yelite opened a new pull request, #14809: URL: https://github.com/apache/tvm/pull/14809
This PR adds more logic on bias shape in cuBLAS partitioning. This prevents matmul with residual input to be offloaded to cuBLAS as matmul with bias. cuBLAS only support 1D bias input. This PR also fixes the type key for cublas json runtime, so that it can be deserialized correctly. cc @masahi @vinx13 -- 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]
