anirudh2290 commented on a change in pull request #15118: Conversion from FP32
model to Mixed Precision model
URL: https://github.com/apache/incubator-mxnet/pull/15118#discussion_r293171585
##########
File path: python/mxnet/module/executor_group.py
##########
@@ -651,6 +652,20 @@ def _bind_ith_exec(self, i, data_shapes, label_shapes,
shared_group):
input_shapes.update(dict(label_shapes))
input_types = {x.name: x.dtype for x in data_shapes}
+ attr_dict = self.symbol.attr_dict()
+
+ for sym_name in self.symbol.list_arguments():
Review comment:
this is actually improving AMP support for the module api. For example after
amp_cast nodes are added dtypes of params won't be able to be inferred easily,
so before we provide a default we check the symbol itself if it has a dtype and
provide that info to simple_bind. I can refactor to make it easier to read and
also think if I can add some test for this.
----------------------------------------------------------------
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
