Thanks @fhieber is your approach for Sockeye's configuration file generalizable 
to something we could build in to mxnet? Can you point me to an example where 
you do this?

> My main question about this though would be how exported HybridBlocks (i.e. 
> SymbolBlocks) behave w.r.t changing input shapes? Before Gluon, we had to 
> make heavy use of the BucketingModule in MXNet to account for varying shapes 
> at inference time.

Nothing would change for that problem with exporting models. The restrictions 
around input shapes is specific to the MXNet engine memory allocation scheme 
where reshaping causes re-allocation. Its much lower level in the stack. If you 
wanna discuss that lets start a separate GitHub issue.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/apache/incubator-mxnet/issues/19535#issuecomment-728231334

Reply via email to