Add conv ops to operator microbenchmark#166331
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/166331
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: ❌ 2 New Failures, 2 Unrelated FailuresAs of commit c7dd98e with merge base a076b4d ( NEW FAILURES - The following jobs have failed:
FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
jainapurva
force-pushed
the
add_conv3d_ops
branch
from
77c0019 to
4f2677d
Compare
| return [config for config in config_list if cuda_config not in config] | ||
|
|
||
|
|
||
| def remove_cpu(config_list): |
There was a problem hiding this comment.
Just curious: Why do you need to remove CPU device on those conv ops? Because they are all on backward path?
There was a problem hiding this comment.
Yes, as backward tests are not needed for cpu benchmarking, I'm removing cpu tests from the final list
|
Darn, it looks like we need to refresh CPU op benchmark baseline again https://github.com/pytorch/pytorch/actions/workflows/operator_benchmark.yml?query=branch%3Amain |
| conv_3d_configs_long = op_bench.cross_product_configs( | ||
| IC=[16, 32], | ||
| OC=[32, 64], | ||
| kernel=[3, 5], | ||
| stride=[1, 2], | ||
| N=[1], | ||
| D=[128], | ||
| H=[128], | ||
| W=[128], | ||
| device=["cpu", "cuda"], | ||
| tags=["long"], | ||
| ) |
There was a problem hiding this comment.
out of curiosity, does this cover the shapes relevant to the recent regression?
There was a problem hiding this comment.
Yes, the shapes of regression are covered, but the enablement of AMP is not covered
|
@pytorchmergebot merge -i |
Merge startedYour change will be merged while ignoring the following 2 checks: operator_benchmark / aarch64-opbenchmark-test / test (cpu_operator_benchmark_short, 1, 1, linux.arm64.m8g.4xlarge), operator_benchmark / x86-opbenchmark-test / test (cpu_operator_benchmark_short, 1, 1, linux.12xlarge) Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Adding `conv` (conv1d, conv2d, conv3d) to the list of operator microbenchmarks run in the CI script (`.ci/pytorch/test.sh`), ensuring convolution operators are now benchmarked alongside existing ones. Pull Request resolved: #166331 Approved by: https://github.com/huydhn, https://github.com/jbschlosser
4 participants
Adding
conv(conv1d, conv2d, conv3d) to the list of operator microbenchmarks run in the CI script (.ci/pytorch/test.sh), ensuring convolution operators are now benchmarked alongside existing ones.