szha commented on a change in pull request #7356: decouple record/train and add 
state readers
URL: https://github.com/apache/incubator-mxnet/pull/7356#discussion_r131586398
 
 

 ##########
 File path: python/mxnet/autograd.py
 ##########
 @@ -112,6 +137,34 @@ def pause(is_train=False):
     return RecordingStateScope(False, is_train)
 
 
+def override_train():
 
 Review comment:
   I agree that we should only introduce features that users need. Just to 
elaborate on the reasoning: set_training() is currently action-based API, and 
users will need to manually revert the values if the change is intended for 
only a small subset, and thus will need to keep the previous states. 
Introducing train() and predict() scopes can help making this simpler since the 
"with" scope is a natural fit for such task. @zackchase what's your view as a 
user?
 
----------------------------------------------------------------
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:
us...@infra.apache.org


With regards,
Apache Git Services

Reply via email to