ANSHUMAN87 commented on pull request #7048: URL: https://github.com/apache/tvm/pull/7048#issuecomment-741855916
> Thanks @ANSHUMAN87. Overall looks pretty good, but I am a little confused at exactly what is happening is this pass. Is it converting matrices from sparse to dense at compile time? Thanks @tkonolige for review! The Densify Op does not need to be present in TVM runtime currently, because its purpose is to convert sparse weights(constant) to dense form. So i have optimized it out. However in case it is required in future to be present in TVM runtime, we can always do it with sparse_to_dense Op with the indices produced from the utility prepare_dense_matrix_from_sparse(). As to your other comments i have tried to add comments in the necessary places. Please have a look. In case anything more required. Please let me know. TIA! ---------------------------------------------------------------- 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]
