gautham-kollu opened a new pull request #12512: [MXNET-938] Allow fusing weight 
ahead of calling forward()
URL: https://github.com/apache/incubator-mxnet/pull/12512
 
 
   ## Description ##
   Issue : MXNET-938
   Root Cause : _rnn_param_concat() takes a while to concat all the params.
   
   This PR allows the user to call .fuse_weights() when running Gluon with 
ndarray there-by preprocessing the fusing-of-weights before calling forward.  
Also, when running inference fusing-of-weights is not required for each batch - 
only once is suffice.
   
   Some of the alternative approaches are 
    1. Fuse Parameters() when created.
    2. Fuse the weights in ".initialize()" if we are running in inference mode 
and if all the parameters are initialized. 
   
   Sending a PR as a first-draft to get some early feedback.

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


With regards,
Apache Git Services

Reply via email to