notably GPT2 conversation and text generation

Models I am using were trained on an RTX 3080 running CUDA 10.2 with tensorflow 2.4 and pytorch 1.8 Bot has been tested in the following environment(s):

• Windows 10 Pro
  • RTX 3080
  • tensorflow 2.3
  • torch 1.7.1
  • CUDA 10.0
• Ubuntu 20.04
  • RTX 1070
  • tensorflow 2.3
  • torch 1.7.1
  • CUDA 10.1

Operation Instructions (barebones atm)

• refer to requirements.txt for python library installation
• train.py looks for training data at data/quotes_raw.json
  • json data must be an array of json objects with a parameter text containing each phrase
• existing models or the model generated from train.py must be placed the model folder
• using config.example.py and data.example.json, create config.py and data.json
  • you must provide your own api keys
• for full operation server.py and bio2.py should both be running

Notable Features

• Solves the issue of ML text generation by running a local server instance
• Using an action queue to avoid too many parallel operations
• Ease of source location though a single reaction, reducing redundancy
• Various inappropriate injokes