LLM-and-RAG-Algorithms

The objective of this repository is to gather methods to build Large Language Models (LLMs) and the Retrieval Augmented generation (RAG) process. These models will be pulled and tracked in the Hugging Face repository.

How to create the Python environment locally

To install the python environment locally for this project, use the following command (in command prompt for Windows and bash terminal for Linux):

conda env create -f environment.yml

Training LLMs on free Google Colab GPUs

If you don't have GPUs on your local computer, Google Colab provides free GPUs (with limits). You can clone this repository following Ashwin's Medium article below.

How to use Google Colaboratory to clone a GitHub Repository to your Google Drive?

To set up your environment on Google Colab, each notebook will have the code snippet below to install the needed dependencies in the requirements.txt if ran in a Google colab notebook. Be sure to replace the <path to requirements.txt> code with the actual path to the requirements.txt in the Google drive.

import sys
import subprocess
import os

IN_COLAB = 'google.colab' in sys.modules

if IN_COLAB:
    print("Running in Google Colab and installing dependencies")
    from google.colab import drive
    drive.mount('/content/gdrive')
    !pip install -r <path to requirements.txt>
else:
    print("Not running in Google Colab")