Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18496
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ❌ 4 New Failures, 1 Unrelated FailureAs of commit 57003f6 with merge base 9b91b43 ( NEW FAILURES - The following jobs have failed:
BROKEN TRUNK - The following job 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
Contributor
This PR needs a
|
meta-codesync
Bot
changed the title
apullin
force-pushed
the
export-D96489903
branch
from
March 25, 2026 22:29
931a377 to
b1a74f8
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Mar 25, 2026
…orch#18496) Summary: Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
meta-codesync
Bot
changed the title
apullin
force-pushed
the
export-D96489903
branch
from
March 26, 2026 15:07
b1a74f8 to
4f1a818
Compare
meta-codesync
Bot
changed the title
apullin
force-pushed
the
export-D96489903
branch
from
March 30, 2026 16:07
4f1a818 to
93d0a70
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Mar 30, 2026
…orch#18496) Summary: Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Mar 30, 2026
…orch#18496) Summary: Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
2 times, most recently
from
March 30, 2026 21:27
559036a to
77d036d
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Mar 30, 2026
…orch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
changed the title
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Apr 13, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
from
April 13, 2026 19:30
77d036d to
00408ce
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Apr 17, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
from
April 17, 2026 16:03
00408ce to
a9c9746
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Apr 20, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
from
April 20, 2026 18:32
a9c9746 to
e10f77c
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
May 12, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Jun 4, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Jun 16, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
from
June 16, 2026 17:08
391f321 to
3718908
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Jun 16, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
from
June 16, 2026 17:15
3718908 to
8af8c68
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Jun 17, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
from
June 17, 2026 00:00
8af8c68 to
25e74e8
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Jun 17, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
from
June 17, 2026 00:05
25e74e8 to
3a31c14
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Jun 19, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
from
June 19, 2026 04:42
3a31c14 to
6817eda
Compare
apullin
added a commit
to apullin/executorch
that referenced
this pull request
Jun 19, 2026
…hen no ops match decomp table (pytorch#18496) Summary: Pull Request resolved: pytorch#18496 Adds an early-exit check to _gen_edge_manager_for_partitioners: before calling program.run_decompositions(table), scan the graph for ops that appear in the decomposition table. If none are found, skip the call entirely. Each run_decompositions call performs a full re-export of the program via make_fx(), re-tracing every node through FakeTensor dispatch. On the EDGE_DO_NOT_DECOMP path this function is called up to 3 times; the early-exit eliminates at least one redundant call where the previous pass already decomposed all matching ops. The check recursively walks control flow submodules (cond/map/scan) to avoid incorrectly skipping when decomposable ops are nested. ## Benchmark Model: small CNN feature extractor (~50K params, 9 conv layers with LayerNorm, targeting Ethos-U55 via the ARM/TOSA lowering pipeline). Graph: ~1200 nodes. lower() before: 82 s lower() after: 71 s Delta: -11 s (-13 %) Differential Revision: D96489903
apullin
force-pushed
the
export-D96489903
branch
from
June 19, 2026 04:48
6817eda to
33061a2
Compare
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
meta-codesync
Bot
changed the title
apullin
force-pushed
the
export-D96489903
branch
from
September 15, 2026 20:02
33061a2 to
1bee7a3
Compare
apullin
requested review from
rascani and
zingo
and removed request for
lucylq
September 15, 2026 20:03
…orch#18496) Summary: Pull Request resolved: pytorch#18496 `_gen_edge_manager_for_partitioners` can call `program.run_decompositions(table)` up to three times. Each call re-exports the program through `make_fx`, retracing every node through FakeTensor dispatch, even when a previous pass already removed every operator covered by the next decomposition table. Before each nonempty-table replay, scan the root and every descendant `GraphModule`. Skip `run_decompositions` when no `call_function` target matches the table. Empty tables still run to preserve functionalization. Overload packets are checked against their constituent overloads, and nested-region graphs with their own decomposition policy conservatively force a replay. This keeps the optimization correct for `cond`, `map`, `scan`, `while_loop`, and `invoke_subgraph` bodies rather than limiting the scan to the previously enumerated control-flow operators. ## Benchmark Synthetic calibration lowering suite, five models: Comparison revision: 79.000 s / 79.516 s This change: 65.637 s / 65.512 s Delta: -17.3% mean / -17.6% warm CombinedControl Ethos-U55, structured `LOWERING.duration_ms`: Comparison revision: 145.801 s / 145.888 s This change: 133.949 s / 132.096 s Delta: -8.8% mean / -9.5% warm Differential Revision: D96489903
This branch had an error being deployed
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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:
_gen_edge_manager_for_partitionerscan callprogram.run_decompositions(table)up to three times. Each call re-exports the program throughmake_fx, retracing every node through FakeTensor dispatch, even when a previous pass already removed every operator covered by the next decomposition table.Before each nonempty-table replay, scan the root and every descendant
GraphModule. Skiprun_decompositionswhen nocall_functiontarget matches the table. Empty tables still run to preserve functionalization. Overload packets are checked against their constituent overloads, and nested-region graphs with their own decomposition policy conservatively force a replay.This keeps the optimization correct for
cond,map,scan,while_loop, andinvoke_subgraphbodies rather than limiting the scan to the previously enumerated control-flow operators.Benchmark
Synthetic calibration lowering suite, five models:
Comparison revision: 79.000 s / 79.516 s
This change: 65.637 s / 65.512 s
Delta: -17.3% mean / -17.6% warm
CombinedControl Ethos-U55, structured
LOWERING.duration_ms:Comparison revision: 145.801 s / 145.888 s
This change: 133.949 s / 132.096 s
Delta: -8.8% mean / -9.5% warm
Differential Revision: D96489903