reminisce commented on a change in pull request #17491: Remove dilation
restriction for conv3d
URL: https://github.com/apache/incubator-mxnet/pull/17491#discussion_r373322284
##########
File path: tests/python/unittest/test_operator.py
##########
@@ -2600,6 +2600,10 @@ def test_convolution_dilated_impulse_response():
for dil in [ (1,1), (2,2), (3,3) ]:
for ks in [ (3,3), (4,4), (2,3), (3,2), (1,1) ]:
test_run_convolution_dilated_impulse_response(dil=dil,
kernel_shape=ks)
+ # 3D
+ for dil in [ (1,1,1), (2,2,2), (3,3,3) ]:
Review comment:
Dilation is required to be an integer greater than 0 in mxnet.
https://github.com/apache/incubator-mxnet/blob/a726c406964b9cd17efa826738a662e09d973972/src/operator/nn/convolution.cc#L240-L241
If it's zero, the kernel size becomes 1. Is there a use case where dilation
needs to be zero? I don't see any technical reasons of not being able to do so
though.
----------------------------------------------------------------
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:
us...@infra.apache.org
With regards,
Apache Git Services
