RuRo edited a comment on issue #14762: How to make nn.HybridSequential handel 
multiple input/output layer?(Gluon)
URL: 
https://github.com/apache/incubator-mxnet/issues/14762#issuecomment-590924323
 
 
   ~~It's really bizarre, that `gluon` doesn't support this out of the box. 
Returning a "`tuple of NDArray/Symbol`" is such a common use case and it is 
even supported in some other `mxnet` APIs (i.e. datasets/dataloaders).~~
   
   Edit: Actually, my bad. I retract my previous statement. It seems, that 
passing a `tuple of NDArray/Symbol` as an argument **_is_** currently 
supported, but, there seems to be a bug, that breaks this functionality when 
one of the Hybrid Blocks happens to be a Symbol Block.
   
   On first glance, this may be related to #17488, as currently [`SymbolBlock` 
just outright 
asserts](https://github.com/apache/incubator-mxnet/blob/5098dbe75fa8f86d688667a02d27f9154cf7c924/python/mxnet/gluon/block.py#L1364),
 that its inputs must be `Symbols` or `NDArrays` even though `Block` and 
`HybridBlock` seem to [support nested lists/tuples of 
`Symbols`/`NDArrays`](https://github.com/apache/incubator-mxnet/blob/5098dbe75fa8f86d688667a02d27f9154cf7c924/python/mxnet/gluon/block.py#L172)
 via `_flatten` and `_in_format`.

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