Relay: A High-Level Compiler for Deep Learning

TL;DR

Relay is a new high-level compiler framework for deep learning that offers a unified, expressive, and portable IR, enabling efficient model compilation across diverse hardware platforms.

Contribution

Relay introduces a functional, statically typed IR that unifies existing DL IRs and supports domain-specific optimizations for diverse hardware.

Findings

Relay achieves competitive performance on CPUs, GPUs, and accelerators.

The IR design balances expressivity, composability, and portability.

Relay's extension mechanisms facilitate hardware-specific optimizations.

Abstract

Frameworks for writing, compiling, and optimizing deep learning (DL) models have recently enabled progress in areas like computer vision and natural language processing. Extending these frameworks to accommodate the rapidly diversifying landscape of DL models and hardware platforms presents challenging tradeoffs between expressivity, composability, and portability. We present Relay, a new compiler framework for DL. Relay's functional, statically typed intermediate representation (IR) unifies and generalizes existing DL IRs to express state-of-the-art models. The introduction of Relay's expressive IR requires careful design of domain-specific optimizations, addressed via Relay's extension mechanisms. Using these extension mechanisms, Relay supports a unified compiler that can target a variety of hardware platforms. Our evaluation demonstrates Relay's competitive performance for a broad class of models and devices (CPUs, GPUs, and emerging accelerators). Relay's design demonstrates how a unified IR can provide expressivity, composability, and portability without compromising performance.