V3 Performance Signal Detected by TorchBench Userbenchmark "torch-nightly" on 'fd56aa0daeec624bcff822fef04fe9339770dc33'

[github-actions[bot]]

TorchBench CI has detected a performance signal or runtime regression.

Base PyTorch commit: d2c3aa1fded6d1a0165eafe6e08816a85a1190d8

Affected PyTorch commit: fd56aa0daeec624bcff822fef04fe9339770dc33

Affected Tests:

• test_eval[DALLE2_pytorch-cuda-eager]_latency: +12.54825%
• test_train[dcgan-cuda-eager]_latency: +32.01213%
• test_train[detectron2_fasterrcnn_r_101_fpn-cuda-eager]_latency: +15.62042%
• test_eval[detectron2_fasterrcnn_r_50_dc5-cuda-eager]_latency: +13.46880%
• test_train[detectron2_fasterrcnn_r_50_fpn-cuda-eager]_latency: +12.28584%
• test_eval[hf_Bert-cuda-eager]_latency: +10.44053%
• test_train[lennard_jones-cuda-eager]_latency: +29.34321%
• test_eval[maml-cuda-eager]_latency: +24.93444%
• test_train[maml_omniglot-cuda-eager]_latency: +19.75669%
• test_train[mobilenet_v3_large-cuda-eager]_latency: +15.09031%
• test_eval[mobilenet_v3_large-cuda-eager]_latency: +10.93895%
• test_train[opacus_cifar10-cuda-eager]_latency: +15.84414%
• test_eval[opacus_cifar10-cuda-eager]_latency: +15.64927%
• test_train[phlippe_densenet-cuda-eager]_latency: +14.62465%
• test_eval[phlippe_densenet-cuda-eager]_latency: +15.91092%
• test_train[phlippe_resnet-cuda-eager]_latency: +19.91591%
• test_eval[phlippe_resnet-cuda-eager]_latency: +19.88433%
• test_eval[pyhpc_turbulent_kinetic_energy-cuda-eager]_latency: +12.15181%
• test_train[pytorch_CycleGAN_and_pix2pix-cuda-eager]_latency: +20.07919%
• test_eval[pytorch_CycleGAN_and_pix2pix-cuda-eager]_latency: +14.88968%
• test_train[resnet18-cuda-eager]_latency: +21.09336%
• test_eval[shufflenet_v2_x1_0-cuda-eager]_latency: +10.72718%
• test_train[speech_transformer-cuda-eager]_latency: +10.47227%
• test_train[timm_efficientnet-cuda-eager]_latency: +10.42307%
• test_eval[timm_regnet-cuda-eager]_latency: +11.28450%
• test_train[tts_angular-cuda-eager]_latency: +14.82836%
• test_eval[vision_maskrcnn-cuda-eager]_latency: +10.08274%
• test_eval[yolov3-cuda-eager]_latency: +17.42123%

Tests that were no longer run on affected commit:

Tests that were newly added on affected commit:

Runtime regressions found? No runtime errors were found in the new benchmarks run--you are all good there!

GitHub workflow that triggered this issue: https://github.com/pytorch/benchmark/actions/runs/6747563114

cc @xuzhao9

[xuzhao9]

Bisection: https://github.com/pytorch/benchmark/actions/runs/6748552362

[xuzhao9]

  "target_repo": "pytorch",
  "start": "59869903b3a54acb3ac3a93470fc595d1de3fa7a",
  "end": "4a17693d193a8b23109ac3129e5cde92babfe191",
  "result": []
}