[GitHub] zachgk commented on issue #12487: [Scala] PrefetchingIter fails with labels and predictions should have the same length

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

zachgk commented on issue #12487: [Scala] PrefetchingIter fails with labels and predictions should have the same length
URL: https://github.com/apache/incubator-mxnet/issues/12487#issuecomment-453707698
 
 
   @mariussoutier The PrefetchingIter has an IndexedSeq of inputs because some networks might require multiple data inputs. That is why the elements are read in parallel and there must be the same number of elements from each component iterator (since the network won't function with some inputs missing). You can see an example of the equivalent python function at https://mxnet.incubator.apache.org/api/python/io/io.html.
   
   I think you are wanting to combine the two iterators such that it will feed the network one input that comes from either of the source iterators. In that case, you might want to create a new DataIter.
   
   If you did intend to read them in parallel but allow missing elements, you need to modify one or both of the iterators to return a default for the missing elements. 

----------------------------------------------------------------
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:
users@infra.apache.org


With regards,
Apache Git Services