Fine-Tuning Gemma2 Models for Tamil News Translation with HuggingFace BnB, Uploading Model Weights and Utilizing Personal Model with Haystack (All Code Included)

Image generated with Google Labs ImageFX

First of all, Merry Christmas! Hi Mom! Nope, I still haven’t learned Tamil… Instead, I will take the easy road and let LLMs handle it for me. After all, the season is about giving, and what better gift than the joy of making LLMs work for you? Next token generation has been all the hype for nearly two years now and Santa has a sleigh-full of model options in all shapes and sizes. So he’s delivering a complete guide on fine-tuning these smaller LLM models in comprehensive detail (not for the faint of heart). We will unwrap each step of the process from preparing the data to parameter efficient fine-tuning. Even uploading the model weights to a repo and then utilize that now globally available model, locally! ALL for FREE!

NOTE: TL;DR & only looking for code, go to the bottom.

Link for Free Read

This Christmas I am grateful for Open Source

Note: You will need a Google/Gmail Account and a HuggingFace Account to follow this guide. Both are FREE to obtain!

1. Google Colab — Google Colab (Colaboratory) is a cloud-based platform that provides an interactive…