🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138922

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

❌ 30 New Failures, 1 Unrelated Failure

As of commit 794f6a4 with merge base c3fbec7 ():

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@pytorchbot rebase -b main

@pytorchbot started a rebase job onto refs/remotes/origin/main. Check the current status here

Successfully rebased gh/bobrenjc93/90/orig onto refs/remotes/origin/main, please pull locally before adding more changes (for example, via ghstack checkout https://github.com/pytorch/pytorch/pull/138922)

@bobrenjc93 I believe this might be causing perf regression on torchbench: https://hud.pytorch.org/benchmark/torchbench/inductor_no_cudagraphs?dashboard=torchinductor&startTime=Mon,%2018%20Nov%202024%2022:00:43%20GMT&stopTime=Mon,%2025%20Nov%202024%2022:00:43%20GMT&granularity=hour&mode=training&dtype=amp&deviceName=cuda%20(a100)&lBranch=gh/bobrenjc93/90/head&lCommit=e67d26992e8a473aec6ede13b8cb928208c3df34&rBranch=main&rCommit=1bdb92cbff6f8f6fd1842e8fd9969bca688eaea3

I also bisected on torchbench llama model (on A100 devvm):
python benchmarks/dynamo/torchbench.py --performance --cold-start-latency --inductor --disable-cudagraphs --device cuda --inference --bfloat16 --print-memory --output output.csv --only llama --batch-size 32
5268754 (11/22 nightly): 2.177902x
995e307: 2.196x
c513f01: 2.198x
11c786d: 2.160x
ba5253d: 1.748x <- this PR
51b6126 (11/23 nightly): 1.781900x

Discussed in Inductor group chat and the group agrees to revert this PR

@pytorchbot revert -m "perf regression on torchbench"

❌ 🤖 pytorchbot command failed:

@pytorchbot revert: error: the following arguments are required: -c/--classification

usage: @pytorchbot revert -m MESSAGE -c
                          {nosignal,ignoredsignal,landrace,weird,ghfirst}

Try @pytorchbot --help for more info.

@pytorchbot revert -m "perf regression on torchbench" -c nosignal

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

@bobrenjc93 your PR has been successfully reverted.

new benchmark runs as of 11/25

base: https://github.com/pytorch/pytorch/actions/runs/12023899968
head: https://github.com/pytorch/pytorch/actions/runs/12023903827

Endorsing the NN module spec changes, although we will have to carefully check if this actually fixes all the benchmarks