samskalicky commented on a change in pull request #17569: Adding sparse support
to MXTensor for custom operators
URL: https://github.com/apache/incubator-mxnet/pull/17569#discussion_r392012930
##########
File path: example/extensions/lib_custom_op/gemm_lib.cc
##########
@@ -140,6 +140,22 @@ MXReturnValue inferType(std::map<std::string,
std::string> attrs,
return MX_SUCCESS;
}
+MXReturnValue inferSType(std::map<std::string, std::string> attrs,
+ std::vector<int> &instypes,
+ std::vector<int> &outstypes) {
+ // validate inputs
+ for (unsigned i = 0; i < instypes.size(); i++) {
+ if (instypes[i] != kDefaultStorage) {
+ std::cout << "Expected input " << i << " to have dense storage type" <<
std::endl;
+ return MX_FAIL;
+ }
+ }
+ for (unsigned i = 0; i < instypes.size(); i++) {
+ outstypes[i] = instypes[0];
Review comment:
Is this going to work? you're looping over instypes.size and using that to
index into outstypes. For Gemm i think we have 2 inputs and 1 output
----------------------------------------------------------------
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]
With regards,
Apache Git Services
