[GitHub] [incubator-mxnet] bgawrych commented on issue #21075: Large convolution fails with memory error on CPU but work on GPU with way less memory

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

bgawrych commented on issue #21075:
URL: https://github.com/apache/incubator-mxnet/issues/21075#issuecomment-1178737090

   @matteosal with oneDNN implementation of convolution it also works - seems like you have disabled/not builded oneDNN support.
   Naive implementation of convolution uses im2col which requires ~100GB memory in this case. With GPU you're no facing such issue, because you're probably using cuDNN library implementation. Can you confirm this by posting `mx.runtime.Features()`


-- 
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.

To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org