[GitHub] [incubator-mxnet] Neutron3529 commented on issue #19333: [Bug] cudnn GRU backward errors on train_mode=False

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

Neutron3529 commented on issue #19333:
URL: https://github.com/apache/incubator-mxnet/issues/19333#issuecomment-707618663


   where is the document for `train_mode` parameter for `backward` function?
   I tried to figure out it, but find nothing:
   [Here](https://mxnet.apache.org/versions/1.7.0/api/python/docs/_modules/mxnet/executor_manager.html#DataParallelExecutorGroup.backward) is what I found
   ```
   
   
   [docs]
       def forward(self, is_train=False):
           """Perform a forward pass on each executor."""
           for texec in self.train_execs:
               texec.forward(is_train=is_train)
   
   
   [docs]
       def backward(self):
           """Perform a backward pass on each executor."""
           for texec in self.train_execs:
               texec.backward()
   ```
   there is no `train_mode=False` for backward.
   
   BTW, why mark `train_mode=False` to a backward function?
   the only meaning of a backward process is to train the network


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



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