Fine-Tuning of Large Language Models for Solving DSA Problems Using QLoRA

Fine-tuning large language models (LLMs) for domain specific tasks is often an expensive resource intensive procedure requiring large computing and memory resources. In this paper, We introduce finetuned-leetcode-CodeLlama-7b, a version of CodeLlama that was fine tuned for solving LeetCode problems, specifically. Using QLoRA (Quantized Low Rank Adaptation) techniques, we show that smaller models achieve high performance when solving Data Structure and Algorithm (DSA) problems, providing a cost effective alternative to larger models. The proposed incremental approach reduces memory and computational needs, keeping high accuracy while providing advanced problem-solving in resource constrained environments.