sxjscience commented on a change in pull request #14912: Add fp16 and fp64
support for topk
URL: https://github.com/apache/incubator-mxnet/pull/14912#discussion_r285424599
##########
File path: src/operator/tensor/ordering_op-inl.h
##########
@@ -401,8 +401,6 @@ void TopKImpl(const RunContext &ctx,
mxnet::op::SortByKeyWorkspaceSize<int, int, xpu>(src.Size()));
temp_size = std::max(temp_size,
mxnet::op::SortByKeyWorkspaceSize<int, DType, xpu>(src.Size()));
- temp_size = std::max(temp_size,
- mxnet::op::SortByKeyWorkspaceSize<DType, int, xpu>(src.Size()));
Review comment:
Removing this line may break some corner cases uncovered by the existing
test. Since we need to call `SortByKey<DType, int>(dat, ind, ...)`,
`SortByKey<int, DType>(batch_id, dat, ...)` and `SortByKey<int, int>(batch_id,
ind, ...)`, we should make sure that the temporary storage has enough size for
all cases.
----------------------------------------------------------------
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
