Tencent / TPAT

TensorRT Plugin Autogen Tool
Apache License 2.0
367 stars 42 forks source link

No radical Subgraph optimization for TensorRT #23

Open wugoukanle opened 1 year ago

wugoukanle commented 1 year ago

TPAT in fact is one node by one node to optimize,namely use TVM Ansor to auto tuning one node every time,butTVM is optimized based on subgraph,as follows: 图片 I want to know:radical subgraph optimization for TPAT is work when using TVM subgraph?modify the TVM code can realize this function?

zhuwenxi commented 1 year ago

That's correct. Ansor does auto schedule at the subgraph level.

I also agree subgraph optimization for TPAT is feasible and benifical for users. Infact, we're already working on developing this feature. Hopefully we will realese this feature 2023H1. Any suggestion/proposal are appreciated, thank you!