Terra: Imperative-Symbolic Co-Execution of Imperative Deep Learning Programs
This addresses the usability-performance trade-off for developers using imperative deep learning frameworks, though it is incremental as it builds on prior work combining imperative and symbolic execution.
The paper tackles the limitation of existing systems that cannot fully support imperative deep learning programs with unsupported Python features, by proposing Terra, an imperative-symbolic co-execution system that handles any such programs while achieving optimized performance, speeding up all ten tested programs where a state-of-the-art system fails on five.
Imperative programming allows users to implement their deep neural networks (DNNs) easily and has become an essential part of recent deep learning (DL) frameworks. Recently, several systems have been proposed to combine the usability of imperative programming with the optimized performance of symbolic graph execution. Such systems convert imperative Python DL programs to optimized symbolic graphs and execute them. However, they cannot fully support the usability of imperative programming. For example, if an imperative DL program contains a Python feature with no corresponding symbolic representation (e.g., third-party library calls or unsupported dynamic control flows) they fail to execute the program. To overcome this limitation, we propose Terra, an imperative-symbolic co-execution system that can handle any imperative DL programs while achieving the optimized performance of symbolic graph execution. To achieve this, Terra builds a symbolic graph by decoupling DL operations from Python features. Then, Terra conducts the imperative execution to support all Python features, while delegating the decoupled operations to the symbolic execution. We evaluated the performance improvement and coverage of Terra with ten imperative DL programs for several DNN architectures. The results show that Terra can speed up the execution of all ten imperative DL programs, whereas AutoGraph, one of the state-of-the-art systems, fails to execute five of them.