JANUS: Fast and Flexible Deep Learning via Symbolic Graph Execution of Imperative Programs
This addresses the problem for deep learning researchers and practitioners who face trade-offs between performance and programmability in existing frameworks, representing a novel method rather than an incremental improvement.
The paper tackles the challenge of deep learning frameworks needing both high performance and usability by introducing JANUS, a system that converts imperative Python programs into efficient symbolic dataflow graphs, achieving fast training while maintaining flexibility.
The rapid evolution of deep neural networks is demanding deep learning (DL) frameworks not only to satisfy the requirement of quickly executing large computations, but also to support straightforward programming models for quickly implementing and experimenting with complex network structures. However, existing frameworks fail to excel in both departments simultaneously, leading to diverged efforts for optimizing performance and improving usability. This paper presents JANUS, a system that combines the advantages from both sides by transparently converting an imperative DL program written in Python, the de-facto scripting language for DL, into an efficiently executable symbolic dataflow graph. JANUS can convert various dynamic features of Python, including dynamic control flow, dynamic types, and impure functions, into elements of a symbolic dataflow graph. Experiments demonstrate that JANUS can achieve fast DL training by exploiting the techniques imposed by symbolic graph-based DL frameworks, while maintaining the simple and flexible programmability of imperative DL frameworks at the same time.