anirudh2290 commented on a change in pull request #15886: Graph Partition API
URL: https://github.com/apache/incubator-mxnet/pull/15886#discussion_r317858868
##########
File path: python/mxnet/symbol/symbol.py
##########
@@ -1437,6 +1437,54 @@ def _gen_atomic_symbol(self):
return Symbol(handle)
+ def optimize_for(self, backend, args=None, **kwargs):
Review comment:
Haven't looked at the whole PR yet since this has been still tagged WIP.
Let me know if its tagged incorrectly. Can you explain the reasoning behind
optimize_for API triggering infer shape type when ndarrays are passed. All the
inference APIs currently take either shape or type. For example if I want to
call graph partitioning on my symbol but I dont have a pretrained model to load
arg_params from and then i will provide a shape and type dict to partition
followed by simple_bind. Also, partial inference is important here as mentioned
in other comments otherwise this may have many failure scenarios.
----------------------------------------------------------------
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
