[Misc] Add conftest plugin for applying forking decorator

[kevin314]

Continuation of #7053--Adds a fixture to wrap test functions with the fork_new_process_for_each_test function through a pytest command-line option (--streamed-fork).

FIX #7053

---

PR Checklist (Click to Expand)

Thank you for your contribution to vLLM! Before submitting the pull request, please ensure the PR meets the following criteria. This helps vLLM maintain the code quality and improve the efficiency of the review process.

PR Title and Classification

Only specific types of PRs will be reviewed. The PR title is prefixed appropriately to indicate the type of change. Please use one of the following:

• [Bugfix] for bug fixes.
• [CI/Build] for build or continuous integration improvements.
• [Doc] for documentation fixes and improvements.
• [Model] for adding a new model or improving an existing model. Model name should appear in the title.
• [Frontend] For changes on the vLLM frontend (e.g., OpenAI API server, LLM class, etc.)
• [Kernel] for changes affecting CUDA kernels or other compute kernels.
• [Core] for changes in the core vLLM logic (e.g., LLMEngine, AsyncLLMEngine, Scheduler, etc.)
• [Hardware][Vendor] for hardware-specific changes. Vendor name should appear in the prefix (e.g., [Hardware][AMD]).
• [Misc] for PRs that do not fit the above categories. Please use this sparingly.

Note: If the PR spans more than one category, please include all relevant prefixes.

Code Quality

The PR need to meet the following code quality standards:

• We adhere to Google Python style guide and Google C++ style guide.
• Pass all linter checks. Please use format.sh to format your code.
• The code need to be well-documented to ensure future contributors can easily understand the code.
• Include sufficient tests to ensure the project to stay correct and robust. This includes both unit tests and integration tests.
• Please add documentation to docs/source/ if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.

Adding or changing kernels

Each custom kernel needs a schema and one or more implementations to be registered with PyTorch.

• Make sure custom ops are registered following PyTorch guidelines: Custom C++ and CUDA Operators and The Custom Operators Manual
• Custom operations that return Tensors require meta-functions. Meta-functions should be implemented and registered in python so that dynamic dims can be handled automatically. See above documents for a description of meta-functions.
• Use torch.libary.opcheck() to test the function registration and meta-function for any registered ops. See tests/kernels for examples.
• When changing the C++ signature of an existing op, the schema must be updated to reflect the changes.
• If a new custom type is needed, see the following document: Custom Class Support in PT2.

Notes for Large Changes

Please keep the changes as concise as possible. For major architectural changes (>500 LOC excluding kernel/data/config/test), we would expect a GitHub issue (RFC) discussing the technical design and justification. Otherwise, we will tag it with rfc-required and might not go through the PR.

What to Expect for the Reviews

The goal of the vLLM team is to be a transparent reviewing machine. We would like to make the review process transparent and efficient and make sure no contributor feel confused or frustrated. However, the vLLM team is small, so we need to prioritize some PRs over others. Here is what you can expect from the review process:

• After the PR is submitted, the PR will be assigned to a reviewer. Every reviewer will pick up the PRs based on their expertise and availability.
• After the PR is assigned, the reviewer will provide status update every 2-3 days. If the PR is not reviewed within 7 days, please feel free to ping the reviewer or the vLLM team.
• After the review, the reviewer will put an action-required label on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR.
• Please respond to all comments within a reasonable time frame. If a comment isn't clear or you disagree with a suggestion, feel free to ask for clarification or discuss the suggestion.

Thank You

Finally, thank you for taking the time to read these guidelines and for your interest in contributing to vLLM. Your contributions make vLLM a great tool for everyone!

[github-actions[bot]]

👋 Hi! Thank you for contributing to the vLLM project. Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[kevin314]

Quick benchmark with test_basic_correctness.py seems to show the forking adds some overhead to running tests: without forking => ~170s with forking => ~300s (both with and without capturing output with -s) this is with 36 vcpu on runpod

[youkaichao]

do you run the one gpu test or the 2 gpu test?

[DarkLight1337]

To better debug CUDA reinitialization errors, I wonder whether we can use unittest.mock.patch to instrument torch.cuda._lazy_init to print out the stack trace, so we can see which code is responsible?

[youkaichao]

To better debug CUDA reinitialization errors, I wonder whether we can use unittest.mock.patch to instrument torch.cuda._lazy_init to print out the stack trace, so we can see which code is responsible?

it will print the stack trace for every test, even if the test does not trigger CUDA reinitialization errors.

[DarkLight1337]

To better debug CUDA reinitialization errors, I wonder whether we can use unittest.mock.patch to instrument torch.cuda._lazy_init to print out the stack trace, so we can see which code is responsible?

it will print the stack trace for every test, even if the test does not trigger CUDA reinitialization errors.

Perhaps we can collect the stack trace and only print it out when the test fails?

[youkaichao]

To better debug CUDA reinitialization errors, I wonder whether we can use unittest.mock.patch to instrument torch.cuda._lazy_init to print out the stack trace, so we can see which code is responsible?

it will print the stack trace for every test, even if the test does not trigger CUDA reinitialization errors.

Perhaps we can collect the stack trace and only print it out when the test fails?

that would be tooo complicated

[kevin314]

Excluding multi gpu tests for a total of 14 tests in test_basic_correctness.py

Running mprof run --include-children pytest -s ./tests/basic_correctness/test_basic_correctness.py https://pypi.org/project/memory-profiler/ (Running without profiler has similar execution times)

Test Scenario	Execution Time (s)	Peak Memory Usage (MB)
No forking (Output captured - without -s)	144.84	19815
No forking (Output not captured - with -s)	138.64, 139.63, 139.48	19820, 19827, 19745
With forking (Output captured - without -s)	216.45	11178
With forking (Output not captured - with -s)	211.82, 223.52, 214.65	11211, 11224, 11179

[youkaichao]

@kevin314 can you also test the time spent by using https://pypi.org/project/pytest-forked/ , and see if the increased test time is caused by fork itself or the fork_new_process_for_each_test implementation?

[kevin314]

@kevin314 can you also test the time spent by using https://pypi.org/project/pytest-forked/ , and see if the increased test time is caused by fork itself or the fork_new_process_for_each_test implementation?

mprof run --include-children pytest --forked ./tests/basic_correctness/test_basic_correctness.py

Test Scenario	Execution Time (s)	Memory Usage (MB)
Forking with pytest-forked	216.25, 221.13, 220.77	11055, 11175, 11021

Looks like it may just be the forking itself

[youkaichao]

thanks for the experiments! then I think it does not make sense to use fork by default, especially for the thousands of small experiments in kernels tests, etc.

to proceed, I think we can remove all manual decorators for fork_new_process_for_each_test , and use --streamed-fork in some of the test commands in https://github.com/vllm-project/vllm/blob/main/.buildkite/test-pipeline.yaml

[DarkLight1337]

Let's try to run the tests with this change.

[kevin314]

This is probably ready for a review--I think the CI is failing due to dependencies on changes in the pipeline file that are part of this PR, if anyone has ideas on how to deal with that

[github-actions[bot]]

This pull request has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this pull request should remain open. Thank you!

[mergify[bot]]

This pull request has merge conflicts that must be resolved before it can be merged. Please rebase the PR, @kevin314.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[github-actions[bot]]

This pull request has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this pull request should remain open. Thank you!

[github-actions[bot]]

This pull request has been automatically closed due to inactivity. Please feel free to reopen if you intend to continue working on it. Thank you!