[Collage] PruneCandidates and demo_collage_partition.py #12105
Merged
Conversation
This file contains hidden or 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
See https://github.com/apache/tvm-rfcs/blob/main/rfcs/0062-collage.md. This completes our checkin of our Collage 'sketch' branch into main. Special thanks to Matthew Barrett for his help getting this over the line. The only C++ functionality added here is for 'pruning' candidates. This is a somewhat speculative algorithm (and I've called that out in the comments) which tries to elide candidate partitions which will 'obviously' not contribute to the final optimal partitioning. For largish models such as GPT2 this can significantly reduce the number of candidates we need to actually measure latency on. I beefed up the MockCostEstimator to make it possible to assert pruning occured from within the test_pass_collage_partition.py unit test. The rest of this PR adds the demo_collage_partition.py driver file we've been using to test and measure perfomance differences against various baseline (though only for the CUDA ecosystem). To eliminate loading time the models of interest are directly expressed in Relay text form in menangerie.py.
jwfromm
reviewed
Jul 15, 2022
|
|
||
| # CAUTION: Requires some changes in python/tvm/autotvm/task/dispatcher.py | ||
| # so that AutoTVM tuning records can be cached between runs and between | ||
| # models. See https://github.com/mbs-octoml/mbs-tvm/tree/mbs-collage-hacks. |
Contributor
There was a problem hiding this comment.
Just noting for posterity, these hacks are needed because autotvm isnt properly caching results? Does that lead to much longer tuning times than necessary or some other breakage?
Contributor
Author
There was a problem hiding this comment.
It's so that the autotvm tuning helpers in demo_collage_partition.py can use the existing tuning records as a cache which can be shared overall all models. Ie a poor man's TRS.
jwfromm
approved these changes
Jul 15, 2022
xinetzone
pushed a commit
to daobook/tvm
that referenced
this pull request
Nov 25, 2022
* [Collage] PruneCandidates and demo_collage_partition.py See https://github.com/apache/tvm-rfcs/blob/main/rfcs/0062-collage.md. This completes our checkin of our Collage 'sketch' branch into main. Special thanks to Matthew Barrett for his help getting this over the line. The only C++ functionality added here is for 'pruning' candidates. This is a somewhat speculative algorithm (and I've called that out in the comments) which tries to elide candidate partitions which will 'obviously' not contribute to the final optimal partitioning. For largish models such as GPT2 this can significantly reduce the number of candidates we need to actually measure latency on. I beefed up the MockCostEstimator to make it possible to assert pruning occured from within the test_pass_collage_partition.py unit test. The rest of this PR adds the demo_collage_partition.py driver file we've been using to test and measure perfomance differences against various baseline (though only for the CUDA ecosystem). To eliminate loading time the models of interest are directly expressed in Relay text form in menangerie.py. * - lint
mikeseven
pushed a commit
to mikeseven/tvm
that referenced
this pull request
Sep 27, 2023
* [Collage] PruneCandidates and demo_collage_partition.py See https://github.com/apache/tvm-rfcs/blob/main/rfcs/0062-collage.md. This completes our checkin of our Collage 'sketch' branch into main. Special thanks to Matthew Barrett for his help getting this over the line. The only C++ functionality added here is for 'pruning' candidates. This is a somewhat speculative algorithm (and I've called that out in the comments) which tries to elide candidate partitions which will 'obviously' not contribute to the final optimal partitioning. For largish models such as GPT2 this can significantly reduce the number of candidates we need to actually measure latency on. I beefed up the MockCostEstimator to make it possible to assert pruning occured from within the test_pass_collage_partition.py unit test. The rest of this PR adds the demo_collage_partition.py driver file we've been using to test and measure perfomance differences against various baseline (though only for the CUDA ecosystem). To eliminate loading time the models of interest are directly expressed in Relay text form in menangerie.py. * - lint
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
See https://github.com/apache/tvm-rfcs/blob/main/rfcs/0062-collage.md.
This completes our checkin of our Collage 'sketch' branch into main. Special thanks
to Matthew Barrett for his help getting this over the line.
The only C++ functionality added here is for 'pruning' candidates. This is a somewhat
speculative algorithm (and I've called that out in the comments) which tries to
elide candidate partitions which will 'obviously' not contribute to the final optimal
partitioning. For largish models such as GPT2 this can significantly reduce the number of
candidates we need to actually measure latency on. I beefed up the MockCostEstimator to
make it possible to assert pruning occured from within the test_pass_collage_partition.py
unit test.
The rest of this PR adds the demo_collage_partition.py driver file we've been using
to test and measure perfomance differences against various baseline (though only
for the CUDA ecosystem). To eliminate loading time the models of interest are directly
expressed in Relay text form in menangerie.py.