[Collage] CombinerRule and CandidatePartition::EstimateCost #12078
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
|
@mbaret when you get a chance, thanks. |
|
CI failure is a flake. |
See https://github.com/apache/tvm-rfcs/blob/main/rfcs/0062-collage.md. We complete the PartitionRule sub-class hierarchy with the addition of CombinePartitionRule, which allows disjoint candidate partitions to be unioned based on simple rules. - By TOpPattern kind, eg a kOutElemwiseFusable and kBroadcast. - A tuple argument with injective fields. - The projection from an injective group (obviously of tuple type) - Combinations of the above. These let us mimic many common fusion strategies, including TVMs, so that the candidates explored during Collage search are as large as possible to expose possible fusion opportunities but no larger. Also completes CandidatePartition with the EstimateCost method, which is used during search to construct a stand-alone IRModule for latency estimation. Finish units tests for PartitionRule and CandidatePartition.
mbs-octoml
force-pushed
the
mbs-collage-combine-rules
branch
from
97f062b
to
d3b608e
Compare
masahi
pushed a commit
to masahi/tvm
that referenced
this pull request
Jul 15, 2022
…2078) * [Collage] CombinerRule and CandidatePartition::EstimateCost See https://github.com/apache/tvm-rfcs/blob/main/rfcs/0062-collage.md. We complete the PartitionRule sub-class hierarchy with the addition of CombinePartitionRule, which allows disjoint candidate partitions to be unioned based on simple rules. - By TOpPattern kind, eg a kOutElemwiseFusable and kBroadcast. - A tuple argument with injective fields. - The projection from an injective group (obviously of tuple type) - Combinations of the above. These let us mimic many common fusion strategies, including TVMs, so that the candidates explored during Collage search are as large as possible to expose possible fusion opportunities but no larger. Also completes CandidatePartition with the EstimateCost method, which is used during search to construct a stand-alone IRModule for latency estimation. Finish units tests for PartitionRule and CandidatePartition. * - fix relay.collage ffi prefix.
xinetzone
pushed a commit
to daobook/tvm
that referenced
this pull request
Nov 25, 2022
…2078) * [Collage] CombinerRule and CandidatePartition::EstimateCost See https://github.com/apache/tvm-rfcs/blob/main/rfcs/0062-collage.md. We complete the PartitionRule sub-class hierarchy with the addition of CombinePartitionRule, which allows disjoint candidate partitions to be unioned based on simple rules. - By TOpPattern kind, eg a kOutElemwiseFusable and kBroadcast. - A tuple argument with injective fields. - The projection from an injective group (obviously of tuple type) - Combinations of the above. These let us mimic many common fusion strategies, including TVMs, so that the candidates explored during Collage search are as large as possible to expose possible fusion opportunities but no larger. Also completes CandidatePartition with the EstimateCost method, which is used during search to construct a stand-alone IRModule for latency estimation. Finish units tests for PartitionRule and CandidatePartition. * - fix relay.collage ffi prefix.
mikeseven
pushed a commit
to mikeseven/tvm
that referenced
this pull request
Sep 27, 2023
…2078) * [Collage] CombinerRule and CandidatePartition::EstimateCost See https://github.com/apache/tvm-rfcs/blob/main/rfcs/0062-collage.md. We complete the PartitionRule sub-class hierarchy with the addition of CombinePartitionRule, which allows disjoint candidate partitions to be unioned based on simple rules. - By TOpPattern kind, eg a kOutElemwiseFusable and kBroadcast. - A tuple argument with injective fields. - The projection from an injective group (obviously of tuple type) - Combinations of the above. These let us mimic many common fusion strategies, including TVMs, so that the candidates explored during Collage search are as large as possible to expose possible fusion opportunities but no larger. Also completes CandidatePartition with the EstimateCost method, which is used during search to construct a stand-alone IRModule for latency estimation. Finish units tests for PartitionRule and CandidatePartition. * - fix relay.collage ffi prefix.
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.
See https://github.com/apache/tvm-rfcs/blob/main/rfcs/0062-collage.md.
We complete the PartitionRule sub-class hierarchy with the addition of
CombinePartitionRule, which allows disjoint candidate partitions to be
unioned based on simple rules.
These let us mimic many common fusion strategies, including TVMs, so that
the candidates explored during Collage search are as large as possible to
expose possible fusion opportunities but no larger.
Also completes CandidatePartition with the EstimateCost method, which is
used during search to construct a stand-alone IRModule for latency estimation.
Thank you @mbaret for authoring the MockCostEstimator on which those
tests depend.
Finish units tests for PartitionRule and CandidatePartition.