SOL: Effortless Device Support for AI Frameworks without Source Code Changes
This addresses the maintenance overhead for vendors and developers in AI by enabling effortless device support across frameworks like PyTorch and TensorFlow, though it is incremental as it builds on existing middleware concepts.
The paper tackles the problem of efficiently supporting diverse hardware accelerators in AI frameworks without requiring source code modifications, introducing SOL as a middleware that provides a hardware abstraction layer and demonstrating it with PyTorch backends for CPUs, GPUs, and vector processors.
Modern high performance computing clusters heavily rely on accelerators to overcome the limited compute power of CPUs. These supercomputers run various applications from different domains such as simulations, numerical applications or artificial intelligence (AI). As a result, vendors need to be able to efficiently run a wide variety of workloads on their hardware. In the AI domain this is in particular exacerbated by the existence of a number of popular frameworks (e.g, PyTorch, TensorFlow, etc.) that have no common code base, and can vary in functionality. The code of these frameworks evolves quickly, making it expensive to keep up with all changes and potentially forcing developers to go through constant rounds of upstreaming. In this paper we explore how to provide hardware support in AI frameworks without changing the framework's source code in order to minimize maintenance overhead. We introduce SOL, an AI acceleration middleware that provides a hardware abstraction layer that allows us to transparently support heterogeneous hardware. As a proof of concept, we implemented SOL for PyTorch with three backends: CPUs, GPUs and vector processors.