Repository navigation
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/4529
Note: Links to docs will display an error until the docs builds have been completed. This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This was referenced Oct 8, 2026
meta-cla
Bot
added
the
CLA Signed
github-actions
Bot
added
Refactoring
Contributor
Author
huydhn
added a commit
that referenced
this pull request
Oct 8, 2026
Two ways shard 3 failed on OSDC that have nothing to do with the tests. The job hung to its 120-minute timeout after the script had already exited 0. run_with_env_secrets.py drains the step's stdout until EOF, and EOF only arrives once every process holding the write end has closed it. A ps dump named the holder: Xvfb, started by the test setup and outliving pytest. Under linux_job_v2 the surrounding `docker run` reaped it for us. So dump the surviving processes and reap them before exiting. -n auto gives 16 workers on the GPU pods, and 16 concurrent CUDA contexts on one A10G leave no room: a test that spawns its own CUDA subprocess dies in cuDevicePrimaryCtxRetain with CUDA_ERROR_OUT_OF_MEMORY before allocating anything, and ordinary Triton tests start failing to allocate too. Which tests lose the race varies run to run, so cap GPU concurrency at 4 rather than chase them. Shard 3 goes from ~650s to ~1130s, well inside the timeout. The CPU path keeps its measured 24. test_dreamer_v3_checkpoint_resume_processes is also deselected from the xdist runs and added to the serial shard. The existing quarantine is path-based and moving all of test_dreamer_v3.py would cost ~1300s of the shard's ~2950s, so quarantine just the one test. Authored with Claude Code. ghstack-source-id: 5bf4013 Pull-Request: #4529
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
Two ways shard 3 failed on OSDC that have nothing to do with the tests.
The job hung to its 120-minute timeout after the script had already exited
0. run_with_env_secrets.py drains the step's stdout until EOF, and EOF only
arrives once every process holding the write end has closed it. A ps dump
named the holder: Xvfb, started by the test setup and outliving pytest.
Under linux_job_v2 the surrounding
docker runreaped it for us. So dumpthe surviving processes and reap them before exiting.
-n auto gives 16 workers on the GPU pods, and 16 concurrent CUDA contexts on
one A10G leave no room: a test that spawns its own CUDA subprocess dies in
cuDevicePrimaryCtxRetain with CUDA_ERROR_OUT_OF_MEMORY before allocating
anything, and ordinary Triton tests start failing to allocate too. Which
tests lose the race varies run to run, so cap GPU concurrency at 4 rather
than chase them. Shard 3 goes from ~650s to ~1130s, well inside the timeout.
The CPU path keeps its measured 24.
test_dreamer_v3_checkpoint_resume_processes is also deselected from the
xdist runs and added to the serial shard. The existing quarantine is
path-based and moving all of test_dreamer_v3.py would cost ~1300s of the
shard's ~2950s, so quarantine just the one test.
Authored with Claude Code.