sxjscience commented on a change in pull request #18492:
URL: https://github.com/apache/incubator-mxnet/pull/18492#discussion_r436043953
##########
File path: src/operator/nn/cudnn/cudnn_batch_norm-inl.h
##########
@@ -228,7 +228,7 @@ class CuDNNBatchNormOp {
&a,
&b,
&a,
- req[cudnnbatchnorm::kGamma] == kWriteTo ? &b: &b_add,
+ req[cudnnbatchnorm::kGamma] == kAddTo ? &b_add : &b,
Review comment:
Here, we just test with `req[cudnnbatchnorm::kGamma] == kAddTo`. We
won't be able to set the gradient accumulation appropriately if
`req[cudnnbatchnorm::kGamma] == kWriteTo && req[cudnnbatchnorm::kBeta] ==
kAddTo`.
----------------------------------------------------------------
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]
