Join us for this interactive PyTorch Live Q&A series where we discuss various topics that are top of mind for the PyTorch Community.
This talk first dives deep into TorchInductor design principles, lowering and codegen followed by the backend integration points in Pytorch compiler stack. It will explain three types of IR used across the stack:
- torch IR produced by Dynamo
- AtenIR produced by AoTAutograd
- Loop-level IR used in Inductor
We will also introduce the infrastructure and utilities available for backend integration, including a IR-agnostic Pattern Matcher and a Graph Partitioner.
Review in advance: https://hubs.la/Q01yGw_B0