/llm_trainer

Generic script to fine tune LLMs using HuggingFace

Primary LanguagePythonMIT LicenseMIT

LLM Tuner

Generic trainer script to fine tune LLMs in an alpaca formatted dataset.

The trainer will use the text data in the "text" column to tune the LLM.

Check more about the alpaca format here

Instructions:

  • Install requirements with pip install -r requirements.txt
  • Customize train.py script training variables according to your training specs
  • Run train.py

Credits:

This is based on the work of Abhishek Thakur in the video Train LLMs in just 50 lines of code!

TODOs:

  • Add modifiable command line arguments