You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2022/06/24 19:57:23 UTC

[GitHub] [tvm] zxybazh opened a new pull request, #11892: [MetaSchedule] Enable Adapative Training For XGBoost Cost Model

zxybazh opened a new pull request, #11892:
URL: https://github.com/apache/tvm/pull/11892

   CostModel retraining is a time consuming part for MetaSchedule tuning, similar to AutoScheduler, we can alleviate it with an adapative way of increasing waiting period between each retraining. This PR introduced an argument called `adpative_training` in `TuneConfig` and the constructor of `XGBoostModel` to enable the capability. Testing tuning scripts are also updated.


-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] junrushao1994 commented on pull request #11892: [MetaSchedule] Enable Adapative Training For XGBoost Cost Model

Posted by GitBox <gi...@apache.org>.
junrushao1994 commented on PR #11892:
URL: https://github.com/apache/tvm/pull/11892#issuecomment-1168278416

   please lmk when the PR is ready for review!


-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] junrushao1994 commented on a diff in pull request #11892: [MetaSchedule] Enable Adapative Training For XGBoost Cost Model

Posted by GitBox <gi...@apache.org>.
junrushao1994 commented on code in PR #11892:
URL: https://github.com/apache/tvm/pull/11892#discussion_r908087215


##########
python/tvm/auto_scheduler/search_task.py:
##########
@@ -481,7 +481,7 @@ def __init__(
             desc,
         )
 
-    def tune(self, tuning_options, search_policy=None):
+    def tune(self, tuning_options, search_policy=None, adaptive_training=False):

Review Comment:
   i see...that makes sense. let's keep the flag here for tune_te then :-)



-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] zxybazh commented on a diff in pull request #11892: [MetaSchedule] Enable Adapative Training For XGBoost Cost Model

Posted by GitBox <gi...@apache.org>.
zxybazh commented on code in PR #11892:
URL: https://github.com/apache/tvm/pull/11892#discussion_r908083091


##########
python/tvm/auto_scheduler/search_task.py:
##########
@@ -481,7 +481,7 @@ def __init__(
             desc,
         )
 
-    def tune(self, tuning_options, search_policy=None):
+    def tune(self, tuning_options, search_policy=None, adaptive_training=False):

Review Comment:
   This only impacts `tune_te`, which does not call thru `TaskScheduler.tune` but `SearchTask.tune` instead. If we don't want to enable that for `tune_te` I can just remove that change.



-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] junrushao1994 commented on pull request #11892: [MetaSchedule] Enable Adapative Training For XGBoost Cost Model

Posted by GitBox <gi...@apache.org>.
junrushao1994 commented on PR #11892:
URL: https://github.com/apache/tvm/pull/11892#issuecomment-1168280544

   Otherwise LGTM


-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] junrushao1994 merged pull request #11892: [MetaSchedule] Enable Adapative Training For XGBoost Cost Model

Posted by GitBox <gi...@apache.org>.
junrushao1994 merged PR #11892:
URL: https://github.com/apache/tvm/pull/11892


-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] junrushao1994 commented on a diff in pull request #11892: [MetaSchedule] Enable Adapative Training For XGBoost Cost Model

Posted by GitBox <gi...@apache.org>.
junrushao1994 commented on code in PR #11892:
URL: https://github.com/apache/tvm/pull/11892#discussion_r908078161


##########
python/tvm/auto_scheduler/search_task.py:
##########
@@ -481,7 +481,7 @@ def __init__(
             desc,
         )
 
-    def tune(self, tuning_options, search_policy=None):
+    def tune(self, tuning_options, search_policy=None, adaptive_training=False):

Review Comment:
   probably we dont need to adjust auto_scheduler's interface



-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org