FrozenGene commented on a change in pull request #7233:
URL: https://github.com/apache/tvm/pull/7233#discussion_r559881396
##########
File path: python/tvm/autotvm/measure/measure_methods.py
##########
@@ -560,9 +561,11 @@ def run_through_rpc(
raise AttributeError(
"Please make sure USE_RANDOM is ON in the config.cmake " "on
the remote devices"
)
- args = [nd.empty(x[0], dtype=x[1], ctx=ctx) for x in
build_result.arg_info]
- for arg in args:
- random_fill(arg)
+ args = [nd.array(np.zeros(x[0], dtype=x[1]), ctx=ctx) for x in
build_result.arg_info]
Review comment:
This solution maybe have potential problem in the future. If we import a
model from other frontend(like tf), if our compiler could leverage the model
layer name and become into compiler func name so that we could debug (profile)
/ partition subgraph more easily, maye we couldn’t see ‘scatter' any more.
----------------------------------------------------------------
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]
