Pathways: Asynchronous Distributed Dataflow for ML
This system enables more efficient exploration of new ML research ideas and parallelism patterns while maintaining state-of-the-art performance for large-scale models.
The authors tackled the challenge of orchestrating large-scale ML computations across thousands of accelerators by developing Pathways, an asynchronous distributed dataflow system that achieves ~100% accelerator utilization on 2048 TPUs for SPMD computations and maintains comparable throughput for complex Transformer models with pipelining or cross-island sharding.
We present the design of a new large scale orchestration layer for accelerators. Our system, Pathways, is explicitly designed to enable exploration of new systems and ML research ideas, while retaining state of the art performance for current models. Pathways uses a sharded dataflow graph of asynchronous operators that consume and produce futures, and efficiently gang-schedules heterogeneous parallel computations on thousands of accelerators while coordinating data transfers over their dedicated interconnects. Pathways makes use of a novel asynchronous distributed dataflow design that lets the control plane execute in parallel despite dependencies in the data plane. This design, with careful engineering, allows Pathways to adopt a single-controller model that makes it easier to express complex new parallelism patterns. We demonstrate that Pathways can achieve performance parity (~100% accelerator utilization) with state-of-the-art systems when running SPMD computations over 2048 TPUs, while also delivering throughput comparable to the SPMD case for Transformer models that are pipelined across 16 stages, or sharded across two islands of accelerators connected over a data center network.