lfengad commented on issue #4990: [TF][Relay] BatchNorm support with run-time 
mean and variance calculation
URL: https://github.com/apache/incubator-tvm/pull/4990#issuecomment-595147201
 
 
   > I think we needn't add `_empty_for_training_mode_inference`. If we find 
`mean` / `variance` is `VarNode`, we should call `Mean` and `Variance`.
   > 
   > I don't think we should add `is_training` flag to relay `BatchNorm`. This 
should be done by users to make sure TF's model `BatchNorm`'s `is_training` 
flag be false. However, we still have user cases like you mention, so we could 
support as current implementation and don't add attribute to `BatchNorm`.
   
   Thank you so much for the quick reply!  
   Yeah, our current implementation is just to check whether `mean` / 
`variance` is empty `VarNode` (with zero dimension),  and then call `Mean` and 
`Variance` in BatchNormToInferUnpack.  Also as I understand, if `mean` / 
`variance` is `VarNode` but with non-zero dimension, it still has the 
possibility to hold the given pre-defined constant values and thus cannot be 
replaced with `Mean` \ `Variance`.   
   Thank you for the discussion!

----------------------------------------------------------------
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:
[email protected]


With regards,
Apache Git Services

Reply via email to