a53ad85913 · tduncan/Nvidia-Optimization-Stack-VLLM

.buildkite/intel_jobs/basic_correctness.yaml

+1 -2

@@ -18,11 +18,10 @@ steps:

18 source_file_dependencies:

19 - vllm/

20 - tests/basic_correctness/test_cumem.py

21 - # Remove VLLM_DISABLE_COMPILE_CACHE=1 once intel/intel-xpu-backend-for-triton#7682 is fixed

21 commands:

22 - >-

23 bash .buildkite/scripts/hardware_ci/run-intel-test.sh

24 'cd tests &&

25 export VLLM_WORKER_MULTIPROC_METHOD=spawn &&

27 - VLLM_DISABLE_COMPILE_CACHE=1 pytest -v -s basic_correctness/test_cpu_offload.py &&

26 + pytest -v -s basic_correctness/test_cpu_offload.py &&

27 pytest -v -s basic_correctness/test_mem.py::test_end_to_end'

.buildkite/intel_jobs/benchmarks_intel.yaml

+1 -2

@@ -19,9 +19,8 @@ steps:

19 source_file_dependencies:

20 - vllm/

21 - tests/benchmarks/

22 - # Remove VLLM_DISABLE_COMPILE_CACHE=1 once intel/intel-xpu-backend-for-triton#7682 is fixed

22 commands:

23 - >-

24 bash .buildkite/scripts/hardware_ci/run-intel-test.sh

25 'cd tests &&

27 - VLLM_DISABLE_COMPILE_CACHE=1 pytest -v -s benchmarks/'

26 + pytest -v -s benchmarks/'

.buildkite/intel_jobs/model_executor_intel.yaml

+1 -2

@@ -21,7 +21,6 @@ steps:

21 - vllm/config/model.py

22 - vllm/model_executor

23 - tests/model_executor

24 - # Remove VLLM_DISABLE_COMPILE_CACHE=1 once intel/intel-xpu-backend-for-triton#7682 is fixed

24 commands:

25 - >-

26 bash .buildkite/scripts/hardware_ci/run-intel-test.sh

@@ -31,4 +30,4 @@ steps:

30 export VLLM_WORKER_MULTIPROC_METHOD=spawn &&

31 export PYTHONFAULTHANDLER=1 &&

32 cd tests &&

34 - VLLM_DISABLE_COMPILE_CACHE=1 pytest -v -s model_executor -m "not slow_test" --ignore="model_executor/layers/test_rocm_unquantized_gemm.py" --deselect="tests/model_executor/model_loader/test_reload.py::test_kv_scale_reload"'

33 + pytest -v -s model_executor -m "not slow_test" --ignore="model_executor/layers/test_rocm_unquantized_gemm.py" --deselect="tests/model_executor/model_loader/test_reload.py::test_kv_scale_reload"'

.buildkite/intel_jobs/model_runner_v2_intel.yaml

+1 -2

@@ -22,14 +22,13 @@ steps:

22 - vllm/v1/attention/

23 - tests/v1/engine/test_llm_engine.py

24 - tests/v1/e2e/

25 - # Remove VLLM_DISABLE_COMPILE_CACHE=1 once intel/intel-xpu-backend-for-triton#7682 is fixed

25 commands:

26 - >-

27 bash .buildkite/scripts/hardware_ci/run-intel-test.sh

28 'export VLLM_USE_V2_MODEL_RUNNER=1 &&

29 cd tests &&

30 pytest -v -s v1/engine/test_llm_engine.py -k "not test_engine_metrics" &&

32 - VLLM_DISABLE_COMPILE_CACHE=1 pytest -v -s v1/e2e/general/test_context_length.py &&

31 + pytest -v -s v1/e2e/general/test_context_length.py &&

32 ENFORCE_EAGER=1 pytest -v -s v1/e2e/general/test_async_scheduling.py -k "not ngram" &&

33 pytest -v -s entrypoints/llm/test_struct_output_generate.py -k "xgrammar and not speculative_config6 and not speculative_config7 and not speculative_config8 and not speculative_config0" &&

34 pytest -v -s v1/e2e/general/test_min_tokens.py'

.buildkite/intel_jobs/models_distributed_intel.yaml

+1 -2

@@ -20,9 +20,8 @@ steps:

20 - vllm/model_executor/model_loader/sharded_state_loader.py

21 - vllm/model_executor/models/

22 - tests/model_executor/model_loader/test_sharded_state_loader.py

23 - # Remove VLLM_DISABLE_COMPILE_CACHE=1 once intel/intel-xpu-backend-for-triton#7682 is fixed

23 commands:

24 - >-

25 bash .buildkite/scripts/hardware_ci/run-intel-test.sh

26 'cd tests &&

28 - VLLM_DISABLE_COMPILE_CACHE=1 pytest -v -s model_executor/model_loader/test_sharded_state_loader.py -m "not slow_test"'

27 + pytest -v -s model_executor/model_loader/test_sharded_state_loader.py -m "not slow_test"'

.buildkite/intel_jobs/samplers_intel.yaml

+1 -2

@@ -22,9 +22,8 @@ steps:

22 - tests/samplers

23 - tests/conftest.py

24 - vllm/entrypoints/generate/beam_search

25 - # Remove VLLM_DISABLE_COMPILE_CACHE=1 once intel/intel-xpu-backend-for-triton#7682 is fixed

25 commands:

26 - >-

27 bash .buildkite/scripts/hardware_ci/run-intel-test.sh

28 'cd tests &&

30 - VLLM_XPU_USE_SAMPLER_KERNEL=1 VLLM_DISABLE_COMPILE_CACHE=1 pytest -v -s samplers'

29 + VLLM_XPU_USE_SAMPLER_KERNEL=1 pytest -v -s samplers'

.buildkite/intel_jobs/test-intel.yaml

+1 -2

@@ -92,7 +92,6 @@ steps:

92 source_file_dependencies:

93 - vllm/

94 - .buildkite/intel_jobs/test-intel.yaml

95 - # Remove VLLM_DISABLE_COMPILE_CACHE=1 once intel/intel-xpu-backend-for-triton#7682 is fixed

95 commands:

96 - >-

97 bash .buildkite/scripts/hardware_ci/run-intel-test.sh

@@ -103,7 +102,7 @@ steps:

102 pytest -v -s v1/worker --ignore=v1/worker/test_gpu_model_runner.py --ignore=v1/worker/test_worker_memory_snapshot.py &&

103 pytest -v -s v1/structured_output &&

104 pytest -v -s v1/test_serial_utils.py &&

106 - VLLM_DISABLE_COMPILE_CACHE=1 pytest -v -s v1/e2e/general/test_correctness_sliding_window.py &&

105 + pytest -v -s v1/e2e/general/test_correctness_sliding_window.py &&

106 pytest -v -s v1/spec_decode --ignore=v1/spec_decode/test_max_len.py --ignore=v1/spec_decode/test_speculators_eagle3.py --ignore=v1/spec_decode/test_acceptance_length.py --ignore=v1/spec_decode/test_speculators_correctness.py &&

107 pytest -v -s v1/kv_connector/unit --ignore=v1/kv_connector/unit/test_lmcache_integration.py --ignore=v1/kv_connector/unit/test_hf3fs_client.py --ignore=v1/kv_connector/unit/test_hf3fs_connector.py --ignore=v1/kv_connector/unit/test_hf3fs_metadata_server.py --ignore=v1/kv_connector/unit/test_offloading_connector.py'

108 - label: "XPU server test"

.buildkite/scripts/hardware_ci/run-intel-ci-test.sh

+2 -2

@@ -28,14 +28,14 @@ case "${test_suite}" in

28 ;;

29 v1)

30 cd tests

31 - # Remove VLLM_DISABLE_COMPILE_CACHE=1 once intel/intel-xpu-backend-for-triton#7682 is fixed

31 +

32 pytest -v -s v1/core --ignore=v1/core/test_reset_prefix_cache_e2e.py --ignore=v1/core/test_scheduler_e2e.py

33 pytest -v -s v1/engine --ignore=v1/engine/test_output_processor.py

34 pytest -v -s v1/sample --ignore=v1/sample/test_logprobs.py --ignore=v1/sample/test_logprobs_e2e.py -k "not test_topk_only and not test_topp_only and not test_topk_and_topp"

35 pytest -v -s v1/worker --ignore=v1/worker/test_gpu_model_runner.py --ignore=v1/worker/test_worker_memory_snapshot.py

36 pytest -v -s v1/structured_output

37 pytest -v -s v1/test_serial_utils.py

38 - VLLM_DISABLE_COMPILE_CACHE=1 pytest -v -s v1/e2e/general/test_correctness_sliding_window.py

38 + pytest -v -s v1/e2e/general/test_correctness_sliding_window.py

39 pytest -v -s v1/spec_decode --ignore=v1/spec_decode/test_max_len.py --ignore=v1/spec_decode/test_speculators_eagle3.py --ignore=v1/spec_decode/test_acceptance_length.py --ignore=v1/spec_decode/test_speculators_correctness.py

40 pytest -v -s v1/kv_connector/unit --ignore=v1/kv_connector/unit/test_multi_connector.py --ignore=v1/kv_connector/unit/test_example_connector.py --ignore=v1/kv_connector/unit/test_lmcache_integration.py --ignore=v1/kv_connector/unit/test_hf3fs_client.py --ignore=v1/kv_connector/unit/test_hf3fs_connector.py --ignore=v1/kv_connector/unit/test_hf3fs_metadata_server.py --ignore=v1/kv_connector/unit/test_offloading_connector.py

41 ;;

.buildkite/scripts/hardware_ci/run-intel-test.sh

+4 -1

@@ -18,7 +18,9 @@ if [[ "${1:-}" == "--dry-run" ]]; then

18 fi

19

20 # Export Python path

21 +# Remove VLLM_DISABLE_COMPILE_CACHE=1 once intel/intel-xpu-backend-for-triton#7682 is fixed

22 export PYTHONPATH=".."

23 +export VLLM_DISABLE_COMPILE_CACHE=1

24

25 ###############################################################################

26 # Helper Functions

@@ -342,7 +344,7 @@ if [[ -z "${ZE_AFFINITY_MASK:-}" ]]; then

344 fi

345

346 export CMDS="${commands}"

345 -export HF_TOKEN ZE_AFFINITY_MASK

347 +export HF_TOKEN ZE_AFFINITY_MASK VLLM_DISABLE_COMPILE_CACHE

348

349 {

350 flock 9

@@ -364,6 +366,7 @@ export HF_TOKEN ZE_AFFINITY_MASK

366 --entrypoint='' \

367 -e HF_TOKEN \

368 -e ZE_AFFINITY_MASK \

369 + -e VLLM_DISABLE_COMPILE_CACHE \

370 -e BUILDKITE_PARALLEL_JOB \

371 -e BUILDKITE_PARALLEL_JOB_COUNT \

372 -e CMDS \

[XPU][CI]Change to use global VLLM_DISABLE_COMPILE_CACHE=1 in Intel GPU CI (#51905)