merrymercy commented on a change in pull request #4151: [AutoTVM] Added an
autotuning Config Library to store autotune results
URL: https://github.com/apache/incubator-tvm/pull/4151#discussion_r357408970
##########
File path: python/tvm/autotvm/tuner/tuner.py
##########
@@ -104,13 +109,25 @@ def tune(self, n_trial, measure_option,
early_stopping=None, callbacks=()):
(Tuner, List of MeasureInput, List of MeasureResult)
with no return value. These callback functions will be called on
every measurement pair. See autotvm/tuner/callback.py for some
examples.
+ allow_skipping: bool, optional
+ Allow the tuner to skip a task if a sufficiently well-tuned entry
already
+ exists in a config library.
+
"""
+ early_stopping = early_stopping or 1e9
+ if GLOBAL_SCOPE.tuning_job:
Review comment:
This is very hacky. We should not use a global variable to hack the behavior
of a tuner.
----------------------------------------------------------------
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
