masahi edited a comment on pull request #7137: URL: https://github.com/apache/tvm/pull/7137#issuecomment-750766311
> How about the perf numbers with default value of `rpn_pre_nms_top_n_test`? It is dominated by NMS which takes 2.1 seconds with default `rpn_pre_nms_top_n_test`. But even in the default setting, PyTorch NMS is fairly fast and it is not a bottleneck at all. So the root issue is our terrible GPU performance. We need to fix our NMS and there is no excuse to cheat in the frontend. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
