sui-hs opened a new issue #14710: BatchNorm gets inconsistent result between 
Module API and NDArray API 
URL: https://github.com/apache/incubator-mxnet/issues/14710
 
 
   ## Description
   when using BatchNorm layer in NN, I found that the results differ in 
Module(Symbol) API and NDArray API. 
   
   ## Environment info (Required)
   MxNet version: 1.3.0
   
   ## Build info (Required if built from source)
   
   ## Error Message:
   1. [with this 
script](https://gist.github.com/sui-hs/ec86ef6622e8363908ebe4202a335567) the 
BatchNorm result is:
   [[-1.1049458   0.45291406]
    [ 1.3138113  -1.3863459 ]
    [-0.20886569  0.9334318 ]]
   
   2. [with this 
script](https://gist.github.com/sui-hs/b36ee22028324b8a6fef6416033e48ed), I got:
   [[-34.330654   31.538832 ]
    [  8.9485655 -47.63322  ]
    [-18.296946   52.223007 ]]
   
   ## Minimum reproducible example
   
   ## Steps to reproduce
   (Paste the commands you ran that produced the error.)
   
   1. run the first script
   2. run the second script
   
   ## What have you tried to solve it?
   
   1. I tried Executor API directly with the first model, and got the second 
result.
   2. I tried with TensorFlow, and got the second result
   

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