🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/158082

• 📄 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.

❌ 1 New Failure

As of commit 844f5c3 with merge base ae86e8f ():

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

• ✅login: Electron4444 / (844f5c3)

@tianyu-l Could you review, please.

I like the speedup. but for me these tests don't run correctly. Did they run fine for you locally?

python test/distributed/tensor/test_tensor_ops.py -k slice for me gives

Mismatched elements: 64 / 64 (100.0%)
Greatest absolute difference: 3.619420289993286 at index (5, 0) (up to 1e-05 allowed)
Greatest relative difference: 72.54373168945312 at index (2, 1) (up to 1.3e-06 allowed)

I had run them locally and everything seemed fine, but it seems I had an error in setting up my environment.
My apologies.
But I think this is an issue of wrong association of values to the tests.
While retesting my PR I read through the last PR in the issue that got closed for inactivity.
In the logs of the performed tests jenkins output the values expected were switched for the corresponding results.

I tried to reproduce this by outputting all the tensors of failed tests to see a pattern, but I sadly couldn't reproduce it.

My guess is that there's an error via the timing during distributed calculation, but this clearly is not working.

Thanks anyways for taking the time.