This paper introduces long-context Granite code models that support effective context windows of up to 128K tokens. Our solution for scaling context length of Granite 3B/8B code models from 2K/4K to 128K consists of a light-weight continual pretraining by gradually increasing its RoPE base frequency with repository-level file packing and length-upsampled long-context data. Additionally, we also release instruction-tuned models with long-context support which are derived by further finetuning the long context base models on a mix of permissively licensed short and long-context instruction-response pairs. While comparing to the original short-context Granite code models, our long-context models achieve significant improvements on long-context tasks without any noticeable performance degradation on regular code completion benchmarks (e.g., HumanEval). We release all our long-context Granite code models under an Apache 2.0 license for both research and commercial use.
Cite
@article{arxiv.2407.13739,
title = {Scaling Granite Code Models to 128K Context},
author = {Matt Stallone and Vaibhav Saxena and Leonid Karlinsky and Bridget McGinn and Tim Bula and Mayank Mishra and Adriana Meza Soria and Gaoyuan Zhang and Aditya Prasad and Yikang Shen and Saptha Surendran and Shanmukha Guttula and Hima Patel and Parameswaran Selvam and Xuan-Hong Dang and Yan Koyfman and Atin Sood and Rogerio Feris and Nirmit Desai and David D. Cox and Ruchir Puri and Rameswar Panda},
journal= {arXiv preprint arXiv:2407.13739},
year = {2024}
}