Fine-tuning Code LLaMA to create a context-aware Python code generation and debugging assistant.
This project aims to fine-tune a large language model (LLM) — specifically Code LLaMA — to perform two tasks:
-
Code Generation Mode:
Generate correct Python code from natural language problem descriptions.
-
Debugging Mode:
Take buggy Python code, identify the errors, fix them, and explain the fix in simple words.
data/ # Datasets: coding problems, buggy codes
scripts/ # Fine-tuning, evaluation, utilities
models/ # Trained models and checkpoints
notebooks/ # Experiment notebooks
results/ # Evaluation results and reports
