[GitHub] [incubator-mxnet] zjost commented on issue #19323: Variable sequence length not handled correctly for BiDirectional layers

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

zjost commented on issue #19323:
URL: https://github.com/apache/incubator-mxnet/issues/19323#issuecomment-708519529


   To be clear, this is a problem for us because we use SageMaker and it makes it such that the same record gets different scores when it's invoked via the endpoint as a single record vs in a Batch Transform job.  Running Batch Transform 1 record at a time takes way too long and we can't control how SageMaker splits the batches.  
   
   Also, any comment on point 1, about how this only seems to work with LSTM, not RNN/GRU, even when using correct cuDNN?


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