@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 0ab67c0222
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 4, 11:15 UTC

[CI] Add key field to all test_areas pipeline steps (#41201)

Signed-off-by: khluu <khluu000@gmail.com> Co-authored-by: Claude <noreply@anthropic.com>

Kevin H. Luu committed Apr 29, 2026 at 16:59 UTC 0ab67c02225c18ed664864f75c3a38c0659f4be1
28 files changed +162 -3
.buildkite/test_areas/attention.yaml
+2
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: V1 attention (H100)
6 + key: v1-attention-h100
7 timeout_in_minutes: 30
8 device: h100
9 source_file_dependencies:
@@ -14,6 +15,7 @@ steps:
15 - pytest -v -s v1/attention
16
17 - label: V1 attention (B200)
18 + key: v1-attention-b200
19 timeout_in_minutes: 30
20 device: b200
21 source_file_dependencies:
.buildkite/test_areas/basic_correctness.yaml
+1
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Basic Correctness
6 + key: basic-correctness
7 timeout_in_minutes: 30
8 device: h200_18gb
9 source_file_dependencies:
.buildkite/test_areas/benchmarks.yaml
+2
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Benchmarks CLI Test
6 + key: benchmarks-cli-test
7 timeout_in_minutes: 20
8 device: h200_18gb
9 source_file_dependencies:
@@ -12,6 +13,7 @@ steps:
13 - pytest -v -s benchmarks/
14
15 - label: Attention Benchmarks Smoke Test (B200)
16 + key: attention-benchmarks-smoke-test-b200
17 device: b200
18 num_gpus: 2
19 optional: true
.buildkite/test_areas/compile.yaml
+13
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Sequence Parallel Correctness Tests (2 GPUs)
6 + key: sequence-parallel-correctness-tests-2-gpus
7 timeout_in_minutes: 50
8 working_dir: "/vllm-workspace/"
9 num_devices: 2
@@ -17,6 +18,7 @@ steps:
18 - pytest -v -s tests/compile/correctness_e2e/test_sequence_parallel.py
19
20 - label: Sequence Parallel Correctness Tests (2xH100)
21 + key: sequence-parallel-correctness-tests-2xh100
22 timeout_in_minutes: 50
23 working_dir: "/vllm-workspace/"
24 device: h100
@@ -27,6 +29,7 @@ steps:
29 - pytest -v -s tests/compile/correctness_e2e/test_sequence_parallel.py
30
31 - label: AsyncTP Correctness Tests (2xH100)
32 + key: asynctp-correctness-tests-2xh100
33 timeout_in_minutes: 50
34 working_dir: "/vllm-workspace/"
35 device: h100
@@ -37,6 +40,7 @@ steps:
40 - pytest -v -s tests/compile/correctness_e2e/test_async_tp.py
41
42 - label: AsyncTP Correctness Tests (B200)
43 + key: asynctp-correctness-tests-b200
44 timeout_in_minutes: 50
45 working_dir: "/vllm-workspace/"
46 device: b200
@@ -47,6 +51,7 @@ steps:
51 - pytest -v -s tests/compile/correctness_e2e/test_async_tp.py
52
53 - label: Distributed Compile Unit Tests (2xH100)
54 + key: distributed-compile-unit-tests-2xh100
55 timeout_in_minutes: 20
56 working_dir: "/vllm-workspace/"
57 device: h100
@@ -60,6 +65,7 @@ steps:
65 - pytest -s -v tests/compile/passes/distributed
66
67 - label: Fusion and Compile Unit Tests (2xB200)
68 + key: fusion-and-compile-unit-tests-2xb200
69 timeout_in_minutes: 20
70 working_dir: "/vllm-workspace/"
71 device: b200
@@ -89,6 +95,7 @@ steps:
95 - pytest -v -s tests/compile/fullgraph/test_full_graph.py::test_fp8_kv_scale_compile
96
97 - label: Fusion E2E Quick (H100)
98 + key: fusion-e2e-quick-h100
99 timeout_in_minutes: 15
100 working_dir: "/vllm-workspace/"
101 device: h100
@@ -107,6 +114,7 @@ steps:
114 - pytest -v -s tests/compile/fusions_e2e/test_tp1_quant.py -k "inductor_partition and not +rms_norm and +quant_fp8 and (qwen3 or deepseek)"
115
116 - label: Fusion E2E Config Sweep (H100)
117 + key: fusion-e2e-config-sweep-h100
118 timeout_in_minutes: 30
119 working_dir: "/vllm-workspace/"
120 device: h100
@@ -126,6 +134,7 @@ steps:
134 - pytest -v -s tests/compile/fusions_e2e/test_tp1_quant.py -k "llama-3"
135
136 - label: Fusion E2E Config Sweep (B200)
137 + key: fusion-e2e-config-sweep-b200
138 timeout_in_minutes: 30
139 working_dir: "/vllm-workspace/"
140 device: b200
@@ -139,6 +148,7 @@ steps:
148 - pytest -v -s tests/compile/fusions_e2e/test_tp1_quant.py -k "inductor_partition and (FLASHINFER and not +rms_norm and (not +quant_fp8 or +quant_fp8 and (qwen3 or deepseek)) or llama-3)"
149
150 - label: Fusion E2E TP2 Quick (H100)
151 + key: fusion-e2e-tp2-quick-h100
152 timeout_in_minutes: 20
153 working_dir: "/vllm-workspace/"
154 device: h100
@@ -156,6 +166,7 @@ steps:
166 - pytest -v -s tests/compile/fusions_e2e/test_tp2_async_tp.py -k "inductor_partition and not +rms_norm and (not +quant_fp8 or +quant_fp8 and (qwen3 or deepseek))"
167
168 - label: Fusion E2E TP2 AR-RMS Config Sweep (H100)
169 + key: fusion-e2e-tp2-ar-rms-config-sweep-h100
170 timeout_in_minutes: 40
171 working_dir: "/vllm-workspace/"
172 device: h100
@@ -175,6 +186,7 @@ steps:
186 - pytest -v -s tests/compile/fusions_e2e/test_tp2_ar_rms.py -k "llama-3"
187
188 - label: Fusion E2E TP2 AsyncTP Config Sweep (H100)
189 + key: fusion-e2e-tp2-asynctp-config-sweep-h100
190 timeout_in_minutes: 40
191 working_dir: "/vllm-workspace/"
192 device: h100
@@ -194,6 +206,7 @@ steps:
206 - pytest -v -s tests/compile/fusions_e2e/test_tp2_async_tp.py -k "llama-3"
207
208 - label: Fusion E2E TP2 (B200)
209 + key: fusion-e2e-tp2-b200
210 timeout_in_minutes: 20
211 working_dir: "/vllm-workspace/"
212 device: b200
.buildkite/test_areas/cuda.yaml
+2
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Platform Tests (CUDA)
6 + key: platform-tests-cuda
7 timeout_in_minutes: 15
8 device: h200_18gb
9 source_file_dependencies:
@@ -13,6 +14,7 @@ steps:
14 - pytest -v -s cuda/test_platform_no_cuda_init.py
15
16 - label: Cudagraph
17 + key: cudagraph
18 timeout_in_minutes: 20
19 source_file_dependencies:
20 - tests/v1/cudagraph
.buildkite/test_areas/disaggregated.yaml
+8
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Distributed NixlConnector PD accuracy (4 GPUs)
6 + key: distributed-nixlconnector-pd-accuracy-4-gpus
7 timeout_in_minutes: 30
8 working_dir: "/vllm-workspace/tests"
9 num_devices: 4
@@ -13,6 +14,7 @@ steps:
14 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
15 - bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
16 - label: Distributed FlashInfer NixlConnector PD accuracy (4 GPUs)
17 + key: distributed-flashinfer-nixlconnector-pd-accuracy-4-gpus
18 timeout_in_minutes: 30
19 working_dir: "/vllm-workspace/tests"
20 num_devices: 4
@@ -24,6 +26,7 @@ steps:
26 - FLASHINFER=1 bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
27
28 - label: DP EP Distributed NixlConnector PD accuracy tests (4 GPUs)
29 + key: dp-ep-distributed-nixlconnector-pd-accuracy-tests-4-gpus
30 timeout_in_minutes: 30
31 working_dir: "/vllm-workspace/tests"
32 num_devices: 4
@@ -35,6 +38,7 @@ steps:
38 - DP_EP=1 bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
39
40 - label: CrossLayer KV layout Distributed NixlConnector PD accuracy tests (4 GPUs)
41 + key: crosslayer-kv-layout-distributed-nixlconnector-pd-accuracy-tests-4-gpus
42 timeout_in_minutes: 30
43 working_dir: "/vllm-workspace/tests"
44 num_devices: 4
@@ -46,6 +50,7 @@ steps:
50 - CROSS_LAYERS_BLOCKS=True bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
51
52 - label: Hybrid SSM NixlConnector PD accuracy tests (4 GPUs)
53 + key: hybrid-ssm-nixlconnector-pd-accuracy-tests-4-gpus
54 timeout_in_minutes: 20
55 working_dir: "/vllm-workspace/tests"
56 num_devices: 4
@@ -57,6 +62,7 @@ steps:
62 - HYBRID_SSM=1 bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
63
64 - label: MultiConnector (Nixl+Offloading) PD accuracy (2 GPUs)
65 + key: multiconnector-nixl-offloading-pd-accuracy-2-gpus
66 timeout_in_minutes: 30
67 working_dir: "/vllm-workspace/tests"
68 num_devices: 2
@@ -71,6 +77,7 @@ steps:
77 - bash v1/kv_connector/nixl_integration/run_multi_connector_accuracy_test.sh
78
79 - label: NixlConnector PD + Spec Decode acceptance (2 GPUs)
80 + key: nixlconnector-pd-spec-decode-acceptance-2-gpus
81 timeout_in_minutes: 30
82 device: a100
83 working_dir: "/vllm-workspace/tests"
@@ -84,6 +91,7 @@ steps:
91 - bash v1/kv_connector/nixl_integration/spec_decode_acceptance_test.sh
92
93 - label: MultiConnector (Nixl+Offloading) PD edge cases (2 GPUs)
94 + key: multiconnector-nixl-offloading-pd-edge-cases-2-gpus
95 timeout_in_minutes: 30
96 working_dir: "/vllm-workspace/tests"
97 num_devices: 2
.buildkite/test_areas/distributed.yaml
+14
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Distributed Comm Ops
6 + key: distributed-comm-ops
7 timeout_in_minutes: 20
8 working_dir: "/vllm-workspace/tests"
9 num_devices: 2
@@ -16,6 +17,7 @@ steps:
17 - pytest -v -s distributed/test_shm_storage.py
18
19 - label: Distributed DP Tests (2 GPUs)
20 + key: distributed-dp-tests-2-gpus
21 timeout_in_minutes: 20
22 working_dir: "/vllm-workspace/tests"
23 num_devices: 2
@@ -37,6 +39,7 @@ steps:
39 - DP_SIZE=2 pytest -v -s entrypoints/openai/test_multi_api_servers.py
40
41 - label: Distributed Compile + RPC Tests (2 GPUs)
42 + key: distributed-compile-rpc-tests-2-gpus
43 timeout_in_minutes: 20
44 working_dir: "/vllm-workspace/tests"
45 num_devices: 2
@@ -59,6 +62,7 @@ steps:
62 - pytest -v -s ./compile/test_wrapper.py
63
64 - label: Distributed Torchrun + Shutdown Tests (2 GPUs)
65 + key: distributed-torchrun-shutdown-tests-2-gpus
66 timeout_in_minutes: 20
67 working_dir: "/vllm-workspace/tests"
68 num_devices: 2
@@ -81,6 +85,7 @@ steps:
85 - pytest -v -s v1/worker/test_worker_memory_snapshot.py
86
87 - label: Distributed Torchrun + Examples (4 GPUs)
88 + key: distributed-torchrun-examples-4-gpus
89 timeout_in_minutes: 30
90 working_dir: "/vllm-workspace"
91 num_devices: 4
@@ -112,6 +117,7 @@ steps:
117 - VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 examples/rl/rlhf_ipc.py
118
119 - label: Distributed DP Tests (4 GPUs)
120 + key: distributed-dp-tests-4-gpus
121 timeout_in_minutes: 30
122 working_dir: "/vllm-workspace/tests"
123 num_devices: 4
@@ -132,6 +138,7 @@ steps:
138 - pytest -v -s distributed/test_utils.py
139
140 - label: Distributed Compile + Comm (4 GPUs)
141 + key: distributed-compile-comm-4-gpus
142 timeout_in_minutes: 30
143 working_dir: "/vllm-workspace/tests"
144 num_devices: 4
@@ -153,6 +160,7 @@ steps:
160 - pytest -v -s distributed/test_multiproc_executor.py::test_multiproc_executor_multi_node
161
162 - label: Distributed Tests (8 GPUs)(H100)
163 + key: distributed-tests-8-gpus-h100
164 timeout_in_minutes: 10
165 device: h100
166 num_devices: 8
@@ -173,6 +181,7 @@ steps:
181 - torchrun --nproc-per-node=8 ../examples/features/torchrun/torchrun_dp_example_offline.py --tp-size=2 --pp-size=1 --dp-size=4 --enable-ep
182
183 - label: Distributed Tests (4 GPUs)(A100)
184 + key: distributed-tests-4-gpus-a100
185 device: a100
186 optional: true
187 num_devices: 4
@@ -187,6 +196,7 @@ steps:
196 - pytest -v -s -x lora/test_mixtral.py
197
198 - label: Distributed Tests (2 GPUs)(H100)
199 + key: distributed-tests-2-gpus-h100
200 timeout_in_minutes: 15
201 device: h100
202 optional: true
@@ -201,6 +211,7 @@ steps:
211 - pytest -v -s tests/distributed/test_packed_tensor.py
212
213 - label: Distributed Tests (2 GPUs)(B200)
214 + key: distributed-tests-2-gpus-b200
215 device: b200
216 optional: true
217 working_dir: "/vllm-workspace/"
@@ -214,6 +225,7 @@ steps:
225
226
227 - label: 2 Node Test (4 GPUs)
228 + key: 2-node-test-4-gpus
229 timeout_in_minutes: 30
230 working_dir: "/vllm-workspace/tests"
231 num_devices: 2
@@ -231,6 +243,7 @@ steps:
243 - ./.buildkite/scripts/run-multi-node-test.sh /vllm-workspace/tests 2 2 $IMAGE_TAG "VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed' && NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed' && python3 ../examples/features/data_parallel/data_parallel_offline.py -dp=2 -tp=1 --dp-num-nodes=2 --dp-node-rank=0 --dp-master-addr=192.168.10.10 --dp-master-port=12345 --enforce-eager --trust-remote-code && VLLM_MULTI_NODE=1 pytest -v -s distributed/test_multi_node_assignment.py && VLLM_MULTI_NODE=1 pytest -v -s distributed/test_pipeline_parallel.py" "VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed' && NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed' && python3 ../examples/features/data_parallel/data_parallel_offline.py -dp=2 -tp=1 --dp-num-nodes=2 --dp-node-rank=1 --dp-master-addr=192.168.10.10 --dp-master-port=12345 --enforce-eager --trust-remote-code"
244
245 - label: Pipeline + Context Parallelism (4 GPUs)
246 + key: pipeline-context-parallelism-4-gpus
247 timeout_in_minutes: 60
248 working_dir: "/vllm-workspace/tests"
249 num_devices: 4
@@ -245,6 +258,7 @@ steps:
258 - pytest -v -s distributed/test_pipeline_parallel.py
259
260 - label: RayExecutorV2 (4 GPUs)
261 + key: rayexecutorv2-4-gpus
262 timeout_in_minutes: 60
263 working_dir: "/vllm-workspace/tests"
264 num_devices: 4
.buildkite/test_areas/e2e_integration.yaml
+5
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: DeepSeek V2-Lite Accuracy
6 + key: deepseek-v2-lite-accuracy
7 timeout_in_minutes: 60
8 device: h100
9 optional: true
@@ -12,6 +13,7 @@ steps:
13 - bash .buildkite/scripts/scheduled_integration_test/deepseek_v2_lite_ep_eplb.sh 0.25 200 8010
14
15 - label: Qwen3-30B-A3B-FP8-block Accuracy
16 + key: qwen3-30b-a3b-fp8-block-accuracy
17 timeout_in_minutes: 60
18 device: h100
19 optional: true
@@ -21,6 +23,7 @@ steps:
23 - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_block_ep_eplb.sh 0.8 200 8020
24
25 - label: Qwen3-30B-A3B-FP8-block Accuracy (B200)
26 + key: qwen3-30b-a3b-fp8-block-accuracy-b200
27 timeout_in_minutes: 60
28 device: b200
29 optional: true
@@ -30,6 +33,7 @@ steps:
33 - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_block_ep_eplb.sh 0.8 200 8020 2 1
34
35 - label: Qwen3-30B-A3B-FP8 DP4 Async EPLB Accuracy
36 + key: qwen3-30b-a3b-fp8-dp4-async-eplb-accuracy
37 timeout_in_minutes: 60
38 device: h100
39 optional: true
@@ -39,6 +43,7 @@ steps:
43 - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_dp4_async_eplb.sh 0.8 200 8050
44
45 - label: DeepSeek V2-Lite Prefetch Offload Accuracy (H100)
46 + key: deepseek-v2-lite-prefetch-offload-accuracy-h100
47 timeout_in_minutes: 60
48 device: h100
49 optional: true
.buildkite/test_areas/engine.yaml
+7
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Engine
6 + key: engine
7 timeout_in_minutes: 15
8 device: h200_18gb
9 source_file_dependencies:
@@ -16,6 +17,7 @@ steps:
17 - pytest -v -s engine test_sequence.py test_config.py test_logger.py test_vllm_port.py
18
19 - label: Engine (1 GPU)
20 + key: engine-1-gpu
21 timeout_in_minutes: 30
22 source_file_dependencies:
23 - vllm/v1/engine/
@@ -25,6 +27,7 @@ steps:
27 - pytest -v -s v1/engine --ignore v1/engine/test_preprocess_error_handling.py
28
29 - label: e2e Scheduling (1 GPU)
30 + key: e2e-scheduling-1-gpu
31 timeout_in_minutes: 30
32 device: h200_18gb
33 source_file_dependencies:
@@ -34,6 +37,7 @@ steps:
37 - pytest -v -s v1/e2e/general/test_async_scheduling.py
38
39 - label: e2e Core (1 GPU)
40 + key: e2e-core-1-gpu
41 timeout_in_minutes: 30
42 source_file_dependencies:
43 - vllm/v1/
@@ -42,6 +46,7 @@ steps:
46 - pytest -v -s v1/e2e/general --ignore v1/e2e/general/test_async_scheduling.py
47
48 - label: V1 e2e (2 GPUs)
49 + key: v1-e2e-2-gpus
50 timeout_in_minutes: 60 # TODO: Fix timeout after we have more confidence in the test stability
51 optional: true
52 num_devices: 2
@@ -58,6 +63,7 @@ steps:
63 - image-build-amd
64
65 - label: V1 e2e (4 GPUs)
66 + key: v1-e2e-4-gpus
67 timeout_in_minutes: 60 # TODO: Fix timeout after we have more confidence in the test stability
68 optional: true
69 num_devices: 4
@@ -74,6 +80,7 @@ steps:
80 - image-build-amd
81
82 - label: V1 e2e (4xH100)
83 + key: v1-e2e-4xh100
84 timeout_in_minutes: 60
85 device: h100
86 num_devices: 4
.buildkite/test_areas/entrypoints.yaml
+10 -1
@@ -2,7 +2,8 @@ group: Entrypoints
2 depends_on:
3 - image-build
4 steps:
5 -- label: Entrypoints Unit Tests
5 +- label: Entrypoints Unit Tests
6 + key: entrypoints-unit-tests
7 timeout_in_minutes: 10
8 working_dir: "/vllm-workspace/tests"
9 source_file_dependencies:
@@ -13,6 +14,7 @@ steps:
14 - pytest -v -s entrypoints/ --ignore=entrypoints/llm --ignore=entrypoints/rpc --ignore=entrypoints/sleep --ignore=entrypoints/serve/instrumentator --ignore=entrypoints/openai --ignore=entrypoints/offline_mode --ignore=entrypoints/test_chat_utils.py --ignore=entrypoints/pooling
15
16 - label: Entrypoints Integration (LLM)
17 + key: entrypoints-integration-llm
18 timeout_in_minutes: 40
19 working_dir: "/vllm-workspace/tests"
20 source_file_dependencies:
@@ -26,6 +28,7 @@ steps:
28 - pytest -v -s entrypoints/offline_mode # Needs to avoid interference with other tests
29
30 - label: Entrypoints Integration (API Server openai - Part 1)
31 + key: entrypoints-integration-api-server-openai-part-1
32 timeout_in_minutes: 50
33 working_dir: "/vllm-workspace/tests"
34 source_file_dependencies:
@@ -43,6 +46,7 @@ steps:
46
47
48 - label: Entrypoints Integration (API Server openai - Part 2)
49 + key: entrypoints-integration-api-server-openai-part-2
50 timeout_in_minutes: 50
51 working_dir: "/vllm-workspace/tests"
52 source_file_dependencies:
@@ -60,6 +64,7 @@ steps:
64 - image-build-amd
65
66 - label: Entrypoints Integration (API Server openai - Part 3)
67 + key: entrypoints-integration-api-server-openai-part-3
68 timeout_in_minutes: 50
69 device: h200_18gb
70 working_dir: "/vllm-workspace/tests"
@@ -72,6 +77,7 @@ steps:
77 - pytest -v -s entrypoints/openai --ignore=entrypoints/openai/chat_completion --ignore=entrypoints/openai/completion --ignore=entrypoints/openai/speech_to_text/ --ignore=entrypoints/openai/correctness/ --ignore=entrypoints/openai/tool_parsers/ --ignore=entrypoints/openai/responses --ignore=entrypoints/openai/test_multi_api_servers.py
78
79 - label: Entrypoints Integration (API Server 2)
80 + key: entrypoints-integration-api-server-2
81 timeout_in_minutes: 130
82 working_dir: "/vllm-workspace/tests"
83 source_file_dependencies:
@@ -86,6 +92,7 @@ steps:
92 - pytest -v -s tool_use
93
94 - label: Entrypoints Integration (Pooling)
95 + key: entrypoints-integration-pooling
96 timeout_in_minutes: 50
97 working_dir: "/vllm-workspace/tests"
98 source_file_dependencies:
@@ -96,6 +103,7 @@ steps:
103 - pytest -v -s entrypoints/pooling
104
105 - label: Entrypoints Integration (Responses API)
106 + key: entrypoints-integration-responses-api
107 timeout_in_minutes: 50
108 working_dir: "/vllm-workspace/tests"
109 source_file_dependencies:
@@ -105,6 +113,7 @@ steps:
113 - pytest -v -s entrypoints/openai/responses
114
115 - label: OpenAI API Correctness
116 + key: openai-api-correctness
117 timeout_in_minutes: 30
118 device: h200_18gb
119 source_file_dependencies:
.buildkite/test_areas/expert_parallelism.yaml
+3
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: EPLB Algorithm
6 + key: eplb-algorithm
7 timeout_in_minutes: 15
8 device: h200_18gb
9 working_dir: "/vllm-workspace/tests"
@@ -15,6 +16,7 @@ steps:
16 - pytest -v -s distributed/test_eplb_utils.py
17
18 - label: EPLB Execution # 17min
19 + key: eplb-execution
20 timeout_in_minutes: 27
21 working_dir: "/vllm-workspace/tests"
22 num_devices: 4
@@ -26,6 +28,7 @@ steps:
28 - pytest -v -s distributed/test_eplb_spec_decode.py
29
30 - label: Elastic EP Scaling Test
31 + key: elastic-ep-scaling-test
32 timeout_in_minutes: 20
33 device: h100
34 working_dir: "/vllm-workspace/tests"
.buildkite/test_areas/kernels.yaml
+15
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: vLLM IR Tests
6 + key: vllm-ir-tests
7 timeout_in_minutes: 10
8 device: h200_18gb
9 working_dir: "/vllm-workspace/"
@@ -14,6 +15,7 @@ steps:
15 - pytest -v -s tests/kernels/ir
16
17 - label: Kernels Core Operation Test
18 + key: kernels-core-operation-test
19 timeout_in_minutes: 75
20 source_file_dependencies:
21 - csrc/
@@ -23,6 +25,7 @@ steps:
25 - pytest -v -s kernels/core --ignore=kernels/core/test_minimax_reduce_rms.py kernels/test_concat_mla_q.py
26
27 - label: Kernels MiniMax Reduce RMS Test (2 GPUs)
28 + key: kernels-minimax-reduce-rms-test-2-gpus
29 timeout_in_minutes: 15
30 num_devices: 2
31 device: h100
@@ -36,6 +39,7 @@ steps:
39 - pytest -v -s kernels/core/test_minimax_reduce_rms.py
40
41 - label: Kernels Attention Test %N
42 + key: kernels-attention-test
43 timeout_in_minutes: 35
44 source_file_dependencies:
45 - csrc/attention/
@@ -49,6 +53,7 @@ steps:
53 parallelism: 2
54
55 - label: Kernels Quantization Test %N
56 + key: kernels-quantization-test
57 timeout_in_minutes: 90
58 source_file_dependencies:
59 - csrc/quantization/
@@ -59,6 +64,7 @@ steps:
64 parallelism: 2
65
66 - label: Kernels MoE Test %N
67 + key: kernels-moe-test
68 timeout_in_minutes: 25
69 source_file_dependencies:
70 - csrc/quantization/cutlass_w8a8/moe/
@@ -74,6 +80,7 @@ steps:
80 parallelism: 5
81
82 - label: Kernels Mamba Test
83 + key: kernels-mamba-test
84 timeout_in_minutes: 45
85 source_file_dependencies:
86 - csrc/mamba/
@@ -83,6 +90,7 @@ steps:
90 - pytest -v -s kernels/mamba
91
92 - label: Kernels DeepGEMM Test (H100)
93 + key: kernels-deepgemm-test-h100
94 timeout_in_minutes: 45
95 device: h100
96 num_devices: 1
@@ -104,6 +112,7 @@ steps:
112 - pytest -v -s quantization/test_cutlass_w4a16.py
113
114 - label: Kernels (B200)
115 + key: kernels-b200
116 timeout_in_minutes: 30
117 working_dir: "/vllm-workspace/"
118 device: b200
@@ -152,6 +161,7 @@ steps:
161 - pytest -v -s tests/models/quantization/test_nvfp4.py
162
163 - label: Kernels Helion Test
164 + key: kernels-helion-test
165 timeout_in_minutes: 30
166 device: h100
167 source_file_dependencies:
@@ -163,6 +173,7 @@ steps:
173
174
175 - label: Kernels FP8 MoE Test (1 H100)
176 + key: kernels-fp8-moe-test-1-h100
177 timeout_in_minutes: 90
178 device: h100
179 num_devices: 1
@@ -179,6 +190,7 @@ steps:
190 - pytest -v -s kernels/moe/test_triton_moe_ptpc_fp8.py
191
192 - label: Kernels FP8 MoE Test (2 H100s)
193 + key: kernels-fp8-moe-test-2-h100s
194 timeout_in_minutes: 90
195 device: h100
196 num_devices: 2
@@ -188,6 +200,7 @@ steps:
200 - pytest -v -s kernels/moe/test_deepep_moe.py
201
202 - label: Kernels Fp4 MoE Test (B200)
203 + key: kernels-fp4-moe-test-b200
204 timeout_in_minutes: 60
205 device: b200
206 num_devices: 1
@@ -200,6 +213,7 @@ steps:
213
214
215 - label: Kernels FusedMoE Layer Test (2 H100s)
216 + key: kernels-fusedmoe-layer-test-2-h100s
217 timeout_in_minutes: 90
218 device: h100
219 num_devices: 2
@@ -216,6 +230,7 @@ steps:
230
231
232 - label: Kernels FusedMoE Layer Test (2 B200s)
233 + key: kernels-fusedmoe-layer-test-2-b200s
234 timeout_in_minutes: 90
235 device: b200
236 num_devices: 2
.buildkite/test_areas/lm_eval.yaml
+11
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: LM Eval Small Models
6 + key: lm-eval-small-models
7 timeout_in_minutes: 75
8 source_file_dependencies:
9 - csrc/
@@ -24,6 +25,7 @@ steps:
25 # - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large.txt --tp-size=4
26
27 - label: LM Eval Large Models (4 GPUs)(H100)
28 + key: lm-eval-large-models-4-gpus-h100
29 device: h100
30 optional: true
31 num_devices: 4
@@ -36,6 +38,7 @@ steps:
38 - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large-hopper.txt --tp-size=4
39
40 - label: LM Eval Small Models (B200)
41 + key: lm-eval-small-models-b200
42 timeout_in_minutes: 120
43 device: b200
44 optional: true
@@ -46,6 +49,7 @@ steps:
49 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-blackwell.txt
50
51 - label: LM Eval Qwen3.5 Models (B200)
52 + key: lm-eval-qwen3-5-models-b200
53 timeout_in_minutes: 120
54 device: b200
55 optional: true
@@ -62,6 +66,7 @@ steps:
66 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-qwen35-blackwell.txt
67
68 - label: LM Eval Large Models (H200)
69 + key: lm-eval-large-models-h200
70 timeout_in_minutes: 60
71 device: h200
72 optional: true
@@ -70,6 +75,7 @@ steps:
75 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-h200.txt
76
77 - label: MoE Refactor Integration Test (H100 - TEMPORARY)
78 + key: moe-refactor-integration-test-h100-temporary
79 device: h100
80 optional: true
81 num_devices: 2
@@ -77,6 +83,7 @@ steps:
83 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/moe-refactor/config-h100.txt
84
85 - label: MoE Refactor Integration Test (B200 - TEMPORARY)
86 + key: moe-refactor-integration-test-b200-temporary
87 device: b200
88 optional: true
89 num_devices: 2
@@ -84,6 +91,7 @@ steps:
91 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/moe-refactor/config-b200.txt
92
93 - label: MoE Refactor Integration Test (B200 DP - TEMPORARY)
94 + key: moe-refactor-integration-test-b200-dp-temporary
95 device: b200
96 optional: true
97 num_devices: 2
@@ -92,6 +100,7 @@ steps:
100
101
102 - label: LM Eval TurboQuant KV Cache
103 + key: lm-eval-turboquant-kv-cache
104 timeout_in_minutes: 75
105 source_file_dependencies:
106 - vllm/model_executor/layers/quantization/turboquant/
@@ -102,6 +111,7 @@ steps:
111 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=evals/gsm8k/configs/models-turboquant.txt
112
113 - label: GPQA Eval (GPT-OSS) (H100)
114 + key: gpqa-eval-gpt-oss-h100
115 timeout_in_minutes: 120
116 device: h100
117 optional: true
@@ -115,6 +125,7 @@ steps:
125 - pytest -s -v evals/gpt_oss/test_gpqa_correctness.py --config-list-file=configs/models-h100.txt
126
127 - label: GPQA Eval (GPT-OSS) (B200)
128 + key: gpqa-eval-gpt-oss-b200
129 timeout_in_minutes: 120
130 device: b200
131 optional: true
.buildkite/test_areas/lora.yaml
+2
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: LoRA %N
6 + key: lora
7 timeout_in_minutes: 30
8 source_file_dependencies:
9 - vllm/lora
@@ -13,6 +14,7 @@ steps:
14
15
16 - label: LoRA TP (Distributed)
17 + key: lora-tp-distributed
18 timeout_in_minutes: 30
19 num_devices: 4
20 source_file_dependencies:
.buildkite/test_areas/misc.yaml
+14 -1
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: V1 Spec Decode
6 + key: v1-spec-decode
7 timeout_in_minutes: 30
8 source_file_dependencies:
9 - vllm/
@@ -18,6 +19,7 @@ steps:
19 - image-build-amd
20
21 - label: V1 Sample + Logits
22 + key: v1-sample-logits
23 timeout_in_minutes: 30
24 device: h200_18gb
25 source_file_dependencies:
@@ -41,6 +43,7 @@ steps:
43 - image-build-amd
44
45 - label: V1 Core + KV + Metrics
46 + key: v1-core-kv-metrics
47 timeout_in_minutes: 30
48 source_file_dependencies:
49 - vllm/
@@ -71,6 +74,7 @@ steps:
74 - image-build-amd
75
76 - label: V1 Others (CPU)
77 + key: v1-others-cpu
78 depends_on:
79 - image-build-cpu
80 source_file_dependencies:
@@ -86,6 +90,7 @@ steps:
90 - pytest -v -s -m 'cpu_test' v1/metrics
91
92 - label: Regression
93 + key: regression
94 timeout_in_minutes: 20
95 device: h200_18gb
96 source_file_dependencies:
@@ -97,6 +102,7 @@ steps:
102 working_dir: "/vllm-workspace/tests" # optional
103
104 - label: Examples
105 + key: examples
106 timeout_in_minutes: 45
107 working_dir: "/vllm-workspace/examples"
108 source_file_dependencies:
@@ -128,6 +134,7 @@ steps:
134 - python3 features/speculative_decoding/spec_decode_offline.py --test --method eagle3 --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 1536
135
136 - label: Metrics, Tracing (2 GPUs)
137 + key: metrics-tracing-2-gpus
138 timeout_in_minutes: 20
139 num_devices: 2
140 source_file_dependencies:
@@ -142,6 +149,7 @@ steps:
149 - pytest -v -s v1/tracing
150
151 - label: Python-only Installation
152 + key: python-only-installation
153 depends_on: ~
154 timeout_in_minutes: 20
155 source_file_dependencies:
@@ -151,6 +159,7 @@ steps:
159 - bash standalone_tests/python_only_compile.sh
160
161 - label: Async Engine, Inputs, Utils, Worker
162 + key: async-engine-inputs-utils-worker
163 timeout_in_minutes: 50
164 source_file_dependencies:
165 - vllm/
@@ -163,7 +172,8 @@ steps:
172 - pytest -v -s utils_
173
174 - label: Async Engine, Inputs, Utils, Worker, Config (CPU)
166 - depends_on:
175 + key: async-engine-inputs-utils-worker-config-cpu
176 + depends_on:
177 - image-build-cpu
178 timeout_in_minutes: 30
179 source_file_dependencies:
@@ -196,6 +206,7 @@ steps:
206 - pytest -v -s config
207
208 - label: Batch Invariance (H100)
209 + key: batch-invariance-h100
210 timeout_in_minutes: 30
211 device: h100
212 source_file_dependencies:
@@ -211,6 +222,7 @@ steps:
222 - VLLM_TEST_MODEL=Qwen/Qwen3-30B-A3B-Thinking-2507-FP8 pytest -v -s v1/determinism/test_batch_invariance.py::test_v1_generation_is_deterministic_across_batch_sizes_with_needle[FLASH_ATTN]
223
224 - label: Batch Invariance (B200)
225 + key: batch-invariance-b200
226 timeout_in_minutes: 30
227 device: b200
228 source_file_dependencies:
@@ -227,6 +239,7 @@ steps:
239 - pytest -v -s v1/determinism/test_nvfp4_batch_invariant.py
240
241 - label: Acceptance Length Test (Large Models) # optional
242 + key: acceptance-length-test-large-models
243 timeout_in_minutes: 25
244 gpu: h100
245 optional: true
.buildkite/test_areas/model_executor.yaml
+1
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Model Executor
6 + key: model-executor
7 timeout_in_minutes: 35
8 source_file_dependencies:
9 - vllm/engine/arg_utils.py
.buildkite/test_areas/model_runner_v2.yaml
+5
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Model Runner V2 Core Tests
6 + key: model-runner-v2-core-tests
7 timeout_in_minutes: 45
8 source_file_dependencies:
9 - vllm/v1/worker/gpu/
@@ -25,6 +26,7 @@ steps:
26 - 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"
27
28 - label: Model Runner V2 Examples
29 + key: model-runner-v2-examples
30 timeout_in_minutes: 45
31 working_dir: "/vllm-workspace/examples"
32 source_file_dependencies:
@@ -60,6 +62,7 @@ steps:
62 - python3 features/speculative_decoding/spec_decode_offline.py --test --method eagle3 --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 1536
63
64 - label: Model Runner V2 Distributed (2 GPUs)
65 + key: model-runner-v2-distributed-2-gpus
66 timeout_in_minutes: 45
67 working_dir: "/vllm-workspace/tests"
68 num_devices: 2
@@ -80,6 +83,7 @@ steps:
83 - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_eagle_dp.py
84
85 - label: Model Runner V2 Pipeline Parallelism (4 GPUs)
86 + key: model-runner-v2-pipeline-parallelism-4-gpus
87 timeout_in_minutes: 60
88 working_dir: "/vllm-workspace/tests"
89 num_devices: 4
@@ -95,6 +99,7 @@ steps:
99 - pytest -v -s distributed/test_pp_cudagraph.py -k "not ray"
100
101 - label: Model Runner V2 Spec Decode
102 + key: model-runner-v2-spec-decode
103 timeout_in_minutes: 30
104 working_dir: "/vllm-workspace/tests"
105 source_file_dependencies:
.buildkite/test_areas/models_basic.yaml
+6
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Basic Models Tests (Initialization)
6 + key: basic-models-tests-initialization
7 timeout_in_minutes: 45
8 torch_nightly: true
9 source_file_dependencies:
@@ -16,6 +17,7 @@ steps:
17 torch_nightly: {}
18
19 - label: Basic Models Tests (Extra Initialization) %N
20 + key: basic-models-tests-extra-initialization
21 timeout_in_minutes: 45
22 source_file_dependencies:
23 - vllm/model_executor/models/
@@ -31,6 +33,7 @@ steps:
33 torch_nightly: {}
34
35 - label: Basic Models Tests (Other)
36 + key: basic-models-tests-other
37 timeout_in_minutes: 45
38 source_file_dependencies:
39 - vllm/
@@ -47,6 +50,7 @@ steps:
50
51
52 - label: Basic Models Test (Other CPU) # 5min
53 + key: basic-models-test-other-cpu
54 depends_on:
55 - image-build-cpu
56 timeout_in_minutes: 10
@@ -59,6 +63,7 @@ steps:
63 - pytest -v -s models/test_utils.py models/test_vision.py
64
65 - label: Transformers Nightly Models
66 + key: transformers-nightly-models
67 working_dir: "/vllm-workspace/"
68 optional: true
69 soft_fail: true
@@ -74,6 +79,7 @@ steps:
79 - VLLM_WORKER_MULTIPROC_METHOD=spawn python3 examples/generate/multimodal/audio_language_offline.py --model-type whisper
80
81 - label: Transformers Backward Compatibility Models Test
82 + key: transformers-backward-compatibility-models-test
83 working_dir: "/vllm-workspace/"
84 optional: true
85 soft_fail: true
.buildkite/test_areas/models_distributed.yaml
+1
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Distributed Model Tests (2 GPUs)
6 + key: distributed-model-tests-2-gpus
7 timeout_in_minutes: 50
8 working_dir: "/vllm-workspace/tests"
9 num_devices: 2
.buildkite/test_areas/models_language.yaml
+7
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Language Models Tests (Standard)
6 + key: language-models-tests-standard
7 timeout_in_minutes: 25
8 source_file_dependencies:
9 - vllm/
@@ -15,6 +16,7 @@ steps:
16 torch_nightly: {}
17
18 - label: Language Models Tests (Extra Standard) %N
19 + key: language-models-tests-extra-standard
20 timeout_in_minutes: 45
21 source_file_dependencies:
22 - vllm/model_executor/models/
@@ -31,6 +33,7 @@ steps:
33 torch_nightly: {}
34
35 - label: Language Models Tests (Hybrid) %N
36 + key: language-models-tests-hybrid
37 timeout_in_minutes: 75
38 source_file_dependencies:
39 - vllm/
@@ -47,6 +50,7 @@ steps:
50 torch_nightly: {}
51
52 - label: Language Models Test (Extended Generation) # 80min
53 + key: language-models-test-extended-generation
54 timeout_in_minutes: 110
55 optional: true
56 source_file_dependencies:
@@ -69,6 +73,7 @@ steps:
73 - pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
74
75 - label: Language Models Test (PPL)
76 + key: language-models-test-ppl
77 timeout_in_minutes: 110
78 device: h200_18gb
79 optional: true
@@ -79,6 +84,7 @@ steps:
84 - pytest -v -s models/language/generation_ppl_test
85
86 - label: Language Models Test (Extended Pooling) # 36min
87 + key: language-models-test-extended-pooling
88 timeout_in_minutes: 50
89 optional: true
90 source_file_dependencies:
@@ -93,6 +99,7 @@ steps:
99 - image-build-amd
100
101 - label: Language Models Test (MTEB)
102 + key: language-models-test-mteb
103 timeout_in_minutes: 110
104 device: h200_18gb
105 optional: true
.buildkite/test_areas/models_multimodal.yaml
+12 -1
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: "Multi-Modal Models (Standard) 1: qwen2"
6 + key: multi-modal-models-standard-1-qwen2
7 timeout_in_minutes: 45
8 device: h200_18gb
9 source_file_dependencies:
@@ -19,6 +20,7 @@ steps:
20 - image-build-amd
21
22 - label: "Multi-Modal Models (Standard) 2: qwen3 + gemma"
23 + key: multi-modal-models-standard-2-qwen3-gemma
24 timeout_in_minutes: 45
25 device: h200_18gb
26 source_file_dependencies:
@@ -36,6 +38,7 @@ steps:
38 - image-build-amd
39
40 - label: "Multi-Modal Models (Standard) 3: llava + qwen2_vl"
41 + key: multi-modal-models-standard-3-llava-qwen2-vl
42 timeout_in_minutes: 45
43 source_file_dependencies:
44 - vllm/
@@ -51,6 +54,7 @@ steps:
54 - image-build-amd
55
56 - label: "Multi-Modal Models (Standard) 4: other + whisper"
57 + key: multi-modal-models-standard-4-other-whisper
58 timeout_in_minutes: 45
59 source_file_dependencies:
60 - vllm/
@@ -67,7 +71,8 @@ steps:
71 - image-build-amd
72
73 - label: Multi-Modal Processor (CPU)
70 - depends_on:
74 + key: multi-modal-processor-cpu
75 + depends_on:
76 - image-build-cpu
77 timeout_in_minutes: 60
78 source_file_dependencies:
@@ -80,6 +85,7 @@ steps:
85 - pytest -v -s models/multimodal/processing --ignore models/multimodal/processing/test_tensor_schema.py
86
87 - label: Multi-Modal Processor # 44min
88 + key: multi-modal-processor
89 timeout_in_minutes: 60
90 device: h200_18gb
91 source_file_dependencies:
@@ -91,6 +97,7 @@ steps:
97 - pytest -v -s models/multimodal/processing/test_tensor_schema.py
98
99 - label: Multi-Modal Accuracy Eval (Small Models) # 50min
100 + key: multi-modal-accuracy-eval-small-models
101 timeout_in_minutes: 70
102 working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
103 source_file_dependencies:
@@ -101,6 +108,7 @@ steps:
108 - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-mm-small.txt --tp-size=1
109
110 - label: Multi-Modal Models (Extended Generation 1)
111 + key: multi-modal-models-extended-generation-1
112 optional: true
113 source_file_dependencies:
114 - vllm/
@@ -117,6 +125,7 @@ steps:
125 - image-build-amd
126
127 - label: Multi-Modal Models (Extended Generation 2)
128 + key: multi-modal-models-extended-generation-2
129 optional: true
130 source_file_dependencies:
131 - vllm/
@@ -126,6 +135,7 @@ steps:
135 - pytest -v -s models/multimodal/generation/test_common.py -m 'split(group=0) and not core_model'
136
137 - label: Multi-Modal Models (Extended Generation 3)
138 + key: multi-modal-models-extended-generation-3
139 optional: true
140 source_file_dependencies:
141 - vllm/
@@ -135,6 +145,7 @@ steps:
145 - pytest -v -s models/multimodal/generation/test_common.py -m 'split(group=1) and not core_model'
146
147 - label: Multi-Modal Models (Extended Pooling)
148 + key: multi-modal-models-extended-pooling
149 optional: true
150 device: h200_18gb
151 source_file_dependencies:
.buildkite/test_areas/plugins.yaml
+1
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Plugin Tests (2 GPUs)
6 + key: plugin-tests-2-gpus
7 timeout_in_minutes: 60
8 working_dir: "/vllm-workspace/tests"
9 num_devices: 2
.buildkite/test_areas/pytorch.yaml
+6
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: PyTorch Compilation Unit Tests
6 + key: pytorch-compilation-unit-tests
7 timeout_in_minutes: 10
8 source_file_dependencies:
9 - vllm/
@@ -18,6 +19,7 @@ steps:
19 - "find compile/ -maxdepth 1 -name 'test_*.py' -print0 | xargs -0 -n1 -I{} pytest -s -v '{}'"
20
21 - label: PyTorch Compilation Unit Tests (H100)
22 + key: pytorch-compilation-unit-tests-h100
23 timeout_in_minutes: 30
24 device: h100
25 num_devices: 1
@@ -28,6 +30,7 @@ steps:
30 - "find compile/h100/ -name 'test_*.py' -print0 | xargs -0 -n1 -I{} pytest -s -v '{}'"
31
32 - label: PyTorch Compilation Passes Unit Tests
33 + key: pytorch-compilation-passes-unit-tests
34 timeout_in_minutes: 20
35 source_file_dependencies:
36 - vllm/
@@ -36,6 +39,7 @@ steps:
39 - pytest -s -v compile/passes --ignore compile/passes/distributed
40
41 - label: PyTorch Fullgraph Smoke Test
42 + key: pytorch-fullgraph-smoke-test
43 timeout_in_minutes: 35
44 source_file_dependencies:
45 - vllm/
@@ -48,6 +52,7 @@ steps:
52 - "find compile/fullgraph/ -name 'test_*.py' -not -name 'test_full_graph.py' -print0 | xargs -0 -n1 -I{} pytest -s -v '{}'"
53
54 - label: PyTorch Fullgraph
55 + key: pytorch-fullgraph
56 timeout_in_minutes: 30
57 device: h200_18gb
58 source_file_dependencies:
@@ -58,6 +63,7 @@ steps:
63 - pytest -v -s compile/fullgraph/test_full_graph.py -k 'not test_fp8_kv_scale_compile'
64
65 - label: Pytorch Nightly Dependency Override Check # 2min
66 + key: pytorch-nightly-dependency-override-check
67 # if this test fails, it means the nightly torch version is not compatible with some
68 # of the dependencies. Please check the error message and add the package to whitelist
69 # in /vllm/tools/pre_commit/generate_nightly_torch_test.py
.buildkite/test_areas/quantization.yaml
+3
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Quantization
6 + key: quantization
7 timeout_in_minutes: 90
8 source_file_dependencies:
9 - csrc/
@@ -21,6 +22,7 @@ steps:
22 - VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s quantization/ --ignore quantization/test_blackwell_moe.py
23
24 - label: Quantized MoE Test (B200)
25 + key: quantized-moe-test-b200
26 timeout_in_minutes: 60
27 working_dir: "/vllm-workspace/"
28 device: b200
@@ -38,6 +40,7 @@ steps:
40 - pytest -s -v tests/quantization/test_blackwell_moe.py
41
42 - label: Quantized Models Test
43 + key: quantized-models-test
44 timeout_in_minutes: 60
45 source_file_dependencies:
46 - vllm/model_executor/layers/quantization
.buildkite/test_areas/ray_compat.yaml
+1
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Ray Dependency Compatibility Check
6 + key: ray-dependency-compatibility-check
7 # Informational only — does not block the pipeline.
8 # If this fails, it means the PR introduces a dependency that
9 # conflicts with Ray's dependency constraints.
.buildkite/test_areas/samplers.yaml
+1
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Samplers Test
6 + key: samplers-test
7 timeout_in_minutes: 75
8 source_file_dependencies:
9 - vllm/model_executor/layers
.buildkite/test_areas/spec_decode.yaml
+8
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Spec Decode Eagle
6 + key: spec-decode-eagle
7 timeout_in_minutes: 30
8 device: h200_18gb
9 source_file_dependencies:
@@ -13,6 +14,7 @@ steps:
14 - pytest -v -s v1/e2e/spec_decode -k "eagle_correctness"
15
16 - label: Spec Decode Eagle Nightly B200
17 + key: spec-decode-eagle-nightly-b200
18 timeout_in_minutes: 30
19 device: b200
20 optional: true
@@ -24,6 +26,7 @@ steps:
26 - pytest -v -s v1/e2e/spec_decode -k "eagle_correctness"
27
28 - label: Spec Decode Speculators + MTP
29 + key: spec-decode-speculators-mtp
30 timeout_in_minutes: 30
31 device: h200_18gb
32 source_file_dependencies:
@@ -35,6 +38,7 @@ steps:
38 - pytest -v -s v1/e2e/spec_decode -k "speculators or mtp_correctness"
39
40 - label: Spec Decode Speculators + MTP Nightly B200
41 + key: spec-decode-speculators-mtp-nightly-b200
42 timeout_in_minutes: 30
43 device: b200
44 optional: true
@@ -47,6 +51,7 @@ steps:
51 - pytest -v -s v1/e2e/spec_decode -k "speculators or mtp_correctness"
52
53 - label: Spec Decode Ngram + Suffix
54 + key: spec-decode-ngram-suffix
55 timeout_in_minutes: 30
56 device: h200_18gb
57 source_file_dependencies:
@@ -57,6 +62,7 @@ steps:
62 - pytest -v -s v1/e2e/spec_decode -k "ngram or suffix"
63
64 - label: Spec Decode Draft Model
65 + key: spec-decode-draft-model
66 timeout_in_minutes: 30
67 device: h200_18gb
68 source_file_dependencies:
@@ -67,6 +73,7 @@ steps:
73 - pytest -v -s v1/e2e/spec_decode -k "draft_model or no_sync or batch_inference"
74
75 - label: Spec Decode Draft Model Nightly B200
76 + key: spec-decode-draft-model-nightly-b200
77 timeout_in_minutes: 30
78 device: b200
79 optional: true
@@ -78,6 +85,7 @@ steps:
85 - pytest -v -s v1/e2e/spec_decode -k "draft_model or no_sync or batch_inference"
86
87 - label: DFlash Speculators Correctness
88 + key: dflash-speculators-correctness
89 timeout_in_minutes: 30
90 device: h100
91 optional: true
.buildkite/test_areas/weight_loading.yaml
+1
@@ -3,6 +3,7 @@ depends_on:
3 - image-build
4 steps:
5 - label: Weight Loading Multiple GPU # 33min
6 + key: weight-loading-multiple-gpu
7 timeout_in_minutes: 45
8 working_dir: "/vllm-workspace/tests"
9 num_devices: 2