ScaleFold: Reducing AlphaFold Initial Training Time to 10 Hours

TL;DR

ScaleFold is a training method that significantly accelerates AlphaFold's training process, reducing pretraining time from seven days to just 10 hours by optimizing communication and computation overheads.

Contribution

The paper introduces ScaleFold, a systematic training approach that enables efficient scaling of AlphaFold training to hundreds of GPUs, achieving substantial speedups over prior methods.

Findings

ScaleFold trained AlphaFold in 10 hours from scratch.

Achieved over 6x speedup in benchmark tests.

Successfully scaled training to 2080 GPUs.

Abstract

AlphaFold2 has been hailed as a breakthrough in protein folding. It can rapidly predict protein structures with lab-grade accuracy. However, its implementation does not include the necessary training code. OpenFold is the first trainable public reimplementation of AlphaFold. AlphaFold training procedure is prohibitively time-consuming, and gets diminishing benefits from scaling to more compute resources. In this work, we conducted a comprehensive analysis on the AlphaFold training procedure based on Openfold, identified that inefficient communications and overhead-dominated computations were the key factors that prevented the AlphaFold training from effective scaling. We introduced ScaleFold, a systematic training method that incorporated optimizations specifically for these factors. ScaleFold successfully scaled the AlphaFold training to 2080 NVIDIA H100 GPUs with high resource utilization. In the MLPerf HPC v3.0 benchmark, ScaleFold finished the OpenFold benchmark in 7.51 minutes, shown over $6\times$ speedup than the baseline. For training the AlphaFold model from scratch, ScaleFold completed the pretraining in 10 hours, a significant improvement over the seven days required by the original AlphaFold pretraining baseline.