LLM Fine-Tuning & Conversion Notebooks

This repository contains Jupyter notebooks for fine-tuning large language models using different approaches and for converting trained models to LiteRT format for edge/mobile deployment.

---

📂 Notebooks Overview

🔹 Fine-tuning with Unsloth

These notebooks use the Unsloth framework for efficient LoRA/QLoRA fine-tuning on GPUs with reduced memory usage.

• functionGemma_fineTune.ipynb
  Fine-tunes the FunctionGemma model using Unsloth for instruction/function-calling style tasks.

• qwen_8b_ocrLatex_fineTune.ipynb
  Fine-tunes the Qwen-8B model using Unsloth for OCR/LaTeX or document understanding tasks.

---

🔹 Fine-tuning with Vanilla Hugging Face

This notebook uses standard Hugging Face transformers, datasets, and Trainer APIs.

• Fine_tune_a_language_model.ipynb
  End-to-end example of fine-tuning a language model using vanilla Hugging Face without Unsloth.

---

🔹 Model Conversion to LiteRT

Used to convert a fine-tuned model into LiteRT format for lightweight inference on edge/mobile devices.

• gemma_to_litertlm.ipynb
  Converts the model produced by functionGemma_fineTune.ipynb into LiteRTLM format.

Shiv Prakash Verma