Kaggle Fine-Tuning Workflow

Overview

Complete pipeline for downloading Kaggle datasets and fine-tuning LLMs. Handles dataset discovery, download via Kaggle CLI, preprocessing into HuggingFace chat format, and training with PEFT/LoRA for memory-efficient fine-tuning.

Prerequisites

pip install kaggle peft transformers accelerate bitsandbytes datasets trl

Set Kaggle API token:

export KAGGLE_API_TOKEN=KGAT_xxxxx

Instructions