cuda partioner supported #14477
Merged
cuda partioner supported #14477
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
Gasoonjia
requested review from
JacobSzwejbka and
larryliu0820
as code owners
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/14477
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 3 Unrelated FailuresAs of commit e8a30b0 with merge base 7b33035 ( NEW FAILURE - The following job has failed:
BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
|
@Gasoonjia has exported this pull request. If you are a Meta employee, you can view the originating diff in D82987193. |
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 22, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Differential Revision: D82987193
This PR needs a
|
larryliu0820
approved these changes
Sep 22, 2025
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 22, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Differential Revision: D82987193
Gasoonjia
force-pushed
the
export-D82987193
branch
from
7154f1f to
c32854e
Compare
|
@Gasoonjia has exported this pull request. If you are a Meta employee, you can view the originating diff in D82987193. |
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 22, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Reviewed By: larryliu0820 Differential Revision: D82987193
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 22, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Reviewed By: larryliu0820 Differential Revision: D82987193
Gasoonjia
force-pushed
the
export-D82987193
branch
from
c32854e to
56480a6
Compare
|
@Gasoonjia has exported this pull request. If you are a Meta employee, you can view the originating diff in D82987193. |
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 22, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Reviewed By: larryliu0820 Differential Revision: D82987193
|
|
||
|
|
||
| @final | ||
| class CudaPartitioner(Partitioner): |
There was a problem hiding this comment.
Mark as experimental
| do_not_decompose = set() | ||
|
|
||
| for node in ep.graph.nodes: | ||
| if node.op == "call_function" and isinstance( |
There was a problem hiding this comment.
can aoti eat control flow hops?
There was a problem hiding this comment.
didn't test that. Let me add a test later.
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 23, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Reviewed By: larryliu0820 Differential Revision: D82987193
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 23, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Reviewed By: larryliu0820 Differential Revision: D82987193
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 23, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Reviewed By: larryliu0820 Differential Revision: D82987193
Gasoonjia
force-pushed
the
export-D82987193
branch
from
56480a6 to
9945631
Compare
|
@Gasoonjia has exported this pull request. If you are a Meta employee, you can view the originating diff in D82987193. |
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 23, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Reviewed By: larryliu0820 Differential Revision: D82987193
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Reviewed By: larryliu0820 Differential Revision: D82987193
Gasoonjia
force-pushed
the
export-D82987193
branch
from
9945631 to
e8a30b0
Compare
|
@Gasoonjia has exported this pull request. If you are a Meta employee, you can view the originating diff in D82987193. |
Gasoonjia
added a commit
to Gasoonjia/executorch-1
that referenced
this pull request
Sep 23, 2025
Summary: This diff introduce partitioner for cuda delegated, which driven by aoti library. The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table. Reviewed By: larryliu0820 Differential Revision: D82987193
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.
Summary:
This diff introduce partitioner for cuda delegated, which driven by aoti library.
The partitioner will partition the input into exactly one partitioned graph that contains all operators from the input graph, also will keep all operators (except the ops that couldn't handled by aoti-cuda lib) away from executorch operator decomposition. Operator will be decomposed in the cuda backend using aoti-cuda specific decomposition table.
Differential Revision: D82987193