rondogency commented on a change in pull request #15969: Partitioning Gluon
HybridBlocks
URL: https://github.com/apache/incubator-mxnet/pull/15969#discussion_r375476278
##########
File path: python/mxnet/gluon/block.py
##########
@@ -967,6 +968,26 @@ def _build_cache(self, *args):
self._cached_op_args.append((False, params[name]))
flags = [('data_indices', data_indices), ('param_indices',
param_indices)] + \
self._flags
+
+ args, _ = _flatten(args, "input")
+ try:
+ for is_arg, i in self._cached_op_args:
+ if not is_arg:
+ i.data()
+ except DeferredInitializationError:
+ self._deferred_infer_shape(*args)
+ for is_arg, i in self._cached_op_args:
+ if not is_arg:
+ i._finish_deferred_init()
+
+ if self._backend:
+ ctx = args[0].context
+ # get list of params in the order of out.list_arguments
+ arg_array = [args[data_names[name]] if name in data_names.keys()
else params[name].data()
+ for name in out.list_arguments()]
+ # Partition the graph.
+ out = out.optimize_for(self._backend, arg_array, ctx,
**self._backend_args)
+
Review comment:
should mention the graph passed in CachedOp is already partitioned using
given backend
----------------------------------------------------------------
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
