[GitHub] sandeep-krishnamurthy commented on a change in pull request #12691: [MXNET-1001] [WIP] Add new cudnn_tune, cudnn_off params in gluon conv blocks

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

sandeep-krishnamurthy commented on a change in pull request #12691: [MXNET-1001] [WIP] Add new cudnn_tune, cudnn_off params in gluon conv blocks
URL: https://github.com/apache/incubator-mxnet/pull/12691#discussion_r223123762
 
 

 ##########
 File path: python/mxnet/gluon/nn/conv_layers.py
 ##########
 @@ -86,10 +86,15 @@ class _Conv(HybridBlock):
         Initializer for the `weight` weights matrix.
     bias_initializer: str or `Initializer`
         Initializer for the bias vector.
+    cudnn_tune: {None, 'fastest', 'limited_workspace', 'off'}, Optional, default='None'
+        Whether to pick convolution algo by running performance test.
+    cudnn_off: boolean, optional, default=False
 
 Review comment:
   Thanks Naveen for the review. Yes, I will add details for all the options.
   cudnn_off = False => Will ignore cudnn_tune and does not use cudnn for this layer. I will update the document to reflect this.
   
   cudnn_tune=None will use environment variable MXNET_CUDNN_AUTOTUNE_DEFAULT value. By default this environment variable value will be '1' i.e., limited_workspace i.e., choose algo based on memory availability.

----------------------------------------------------------------
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