[Ansor][AutoTVM v2.0] Phase 1: The base class for cost models #6187
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
merrymercy
changed the title
merrymercy
force-pushed
the
pr-cost-model
branch
from
9203263 to
e3eb8b3
Compare
merrymercy
force-pushed
the
pr-cost-model
branch
from
e3eb8b3 to
5a49ec7
Compare
|
Will review Tuesday night |
jcf94
reviewed
Aug 3, 2020
| * \param state_scores The predicted scores for all states | ||
| * \param stage_scores The predicted scores for all stages in all stages | ||
| */ | ||
| virtual void PredictStages(const SearchTask& task, const std::vector<State>& states, |
There was a problem hiding this comment.
Do we still need this part? Since out current search policy only apply cost model on a complete schedule.
There was a problem hiding this comment.
This is needed by crossover, and it should be useful for future developments IMO.
There was a problem hiding this comment.
@merrymercy would that be better to also make this as a virtual function, or setup a flag or something to indicate if this cost model supports PredictStages or not instead of just throwing an error?
There was a problem hiding this comment.
Yeah, your suggestion seems to be better.
comaniac
approved these changes
Aug 3, 2020
| * \param state_scores The predicted scores for all states | ||
| * \param stage_scores The predicted scores for all stages in all stages | ||
| */ | ||
| virtual void PredictStages(const SearchTask& task, const std::vector<State>& states, |
There was a problem hiding this comment.
@merrymercy would that be better to also make this as a virtual function, or setup a flag or something to indicate if this cost model supports PredictStages or not instead of just throwing an error?
jcf94
approved these changes
Aug 4, 2020
FrozenGene
reviewed
Aug 6, 2020
Disable test if user doesn't have llvm
FrozenGene
approved these changes
Aug 6, 2020
wjliu1998
pushed a commit
to wjliu1998/incubator-tvm
that referenced
this pull request
Aug 13, 2020
…#6187) * add the base class for cost models * address comments * Update tests/python/unittest/test_auto_scheduler_cost_model.py Disable test if user doesn't have llvm Co-authored-by: Zhao Wu <[email protected]>
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Aug 26, 2020
…#6187) * add the base class for cost models * address comments * Update tests/python/unittest/test_auto_scheduler_cost_model.py Disable test if user doesn't have llvm Co-authored-by: Zhao Wu <[email protected]>
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Aug 26, 2020
…#6187) * add the base class for cost models * address comments * Update tests/python/unittest/test_auto_scheduler_cost_model.py Disable test if user doesn't have llvm Co-authored-by: Zhao Wu <[email protected]>
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Aug 26, 2020
…#6187) * add the base class for cost models * address comments * Update tests/python/unittest/test_auto_scheduler_cost_model.py Disable test if user doesn't have llvm Co-authored-by: Zhao Wu <[email protected]>
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Sep 2, 2020
…#6187) * add the base class for cost models * address comments * Update tests/python/unittest/test_auto_scheduler_cost_model.py Disable test if user doesn't have llvm Co-authored-by: Zhao Wu <[email protected]>
trevor-m
pushed a commit
to neo-ai/tvm
that referenced
this pull request
Sep 3, 2020
…#6187) * add the base class for cost models * address comments * Update tests/python/unittest/test_auto_scheduler_cost_model.py Disable test if user doesn't have llvm Co-authored-by: Zhao Wu <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
For the full upstream plan, see Ansor RFC.
This PR adds the base class for cost models and a random cost model.
Because learning-based models typically use python packages (e.g., xgboost, pytorch, tensorflow), we prefer to implement them in python and use the c++ wrapper PythonBasedCostModel for calling them in c++.
The later PR will introduce a cost model based on manual features and xgboost.