zhuwenxi commented on PR #11326: URL: https://github.com/apache/tvm/pull/11326#issuecomment-1144389118
@comaniac Actually there is still a major issue hasn't been addressed in current my implementation: how to enable libxsmm only in the inference mode? That's because libxsmm doesn't have a DENSE implementation, so we have to use GEMM instead which requires a explicit weight transpose. This won't be a problem in inference mode because the weight is frozen thus we can pre-transpose and store the transposed weight, no performance penaty at all. But I failed to find out a way to check weather the functional-style relay graph is in the inference mode. Do you have any suggestions? -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
