Part of Proceedings of Machine Learning and Systems 4 (MLSys 2022)
James Reed, Zachary DeVito, Horace He, Ansley Ussery, Jason Ansel
Modern deep learning frameworks provide imperative, eager execution programming interfaces embedded in Python to provide a productive development experience. However, deep learning practitioners sometimes need to capture and transform program structure for performance optimization, visualization, analysis, and hardware integration. We study the different designs for program capture and transformation used in deep learning. By designing for typical deep learning use cases rather than long tail ones, it is possible to create a simpler framework for program capture and transformation. We apply this principle in torch.fx, a program capture and transformation library for PyTorch written entirely in Python and optimized for high developer productivity by ML practitioners. We present case studies showing how torch.fx enables workflows previously inaccessible in the PyTorch ecosystem.