[GitHub] KnightOfTheMoonlight commented on issue #9110: 3D dilation support

[GitBox <gi...@apache.org>]

KnightOfTheMoonlight commented on issue #9110: 3D dilation support
URL: https://github.com/apache/incubator-mxnet/issues/9110#issuecomment-362143973
 
 
   @chinakook Thanks for your suggestion. As I can remember mxnet has said 3d conv related things are based on cudnn calculation. But I still tried setting cudnn_off.
   Here is the code: 
   ```
   import mxnet as mx
   print(mx.__version__)
   
   data = mx.nd.arange(1029).reshape((3,1,7,7,7)) # NCDHW
   data_shape=(3,1,7,7,7)
   x = mx.sym.Variable('data')
   conv3D = mx.sym.Convolution(data=x, kernel=(3, 3, 3), pad=(2, 2, 2), dilate=(2, 2, 2), num_filter=2, cudnn_off=True)
   conv3D_exe = conv3D.simple_bind(mx.gpu(),data=data.shape)
   ```
   
   Here is the related error:
   ```
   RuntimeError: simple_bind error. Arguments:
   data: (3L, 1L, 7L, 7L, 7L)
   Error in operator convolution0: [13:42:28] src/operator/./convolution-inl.h:495: Check failed: param_.dilate.Size() == 1U (8 vs. 1) Dilate is not supported in 3d convolution
   ```
   

----------------------------------------------------------------
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:
users@infra.apache.org


With regards,
Apache Git Services