[Relax] Initial setup of relax backend pipeline #17626
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
Member
|
cc @Hzfengsy |
This PR sets up the initial relax backend pipeline, as per thread in https://discuss.tvm.apache.org/t/relax-target-default-pipeline. Particularly, we organized default list of passes in the default pipeline into four stages: * library dispatch passes, * legalize passes, * dataflow lower passes, * finalize passes. The expectation is to have the default pipeline work out of the box, while users can still choose to customize the pipeline by importing these lists and combining them together, with customized passes inserted at proper locations. In this PR, we set up the initial pipelines for CUDA and LLVM backends. The pipelines for both backends will be enhanced and completed in the near future, and meanwhile the default pipeline for more backends (e.g., generic GPUs) will be added. The existing pipeline registry system in python/tvm/relax/pipeline.py will be gradually deprecated after the establishment of the new relax backend pipeline.
MasterJH5574
force-pushed
the
tvm-dev/2025-02-05-init-relax-pipeline
branch
from
32680de to
f66dd7c
Compare
tqchen
approved these changes
Feb 6, 2025
Member
|
@tvm-bot rerun |
1 similar comment
Member
|
@tvm-bot rerun |
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Outdated suggestions cannot be applied.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.
This PR sets up the initial relax backend pipeline, as per thread in https://discuss.tvm.apache.org/t/relax-target-default-pipeline.
Particularly, we organized default list of passes in the default pipeline into four stages:
The expectation is to have the default pipeline work out of the box, while users can still choose to customize the pipeline by importing these lists and combining them together, with customized passes inserted at proper locations.
In this PR, we set up the initial pipelines for CUDA and LLVM backends. The pipelines for both backends will be enhanced and completed in the near future, and meanwhile the default pipeline for more backends (e.g., generic GPUs) will be added.
The existing pipeline registry system in python/tvm/relax/pipeline.py will be gradually deprecated after the establishment of the new relax backend pipeline.