Jerryzcn commented on a change in pull request #11834: Fix mxnet ctc_loss bug
URL: https://github.com/apache/incubator-mxnet/pull/11834#discussion_r205707680
##########
File path: src/operator/contrib/ctc_include/detail/gpu_ctc.h
##########
@@ -411,12 +411,7 @@ GpuCTC<ProbT>::compute_log_probs(const ProbT* const
activations) {
denoms_, out_dim_, num_elements);
// compute denominators for softmax
- denoms_handle = reduce_with_axis<red::sum, false>(
- F<mxnet::op::mshadow_op::exp>(
- log_probs_handle -
- broadcast<0>(reduce_with_axis<red::maximum,
false>(log_probs_handle, 1),
Review comment:
I see. Did not look at other parts of the code thanks! Let me check why my
model is not converging
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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
