From d841c4729057ef4a09615edbd11239b0cd3127fb Mon Sep 17 00:00:00 2001 From: khluu Date: Tue, 12 May 2026 02:14:24 -0700 Subject: [PATCH] [CI] Migrate all gpu_1_queue jobs to h200_18gb MIG Move all test steps that were falling through to gpu_1_queue (L4) to explicitly use device: h200_18gb (H200 18GB MIG partition). This covers 53 steps across 15 test area files including kernels, entrypoints, models, pytorch compilation, quantization, and more. Signed-off-by: khluu Co-Authored-By: Claude Opus 4.6 (1M context) --- .buildkite/test_areas/cuda.yaml | 1 + .buildkite/test_areas/engine.yaml | 2 ++ .buildkite/test_areas/entrypoints.yaml | 8 ++++++++ .buildkite/test_areas/kernels.yaml | 6 ++++++ .buildkite/test_areas/lm_eval.yaml | 3 +++ .buildkite/test_areas/lora.yaml | 1 + .buildkite/test_areas/misc.yaml | 8 ++++++-- .buildkite/test_areas/model_executor.yaml | 1 + .buildkite/test_areas/model_runner_v2.yaml | 3 +++ .buildkite/test_areas/models_basic.yaml | 5 +++++ .buildkite/test_areas/models_language.yaml | 5 +++++ .buildkite/test_areas/models_multimodal.yaml | 6 ++++++ .buildkite/test_areas/pytorch.yaml | 3 +++ .buildkite/test_areas/quantization.yaml | 2 ++ .buildkite/test_areas/samplers.yaml | 1 + 15 files changed, 53 insertions(+), 2 deletions(-) diff --git a/.buildkite/test_areas/cuda.yaml b/.buildkite/test_areas/cuda.yaml index 6254a6ba3dd..85ef28166f7 100644 --- a/.buildkite/test_areas/cuda.yaml +++ b/.buildkite/test_areas/cuda.yaml @@ -20,6 +20,7 @@ steps: - label: Cudagraph key: cudagraph timeout_in_minutes: 20 + device: h200_18gb source_file_dependencies: - tests/v1/cudagraph - vllm/v1/cudagraph_dispatcher.py diff --git a/.buildkite/test_areas/engine.yaml b/.buildkite/test_areas/engine.yaml index c8b2d66a413..4fa394ca98e 100644 --- a/.buildkite/test_areas/engine.yaml +++ b/.buildkite/test_areas/engine.yaml @@ -30,6 +30,7 @@ steps: - label: Engine (1 GPU) key: engine-1-gpu timeout_in_minutes: 30 + device: h200_18gb source_file_dependencies: - vllm/v1/engine/ - tests/v1/engine/ @@ -50,6 +51,7 @@ steps: - label: e2e Core (1 GPU) key: e2e-core-1-gpu timeout_in_minutes: 30 + device: h200_18gb source_file_dependencies: - vllm/v1/ - tests/v1/e2e/general/ diff --git a/.buildkite/test_areas/entrypoints.yaml b/.buildkite/test_areas/entrypoints.yaml index f9ddf260308..f92f2bda10b 100644 --- a/.buildkite/test_areas/entrypoints.yaml +++ b/.buildkite/test_areas/entrypoints.yaml @@ -6,6 +6,7 @@ steps: key: entrypoints-unit-tests timeout_in_minutes: 10 working_dir: "/vllm-workspace/tests" + device: h200_18gb source_file_dependencies: - vllm/entrypoints - tests/entrypoints/ @@ -17,6 +18,7 @@ steps: key: entrypoints-integration-llm timeout_in_minutes: 40 working_dir: "/vllm-workspace/tests" + device: h200_18gb source_file_dependencies: - vllm/ - tests/entrypoints/llm @@ -36,6 +38,7 @@ steps: key: entrypoints-integration-api-server-openai-part-1 timeout_in_minutes: 50 working_dir: "/vllm-workspace/tests" + device: h200_18gb source_file_dependencies: - vllm/ - tests/entrypoints/openai @@ -48,6 +51,7 @@ steps: key: entrypoints-integration-api-server-openai-part-2 timeout_in_minutes: 50 working_dir: "/vllm-workspace/tests" + device: h200_18gb source_file_dependencies: - vllm/ - tests/entrypoints/openai @@ -73,6 +77,7 @@ steps: key: entrypoints-integration-api-server-2 timeout_in_minutes: 130 working_dir: "/vllm-workspace/tests" + device: h200_18gb source_file_dependencies: - vllm/ - tests/entrypoints/rpc @@ -88,6 +93,7 @@ steps: key: entrypoints-integration-speech_to_text timeout_in_minutes: 50 working_dir: "/vllm-workspace/tests" + device: h200_18gb source_file_dependencies: - vllm/ - tests/entrypoints/speech_to_text @@ -99,6 +105,7 @@ steps: key: entrypoints-integration-pooling timeout_in_minutes: 50 working_dir: "/vllm-workspace/tests" + device: h200_18gb source_file_dependencies: - vllm/ - tests/entrypoints/pooling @@ -110,6 +117,7 @@ steps: key: entrypoints-integration-responses-api timeout_in_minutes: 50 working_dir: "/vllm-workspace/tests" + device: h200_18gb source_file_dependencies: - vllm/ - tests/entrypoints/openai/responses diff --git a/.buildkite/test_areas/kernels.yaml b/.buildkite/test_areas/kernels.yaml index e56252637d3..9b70f7b33af 100644 --- a/.buildkite/test_areas/kernels.yaml +++ b/.buildkite/test_areas/kernels.yaml @@ -17,6 +17,7 @@ steps: - label: Kernels Core Operation Test key: kernels-core-operation-test timeout_in_minutes: 75 + device: h200_18gb source_file_dependencies: - csrc/ - tests/kernels/core @@ -41,6 +42,7 @@ steps: - label: Kernels Attention Test %N key: kernels-attention-test timeout_in_minutes: 35 + device: h200_18gb source_file_dependencies: - csrc/attention/ - vllm/v1/attention @@ -55,6 +57,7 @@ steps: - label: Kernels Quantization Test %N key: kernels-quantization-test timeout_in_minutes: 90 + device: h200_18gb source_file_dependencies: - csrc/quantization/ - vllm/model_executor/layers/quantization @@ -66,6 +69,7 @@ steps: - label: Kernels MoE Test %N key: kernels-moe-test timeout_in_minutes: 25 + device: h200_18gb source_file_dependencies: - csrc/quantization/cutlass_w8a8/moe/ - csrc/moe/ @@ -82,6 +86,7 @@ steps: - label: Kernels Mamba Test key: kernels-mamba-test timeout_in_minutes: 45 + device: h200_18gb source_file_dependencies: - csrc/mamba/ - tests/kernels/mamba @@ -91,6 +96,7 @@ steps: - label: Kernels KDA Test timeout_in_minutes: 20 + device: h200_18gb source_file_dependencies: - vllm/model_executor/layers/fla/ops/kda.py - vllm/model_executor/layers/fla/ops/chunk_delta_h.py diff --git a/.buildkite/test_areas/lm_eval.yaml b/.buildkite/test_areas/lm_eval.yaml index e61e57e7104..240db568a35 100644 --- a/.buildkite/test_areas/lm_eval.yaml +++ b/.buildkite/test_areas/lm_eval.yaml @@ -5,6 +5,7 @@ steps: - label: LM Eval Small Models key: lm-eval-small-models timeout_in_minutes: 75 + device: h200_18gb source_file_dependencies: - csrc/ - vllm/model_executor/layers/quantization @@ -114,6 +115,7 @@ steps: - label: LM Eval TurboQuant KV Cache key: lm-eval-turboquant-kv-cache timeout_in_minutes: 75 + device: h200_18gb source_file_dependencies: - vllm/model_executor/layers/quantization/turboquant/ - vllm/v1/attention/backends/turboquant_attn.py @@ -152,6 +154,7 @@ steps: - label: MRCR Eval Small Models timeout_in_minutes: 30 + device: h200_18gb source_file_dependencies: - tests/evals/mrcr/ commands: diff --git a/.buildkite/test_areas/lora.yaml b/.buildkite/test_areas/lora.yaml index 8107f9b37ff..b9d136e5d29 100644 --- a/.buildkite/test_areas/lora.yaml +++ b/.buildkite/test_areas/lora.yaml @@ -5,6 +5,7 @@ steps: - label: LoRA %N key: lora timeout_in_minutes: 30 + device: h200_18gb source_file_dependencies: - vllm/lora - tests/lora diff --git a/.buildkite/test_areas/misc.yaml b/.buildkite/test_areas/misc.yaml index 2a78201a9e4..5bfbd4764e2 100644 --- a/.buildkite/test_areas/misc.yaml +++ b/.buildkite/test_areas/misc.yaml @@ -5,6 +5,7 @@ steps: - label: V1 Spec Decode key: v1-spec-decode timeout_in_minutes: 30 + device: h200_18gb source_file_dependencies: - vllm/config/ - vllm/distributed/ @@ -58,6 +59,7 @@ steps: - label: V1 Core + KV + Metrics key: v1-core-kv-metrics timeout_in_minutes: 30 + device: h200_18gb source_file_dependencies: - vllm/config/ - vllm/distributed/ @@ -153,6 +155,7 @@ steps: key: examples timeout_in_minutes: 45 working_dir: "/vllm-workspace/examples" + device: h200_18gb source_file_dependencies: - vllm/entrypoints - vllm/multimodal @@ -212,6 +215,7 @@ steps: depends_on: ~ optional: true timeout_in_minutes: 20 + device: h200_18gb source_file_dependencies: - tests/standalone_tests/python_only_compile.sh - setup.py @@ -221,6 +225,7 @@ steps: - label: Async Engine, Inputs, Utils, Worker key: async-engine-inputs-utils-worker timeout_in_minutes: 50 + device: h200_18gb source_file_dependencies: - vllm/assets/ - vllm/config/ @@ -338,9 +343,8 @@ steps: - label: Acceptance Length Test (Large Models) # optional key: acceptance-length-test-large-models timeout_in_minutes: 25 - gpu: h100 + device: h200_18gb optional: true - num_gpus: 1 working_dir: "/vllm-workspace/tests" source_file_dependencies: - vllm/v1/spec_decode/ diff --git a/.buildkite/test_areas/model_executor.yaml b/.buildkite/test_areas/model_executor.yaml index c41ef8a7110..77c3b325e70 100644 --- a/.buildkite/test_areas/model_executor.yaml +++ b/.buildkite/test_areas/model_executor.yaml @@ -5,6 +5,7 @@ steps: - label: Model Executor key: model-executor timeout_in_minutes: 35 + device: h200_18gb source_file_dependencies: - vllm/engine/arg_utils.py - vllm/config/model.py diff --git a/.buildkite/test_areas/model_runner_v2.yaml b/.buildkite/test_areas/model_runner_v2.yaml index 9dfd046289e..86663df3008 100644 --- a/.buildkite/test_areas/model_runner_v2.yaml +++ b/.buildkite/test_areas/model_runner_v2.yaml @@ -5,6 +5,7 @@ steps: - label: Model Runner V2 Core Tests key: model-runner-v2-core-tests timeout_in_minutes: 45 + device: h200_18gb source_file_dependencies: - vllm/v1/worker/gpu/ - vllm/v1/worker/gpu_worker.py @@ -29,6 +30,7 @@ steps: key: model-runner-v2-examples timeout_in_minutes: 45 working_dir: "/vllm-workspace/examples" + device: h200_18gb source_file_dependencies: - vllm/v1/worker/gpu/ - vllm/v1/core/sched/ @@ -102,6 +104,7 @@ steps: key: model-runner-v2-spec-decode timeout_in_minutes: 30 working_dir: "/vllm-workspace/tests" + device: h200_18gb source_file_dependencies: - vllm/v1/worker/gpu/ - vllm/v1/worker/gpu_worker.py diff --git a/.buildkite/test_areas/models_basic.yaml b/.buildkite/test_areas/models_basic.yaml index 8fca203de44..a70e46fa036 100644 --- a/.buildkite/test_areas/models_basic.yaml +++ b/.buildkite/test_areas/models_basic.yaml @@ -6,6 +6,7 @@ steps: key: basic-models-tests-initialization timeout_in_minutes: 45 torch_nightly: true + device: h200_18gb source_file_dependencies: - vllm/ - tests/models/test_initialization.py @@ -19,6 +20,7 @@ steps: - label: Basic Models Tests (Extra Initialization) %N key: basic-models-tests-extra-initialization timeout_in_minutes: 45 + device: h200_18gb source_file_dependencies: - vllm/model_executor/models/ - tests/models/test_initialization.py @@ -35,6 +37,7 @@ steps: - label: Basic Models Tests (Other) key: basic-models-tests-other timeout_in_minutes: 45 + device: h200_18gb source_file_dependencies: - vllm/ - tests/models/test_terratorch.py @@ -61,6 +64,7 @@ steps: working_dir: "/vllm-workspace/" optional: true soft_fail: true + device: h200_18gb commands: - pip install --upgrade git+https://github.com/huggingface/transformers - pytest -v -s tests/models/test_initialization.py @@ -77,6 +81,7 @@ steps: working_dir: "/vllm-workspace/" optional: true soft_fail: true + device: h200_18gb commands: - pip install transformers==4.57.5 - pytest -v -s tests/models/test_initialization.py diff --git a/.buildkite/test_areas/models_language.yaml b/.buildkite/test_areas/models_language.yaml index b560c5a4769..659373798b8 100644 --- a/.buildkite/test_areas/models_language.yaml +++ b/.buildkite/test_areas/models_language.yaml @@ -5,6 +5,7 @@ steps: - label: Language Models Tests (Standard) key: language-models-tests-standard timeout_in_minutes: 25 + device: h200_18gb source_file_dependencies: - vllm/ - tests/models/language @@ -18,6 +19,7 @@ steps: - label: Language Models Tests (Extra Standard) %N key: language-models-tests-extra-standard timeout_in_minutes: 45 + device: h200_18gb source_file_dependencies: - vllm/model_executor/models/ - tests/models/language/pooling/test_embedding.py @@ -35,6 +37,7 @@ steps: - label: Language Models Tests (Hybrid) %N key: language-models-tests-hybrid timeout_in_minutes: 75 + device: h200_18gb source_file_dependencies: - vllm/ - tests/models/language/generation @@ -60,6 +63,7 @@ steps: - label: Language Models Test (Extended Generation) # 80min key: language-models-test-extended-generation timeout_in_minutes: 110 + device: h200_18gb optional: true source_file_dependencies: - vllm/ @@ -85,6 +89,7 @@ steps: - label: Language Models Test (Extended Pooling) # 36min key: language-models-test-extended-pooling timeout_in_minutes: 50 + device: h200_18gb optional: true source_file_dependencies: - vllm/ diff --git a/.buildkite/test_areas/models_multimodal.yaml b/.buildkite/test_areas/models_multimodal.yaml index 1f66393df81..dbfa21f203d 100644 --- a/.buildkite/test_areas/models_multimodal.yaml +++ b/.buildkite/test_areas/models_multimodal.yaml @@ -40,6 +40,7 @@ steps: - label: "Multi-Modal Models (Standard) 3: llava + qwen2_vl" key: multi-modal-models-standard-3-llava-qwen2-vl timeout_in_minutes: 45 + device: h200_18gb source_file_dependencies: - vllm/ - tests/models/multimodal @@ -56,6 +57,7 @@ steps: - label: "Multi-Modal Models (Standard) 4: other + whisper" key: multi-modal-models-standard-4-other-whisper timeout_in_minutes: 45 + device: h200_18gb source_file_dependencies: - vllm/ - tests/models/multimodal @@ -94,6 +96,7 @@ steps: - label: Multi-Modal Accuracy Eval (Small Models) # 50min key: multi-modal-accuracy-eval-small-models timeout_in_minutes: 70 + device: h200_18gb working_dir: "/vllm-workspace/.buildkite/lm-eval-harness" source_file_dependencies: - vllm/multimodal/ @@ -105,6 +108,7 @@ steps: - label: Multi-Modal Models (Extended Generation 1) key: multi-modal-models-extended-generation-1 optional: true + device: h200_18gb source_file_dependencies: - vllm/ - tests/models/multimodal/generation @@ -122,6 +126,7 @@ steps: - label: Multi-Modal Models (Extended Generation 2) key: multi-modal-models-extended-generation-2 optional: true + device: h200_18gb source_file_dependencies: - vllm/ - tests/models/multimodal/generation @@ -132,6 +137,7 @@ steps: - label: Multi-Modal Models (Extended Generation 3) key: multi-modal-models-extended-generation-3 optional: true + device: h200_18gb source_file_dependencies: - vllm/ - tests/models/multimodal/generation diff --git a/.buildkite/test_areas/pytorch.yaml b/.buildkite/test_areas/pytorch.yaml index c58a9986beb..f17ac93775f 100644 --- a/.buildkite/test_areas/pytorch.yaml +++ b/.buildkite/test_areas/pytorch.yaml @@ -5,6 +5,7 @@ steps: - label: PyTorch Compilation Unit Tests key: pytorch-compilation-unit-tests timeout_in_minutes: 10 + device: h200_18gb source_file_dependencies: - vllm/__init__.py - vllm/_aiter_ops.py @@ -78,6 +79,7 @@ steps: - label: PyTorch Compilation Passes Unit Tests key: pytorch-compilation-passes-unit-tests timeout_in_minutes: 20 + device: h200_18gb source_file_dependencies: - vllm/__init__.py - vllm/_aiter_ops.py @@ -110,6 +112,7 @@ steps: - label: PyTorch Fullgraph Smoke Test key: pytorch-fullgraph-smoke-test timeout_in_minutes: 35 + device: h200_18gb source_file_dependencies: - vllm/__init__.py - vllm/_aiter_ops.py diff --git a/.buildkite/test_areas/quantization.yaml b/.buildkite/test_areas/quantization.yaml index 8a9a36da448..23e9ce58177 100644 --- a/.buildkite/test_areas/quantization.yaml +++ b/.buildkite/test_areas/quantization.yaml @@ -5,6 +5,7 @@ steps: - label: Quantization key: quantization timeout_in_minutes: 90 + device: h200_18gb source_file_dependencies: - csrc/ - vllm/model_executor/layers/quantization @@ -42,6 +43,7 @@ steps: - label: Quantized Models Test key: quantized-models-test timeout_in_minutes: 60 + device: h200_18gb source_file_dependencies: - vllm/model_executor/layers/quantization - tests/models/quantization diff --git a/.buildkite/test_areas/samplers.yaml b/.buildkite/test_areas/samplers.yaml index 48e9f55571e..110c83163a0 100644 --- a/.buildkite/test_areas/samplers.yaml +++ b/.buildkite/test_areas/samplers.yaml @@ -5,6 +5,7 @@ steps: - label: Samplers Test key: samplers-test timeout_in_minutes: 75 + device: h200_18gb source_file_dependencies: - vllm/model_executor/layers - vllm/sampling_metadata.py