samskalicky commented on a change in pull request #15969: Partitioning Gluon
HybridBlocks
URL: https://github.com/apache/incubator-mxnet/pull/15969#discussion_r374323086
##########
File path: python/mxnet/gluon/block.py
##########
@@ -954,6 +955,22 @@ def _build_cache(self, *args):
unused = ', '.join(list(param_names - set(used_param_names)))
warnings.warn("Parameter %s is not used by any computation. "
"Is this intended?"%unused, stacklevel=4)
+ if self._backend:
+ arg_array = []
+ ctx = args[0].context
+ try:
+ for name in out.list_arguments():
+ if name in data_names.keys():
+ arg_array.append(args[data_names[name]])
+ else:
+ arg_array.append(params.get(name).data())
+
+ except DeferredInitializationError:
Review comment:
Plus, during training we dont have the params yet (shapes/types) so we need
to partition without them. For inference we do have the shapes/types, so it
makes sense to have a flow that works for both use-cases.
The question is how we enable this flow with Gluon-style handling (ie. just
make it work internally by catching the exception, or have the user pass in a
flag to explicitly skip the attempt to get the params)
----------------------------------------------------------------
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
