This legacy 2017 model has been failing with eager_fail_to_run on 100%
of commits since mid-2025, providing zero CI signal while consuming
~5.3M GPU-seconds/week across 7+ benchmark jobs on CUDA, CPU, and ROCm.

Skip it in torchbench.yaml and remove its entries from all 31 expected
accuracy CSV files.


ghstack-source-id: 14abaec
Pull-Request: #175066

)

## Summary

Move CUDA 12.8 GPU tests from per-commit trunk CI to periodic (~3x/day on weekdays).

Both CUDA 12.8 and 13.0 are shipping wheel targets (nightly ships cu126, cu128, cu129, cu130), but their trunk CI test suites have **85-90% failure correlation** -- they almost always fail together. Over a 30-day analysis window covering 97 reverts and 38 significant regression events, **CUDA 12.8 never uniquely caught a regression that 13.0 missed**.

CUDA 13.0 is kept per-commit because:
- It is the **newest** shipping CUDA version
- Most likely to surface **novel breakage** from new CUDA runtime behavior
- Forward-looking CI should protect what's coming, not what's already stable

CUDA 12.8 is moved to periodic because:
- It is **mature and well-understood** -- breakage is less likely and less urgent
- The rare 12.8-only regression can tolerate the ~8-hour periodic detection window
- The 12.8 build job **remains in trunk** because `cross-compile-linux-test` depends on its artifacts

**Estimated savings: ~1,270 GPU-hours/week (~5,080 GPU-hours/month)**

This is the #2 savings opportunity from a broader CI workflow analysis (P2188981399) covering 128 PR+trunk jobs over 30 days. Combined with #175066 (CycleGAN skip, ~310 GPU-hours/week), total savings from this stack: **~1,580 GPU-hours/week (~6,320 GPU-hours/month)**.

### Changes
- `trunk.yml`: remove CUDA 12.8 test job (5 default + 3 distributed + 1 pr_time_benchmarks + 1 libtorch shards) and no-ops build
- `periodic.yml`: add default (5 GPU shards on g6.4xlarge) and distributed (3 multi-GPU shards on g4dn.12xlarge) to existing CUDA 12.8 periodic entry

## Test Plan

- CUDA 12.8 GPU tests continue to run in periodic (3x/day weekdays)
- CUDA 13.0 per-commit coverage is unchanged
- Cross-compile-linux-test continues to work (12.8 build job kept)

Pull Request resolved: #175067
Approved by: https://github.com/malfet
ghstack dependencies: #175066

Summary:
## Summary

Skip the `pytorch_CycleGAN_and_pix2pix` benchmark model from the inductor benchmark suite.

This legacy 2017 model has been failing with `eager_fail_to_run` on 100%
of commits since mid-2025, providing zero CI signal while consuming
~5.3M GPU-seconds/week across 7+ benchmark jobs on CUDA, CPU, and ROCm.

**Estimated savings: ~310 GPU-hours/week (~1,240 GPU-hours/month)**

Skip it in `torchbench.yaml` and remove its entries from all 31 expected
accuracy CSV files. Also remove it from the `higher_fp16` tolerance list.

See P2188981399 for the full CI workflow analysis.

## Test Plan

- CI should pass with CycleGAN skipped (it was already failing 100% of the time)
- No other benchmark models affected

X-link: pytorch/pytorch#175066
Approved by: https://github.com/huydhn, https://github.com/malfet

Reviewed By: atalman

Differential Revision: D93507913

fbshipit-source-id: b9e0b750b38bd2a9afd72eddede007a1eedf0c09

#175066)

## Summary

Skip the `pytorch_CycleGAN_and_pix2pix` benchmark model from the inductor benchmark suite.

This legacy 2017 model has been failing with `eager_fail_to_run` on 100%
of commits since mid-2025, providing zero CI signal while consuming
~5.3M GPU-seconds/week across 7+ benchmark jobs on CUDA, CPU, and ROCm.

**Estimated savings: ~310 GPU-hours/week (~1,240 GPU-hours/month)**

Skip it in `torchbench.yaml` and remove its entries from all 31 expected
accuracy CSV files. Also remove it from the `higher_fp16` tolerance list.

See P2188981399 for the full CI workflow analysis.

## Test Plan

- CI should pass with CycleGAN skipped (it was already failing 100% of the time)
- No other benchmark models affected

Pull Request resolved: #175066
Approved by: https://github.com/huydhn, https://github.com/malfet

(cherry picked from commit 688c943)

)

## Summary

Move CUDA 12.8 GPU tests from per-commit trunk CI to periodic (~3x/day on weekdays).

Both CUDA 12.8 and 13.0 are shipping wheel targets (nightly ships cu126, cu128, cu129, cu130), but their trunk CI test suites have **85-90% failure correlation** -- they almost always fail together. Over a 30-day analysis window covering 97 reverts and 38 significant regression events, **CUDA 12.8 never uniquely caught a regression that 13.0 missed**.

CUDA 13.0 is kept per-commit because:
- It is the **newest** shipping CUDA version
- Most likely to surface **novel breakage** from new CUDA runtime behavior
- Forward-looking CI should protect what's coming, not what's already stable

CUDA 12.8 is moved to periodic because:
- It is **mature and well-understood** -- breakage is less likely and less urgent
- The rare 12.8-only regression can tolerate the ~8-hour periodic detection window
- The 12.8 build job **remains in trunk** because `cross-compile-linux-test` depends on its artifacts

**Estimated savings: ~1,270 GPU-hours/week (~5,080 GPU-hours/month)**

This is the #2 savings opportunity from a broader CI workflow analysis (P2188981399) covering 128 PR+trunk jobs over 30 days. Combined with #175066 (CycleGAN skip, ~310 GPU-hours/week), total savings from this stack: **~1,580 GPU-hours/week (~6,320 GPU-hours/month)**.

### Changes
- `trunk.yml`: remove CUDA 12.8 test job (5 default + 3 distributed + 1 pr_time_benchmarks + 1 libtorch shards) and no-ops build
- `periodic.yml`: add default (5 GPU shards on g6.4xlarge) and distributed (3 multi-GPU shards on g4dn.12xlarge) to existing CUDA 12.8 periodic entry

## Test Plan

- CUDA 12.8 GPU tests continue to run in periodic (3x/day weekdays)
- CUDA 13.0 per-commit coverage is unchanged
- Cross-compile-linux-test continues to work (12.8 build job kept)

Pull Request resolved: #175067
Approved by: https://github.com/malfet
ghstack dependencies: #175066

(cherry picked from commit ef0353f)

#175299)

[benchmark] Skip pytorch_CycleGAN_and_pix2pix from inductor benchmarks (#175066)

## Summary

Skip the `pytorch_CycleGAN_and_pix2pix` benchmark model from the inductor benchmark suite.

This legacy 2017 model has been failing with `eager_fail_to_run` on 100%
of commits since mid-2025, providing zero CI signal while consuming
~5.3M GPU-seconds/week across 7+ benchmark jobs on CUDA, CPU, and ROCm.

**Estimated savings: ~310 GPU-hours/week (~1,240 GPU-hours/month)**

Skip it in `torchbench.yaml` and remove its entries from all 31 expected
accuracy CSV files. Also remove it from the `higher_fp16` tolerance list.

See P2188981399 for the full CI workflow analysis.

## Test Plan

- CI should pass with CycleGAN skipped (it was already failing 100% of the time)
- No other benchmark models affected

Pull Request resolved: #175066
Approved by: https://github.com/huydhn, https://github.com/malfet

(cherry picked from commit 688c943)

Co-authored-by: Eli Uriegas <eliuriegas@meta.com>

)

[CI] Move CUDA 12.8 GPU tests from per-commit trunk to periodic (#175067)

## Summary

Move CUDA 12.8 GPU tests from per-commit trunk CI to periodic (~3x/day on weekdays).

Both CUDA 12.8 and 13.0 are shipping wheel targets (nightly ships cu126, cu128, cu129, cu130), but their trunk CI test suites have **85-90% failure correlation** -- they almost always fail together. Over a 30-day analysis window covering 97 reverts and 38 significant regression events, **CUDA 12.8 never uniquely caught a regression that 13.0 missed**.

CUDA 13.0 is kept per-commit because:
- It is the **newest** shipping CUDA version
- Most likely to surface **novel breakage** from new CUDA runtime behavior
- Forward-looking CI should protect what's coming, not what's already stable

CUDA 12.8 is moved to periodic because:
- It is **mature and well-understood** -- breakage is less likely and less urgent
- The rare 12.8-only regression can tolerate the ~8-hour periodic detection window
- The 12.8 build job **remains in trunk** because `cross-compile-linux-test` depends on its artifacts

**Estimated savings: ~1,270 GPU-hours/week (~5,080 GPU-hours/month)**

This is the #2 savings opportunity from a broader CI workflow analysis (P2188981399) covering 128 PR+trunk jobs over 30 days. Combined with #175066 (CycleGAN skip, ~310 GPU-hours/week), total savings from this stack: **~1,580 GPU-hours/week (~6,320 GPU-hours/month)**.

### Changes
- `trunk.yml`: remove CUDA 12.8 test job (5 default + 3 distributed + 1 pr_time_benchmarks + 1 libtorch shards) and no-ops build
- `periodic.yml`: add default (5 GPU shards on g6.4xlarge) and distributed (3 multi-GPU shards on g4dn.12xlarge) to existing CUDA 12.8 periodic entry

## Test Plan

- CUDA 12.8 GPU tests continue to run in periodic (3x/day weekdays)
- CUDA 13.0 per-commit coverage is unchanged
- Cross-compile-linux-test continues to work (12.8 build job kept)

Pull Request resolved: #175067
Approved by: https://github.com/malfet
ghstack dependencies: #175066

(cherry picked from commit ef0353f)

Co-authored-by: Eli Uriegas <eliuriegas@meta.com>

- Re-enable `detectron2_maskrcnn` skip in skip.all.
- Re-enable all `timm_*` model skips in skip.all.
- Keep explicit upstream PR context comments for `modded_nanogpt`
  and `pytorch_CycleGAN_and_pix2pix`.
- Remove stale expected-accuracy rows for skipped models.

Relevant PRs:
[1] pytorch/pytorch#120299
[2] pytorch/pytorch#164816
[3] pytorch/pytorch#172125
[4] pytorch/pytorch#175066
[5] #2306

This legacy 2017 model has been failing with eager_fail_to_run on 100%
of commits since mid-2025, providing zero CI signal while consuming
~5.3M GPU-seconds/week across 7+ benchmark jobs on CUDA, CPU, and ROCm.

Skip it in torchbench.yaml and remove its entries from all 31 expected
accuracy CSV files.

ghstack-source-id: b745ae2
Pull-Request: pytorch/pytorch#175066