Add vLLM v0.18.1 source tree with KV transfer abort fix
third_party/vllm/ now tracked in git for direct patch management.
Based on vLLM v0.18.1 release with one patch applied:
vllm/v1/core/sched/scheduler.py:
Replace fatal assert with graceful skip when KV transfer callback
arrives for an already-aborted request during PD disaggregated serving.
Future vLLM modifications should be made directly in third_party/vllm/
and committed normally. The patches/ directory is kept as documentation
of what changed from upstream.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
57
third_party/vllm/.buildkite/test_areas/pytorch.yaml
vendored
Normal file
57
third_party/vllm/.buildkite/test_areas/pytorch.yaml
vendored
Normal file
@@ -0,0 +1,57 @@
|
||||
group: PyTorch
|
||||
depends_on:
|
||||
- image-build
|
||||
steps:
|
||||
- label: PyTorch Compilation Unit Tests
|
||||
timeout_in_minutes: 10
|
||||
source_file_dependencies:
|
||||
- vllm/
|
||||
- tests/compile
|
||||
commands:
|
||||
# Run unit tests defined directly under compile/,
|
||||
# not including subdirectories, which are usually heavier
|
||||
# tests covered elsewhere.
|
||||
# Use `find` to launch multiple instances of pytest so that
|
||||
# they do not suffer from https://github.com/vllm-project/vllm/issues/28965
|
||||
# However, find does not normally propagate error codes, so we combine it with xargs
|
||||
# (using -0 for proper path handling)
|
||||
- "find compile/ -maxdepth 1 -name 'test_*.py' -print0 | xargs -0 -n1 -I{} pytest -s -v '{}'"
|
||||
|
||||
- label: PyTorch Compilation Passes Unit Tests
|
||||
timeout_in_minutes: 20
|
||||
source_file_dependencies:
|
||||
- vllm/
|
||||
- tests/compile/passes
|
||||
commands:
|
||||
- pytest -s -v compile/passes --ignore compile/passes/distributed
|
||||
|
||||
- label: PyTorch Fullgraph Smoke Test
|
||||
timeout_in_minutes: 35
|
||||
source_file_dependencies:
|
||||
- vllm/
|
||||
- tests/compile
|
||||
commands:
|
||||
# Run smoke tests under fullgraph directory, except test_full_graph.py
|
||||
# as it is a heavy test that is covered in other steps.
|
||||
# Use `find` to launch multiple instances of pytest so that
|
||||
# they do not suffer from https://github.com/vllm-project/vllm/issues/28965
|
||||
- "find compile/fullgraph/ -name 'test_*.py' -not -name 'test_full_graph.py' -exec pytest -s -v {} \\;"
|
||||
|
||||
- label: PyTorch Fullgraph
|
||||
timeout_in_minutes: 30
|
||||
source_file_dependencies:
|
||||
- vllm/
|
||||
- tests/compile
|
||||
commands:
|
||||
# fp8 kv scales not supported on sm89, tested on Blackwell instead
|
||||
- pytest -v -s compile/fullgraph/test_full_graph.py -k 'not test_fp8_kv_scale_compile'
|
||||
|
||||
- label: Pytorch Nightly Dependency Override Check # 2min
|
||||
# if this test fails, it means the nightly torch version is not compatible with some
|
||||
# of the dependencies. Please check the error message and add the package to whitelist
|
||||
# in /vllm/tools/pre_commit/generate_nightly_torch_test.py
|
||||
soft_fail: true
|
||||
source_file_dependencies:
|
||||
- requirements/nightly_torch_test.txt
|
||||
commands:
|
||||
- bash standalone_tests/pytorch_nightly_dependency.sh
|
||||
Reference in New Issue
Block a user