Breaking the Computation and Communication Abstraction Barrier in Distributed Machine Learning Workloads

TL;DR

This paper introduces CoCoNeT, a high-level framework that unifies computation and communication in distributed machine learning, enabling advanced optimizations and significantly improving performance in training large models.

Contribution

CoCoNeT provides a novel DSL and compiler that treat computation and communication as first-class citizens, facilitating cross-layer optimizations in distributed ML workloads.

Findings

CoCoNeT outperforms existing distributed ML implementations.

Enables optimization of data, model, and pipeline parallelism.

Achieves significant performance improvements with minimal code changes.

Abstract

Recent trend towards increasing large machine learning models require both training and inference tasks to be distributed. Considering the huge cost of training these models, it is imperative to unlock optimizations in computation and communication to obtain best performance. However, current logical separation between computation and communication kernels in deep learning frameworks misses the optimization opportunities across such barrier. Breaking this abstraction with a holistic consideration can provide many optimizations to provide performance improvements in distributed workloads. Manually applying these optimizations needs modifications in underlying computation and communication libraries for each scenario, which is time consuming and error-prone. Therefore, we present CoCoNeT, with a DSL to express a program with both computation and communication. CoCoNeT contains several machine learning aware transformations to optimize a program and a compiler to generate high performance kernels. Providing both computation and communication as first class constructs allows users to work on a high-level abstraction and apply powerful optimizations, such as fusion or overlapping of communication and computation. CoCoNeT enables us to optimize data-, model-and pipeline-parallel workloads in large language models with only a few lines of code. Experiments show CoCoNeT significantly outperforms state-of-the-art distributed machine learning implementations.