CodeACT: Code Adaptive Compute-efficient Tuning Framework for Code LLMs

TL;DR

CodeACT is a framework that improves open-source code language models by selecting high-quality training data and optimizing training strategies, resulting in better performance and reduced computational costs.

Contribution

It introduces the CDAS data sampling method and Dynamic Pack padding, significantly enhancing training efficiency and model performance with less data and resources.

Findings

8.6% performance increase on HumanEval

78% reduction in training time

27% decrease in peak GPU memory usage

Abstract

Large language models (LLMs) have shown great potential in code-related tasks, yet open-source models lag behind their closed-source counterparts. To bridge this performance gap, existing methods generate vast amounts of synthetic data for fine-tuning, leading to inefficiencies in training. Motivated by the need for more effective and efficient training, we propose the Code Adaptive Compute-efficient Tuning (CodeACT) framework. CodeACT introduces the Complexity and Diversity Aware Sampling (CDAS) method to select high-quality training data based on complexity and diversity, and the Dynamic Pack padding strategy to reduce computational resource usage by minimizing padding tokens during training. Experimental results demonstrate that CodeACT-DeepSeek-Coder-6.7B, fine-tuned on only 40% of the EVOL-Instruct data, achieves an 8.6% performance increase on HumanEval, reduces training time by 78%, and decreases peak GPU memory usage by 27%. These findings underscore CodeACT's ability to enhance the performance and efficiency of open-source models. By optimizing both the data selection and training processes, CodeACT offers a comprehensive approach to improving the capabilities of open-source LLMs while significantly reducing computational requirements, addressing the dual challenges of data quality and training efficiency, and paving the way for more resource-efficient and performant models.