tqchen commented on a change in pull request #6334:
URL: https://github.com/apache/incubator-tvm/pull/6334#discussion_r487609121
##########
File path: python/tvm/contrib/graph_runtime.py
##########
@@ -43,24 +43,34 @@ def create(graph_json_str, libmod, ctx):
be used as this purpose. All context should be given for heterogeneous
execution.
+ force_local_graph_runtime : bool
+ When ctx contains only rpc contexts, the GraphRuntime is typically
instantiated
Review comment:
We are migrating away from graph_runtime.create, and instead relying on
the more general module-based interface -
https://github.com/apache/incubator-tvm/blob/master/tests/python/unittest/test_runtime_module_based_interface.py#L102
The idea is to remove layer of abstractions and only make GraphModule a thin
wrapper. Without having to couple the decision of module creation with the
GraphModule wrapping.
So it is preferrable to not complicate the graph_runtime.create. It would be
useful to have a function in micro tvm that creates the module, then call into
GraphModule to wrap the resulting constructed module.
----------------------------------------------------------------
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]
