A high-throughput and memory-efficient inference and serving engine for LLMs
[XPU][CI] Add more test cases in Intel GPU CI (#43663)
Signed-off-by: zengxian <xiangdong.zeng@intel.com>
xiangdong committed
Jun 8, 2026 at 14:21 UTC
94fcdd007fe5cfa2a2f97cf062424cad543404df
4 files changed
+245
-1
.buildkite/intel_jobs/expert_parallelism_intel.yaml
new
+23
@@ -0,0 +1,23 @@
1
+group: Expert Parallelism
2
+depends_on:
3
+ - image-build-xpu
4
+steps:
5
+- label: EPLB Algorithm
6
+ key: eplb-algorithm
7
+ timeout_in_minutes: 45
8
+ device: intel_gpu
9
+ no_plugin: true
10
+ working_dir: "."
11
+ env:
12
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
13
+ REPO: "vllm-ci-test-repo"
14
+ VLLM_TEST_DEVICE: "xpu"
15
+ source_file_dependencies:
16
+ - vllm/distributed/eplb
17
+ - tests/distributed/test_eplb_algo.py
18
+ - tests/distributed/test_eplb_utils.py
19
+ commands:
20
+ - >-
21
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
22
+ 'cd tests &&
23
+ pytest -v -s distributed/test_eplb_algo.py'
.buildkite/intel_jobs/misc_intel.yaml
+109
-1
@@ -38,7 +38,17 @@ steps:
38
REPO: "vllm-ci-test-repo"
39
VLLM_TEST_DEVICE: "xpu"
40
source_file_dependencies:
41
- - vllm/
41
+ - vllm/config/
42
+ - vllm/distributed/
43
+ - vllm/engine/
44
+ - vllm/inputs/
45
+ - vllm/logger.py
46
+ - vllm/model_executor/
47
+ - vllm/platforms/
48
+ - vllm/sampling_params.py
49
+ - vllm/transformers_utils/
50
+ - vllm/utils/
51
+ - vllm/v1/
52
- tests/v1/sample
53
- tests/v1/logits_processors
54
- tests/v1/test_oracle.py
@@ -76,3 +86,101 @@ steps:
86
cd tests &&
87
pytest -v -s v1/kv_offload &&
88
pytest -v -s v1/kv_connector/unit/test_offloading_connector.py'
89
+
90
+- label: Regression
91
+ key: regression
92
+ timeout_in_minutes: 30
93
+ device: intel_gpu
94
+ no_plugin: true
95
+ working_dir: "."
96
+ env:
97
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
98
+ REPO: "vllm-ci-test-repo"
99
+ VLLM_TEST_DEVICE: "xpu"
100
+ source_file_dependencies:
101
+ - vllm/config/
102
+ - vllm/distributed/
103
+ - vllm/engine/
104
+ - vllm/inputs/
105
+ - vllm/model_executor/
106
+ - vllm/multimodal/
107
+ - vllm/platforms/
108
+ - vllm/sampling_params.py
109
+ - vllm/transformers_utils/
110
+ - vllm/utils/
111
+ - vllm/v1/
112
+ - tests/test_regression
113
+ commands:
114
+ - >-
115
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
116
+ 'pip install modelscope &&
117
+ cd tests &&
118
+ pytest -v -s test_regression.py'
119
+
120
+- label: Metrics, Tracing (2 GPUs)
121
+ key: metrics-tracing-2-gpus
122
+ timeout_in_minutes: 30
123
+ num_devices: 2
124
+ device: intel_gpu
125
+ no_plugin: true
126
+ working_dir: "."
127
+ env:
128
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
129
+ REPO: "vllm-ci-test-repo"
130
+ VLLM_TEST_DEVICE: "xpu"
131
+ source_file_dependencies:
132
+ - vllm/config/
133
+ - vllm/distributed/
134
+ - vllm/engine/
135
+ - vllm/inputs/
136
+ - vllm/model_executor/
137
+ - vllm/multimodal/
138
+ - vllm/platforms/
139
+ - vllm/sampling_params.py
140
+ - vllm/tracing/
141
+ - vllm/transformers_utils/
142
+ - vllm/utils/
143
+ - vllm/v1/
144
+ - tests/v1/tracing
145
+ commands:
146
+ - >-
147
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
148
+ 'pip install opentelemetry-sdk\>=1.26.0 opentelemetry-api\>=1.26.0 opentelemetry-exporter-otlp\>=1.26.0 opentelemetry-semantic-conventions-ai\>=0.4.1 &&
149
+ cd tests &&
150
+ pytest -v -s v1/tracing'
151
+
152
+- label: Async Engine, Inputs, Utils, Worker
153
+ key: async-engine-inputs-utils-worker
154
+ timeout_in_minutes: 30
155
+ device: intel_gpu
156
+ no_plugin: true
157
+ working_dir: "."
158
+ env:
159
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
160
+ REPO: "vllm-ci-test-repo"
161
+ VLLM_TEST_DEVICE: "xpu"
162
+ source_file_dependencies:
163
+ - vllm/assets/
164
+ - vllm/config/
165
+ - vllm/distributed/
166
+ - vllm/engine/
167
+ - vllm/inputs/
168
+ - vllm/model_executor/
169
+ - vllm/multimodal/
170
+ - vllm/platforms/
171
+ - vllm/sampling_params.py
172
+ - vllm/tokenizers/
173
+ - vllm/transformers_utils/
174
+ - vllm/utils/
175
+ - vllm/v1/
176
+ - tests/detokenizer
177
+ - tests/multimodal
178
+ - tests/utils_
179
+ commands:
180
+ - >-
181
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
182
+ 'cd tests &&
183
+ pip install av &&
184
+ pytest -v -s detokenizer &&
185
+ pytest -v -s -m "not cpu_test" ./multimodal &&
186
+ pytest -v -s utils_ --ignore=utils_/test_mem_utils.py'
.buildkite/intel_jobs/models_multimodal_intel.yaml
new
+111
@@ -0,0 +1,111 @@
1
+group: Models - Multimodal
2
+depends_on:
3
+ - image-build-xpu
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: intel_gpu
9
+ no_plugin: true
10
+ working_dir: "."
11
+ env:
12
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
13
+ REPO: "vllm-ci-test-repo"
14
+ VLLM_TEST_DEVICE: "xpu"
15
+ source_file_dependencies:
16
+ - vllm/
17
+ - tests/models/multimodal
18
+ commands:
19
+ - >-
20
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
21
+ 'pip install av git+https://github.com/TIGER-AI-Lab/Mantis.git &&
22
+ cd tests &&
23
+ pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "qwen2" &&
24
+ pytest -v -s models/multimodal/generation/test_ultravox.py -m core_model'
25
+
26
+- label: "Multi-Modal Models (Standard) 2: qwen3 + gemma"
27
+ key: multi-modal-models-standard-2-qwen3-gemma
28
+ timeout_in_minutes: 45
29
+ device: intel_gpu
30
+ no_plugin: true
31
+ working_dir: "."
32
+ env:
33
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
34
+ REPO: "vllm-ci-test-repo"
35
+ VLLM_TEST_DEVICE: "xpu"
36
+ source_file_dependencies:
37
+ - vllm/
38
+ - tests/models/multimodal
39
+ commands:
40
+ - >-
41
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
42
+ 'pip install git+https://github.com/TIGER-AI-Lab/Mantis.git &&
43
+ cd tests &&
44
+ pytest -v -s models/multimodal/generation/test_qwen2_5_vl.py -m core_model'
45
+
46
+- label: "Multi-Modal Models (Standard) 3: llava + qwen2_vl"
47
+ key: multi-modal-models-standard-3-llava-qwen2-vl
48
+ timeout_in_minutes: 45
49
+ device: intel_gpu
50
+ no_plugin: true
51
+ working_dir: "."
52
+ env:
53
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
54
+ REPO: "vllm-ci-test-repo"
55
+ VLLM_TEST_DEVICE: "xpu"
56
+ source_file_dependencies:
57
+ - vllm/
58
+ - tests/models/multimodal
59
+ commands:
60
+ - >-
61
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
62
+ 'pip install git+https://github.com/TIGER-AI-Lab/Mantis.git &&
63
+ cd tests &&
64
+ pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "not qwen2 and not qwen3 and not gemma" &&
65
+ pytest -v -s models/multimodal/generation/test_qwen2_vl.py -m core_model'
66
+
67
+- label: "Multi-Modal Models (Standard) 4: other + whisper"
68
+ key: multi-modal-models-standard-4-other-whisper
69
+ timeout_in_minutes: 45
70
+ device: intel_gpu
71
+ no_plugin: true
72
+ working_dir: "."
73
+ env:
74
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
75
+ REPO: "vllm-ci-test-repo"
76
+ VLLM_TEST_DEVICE: "xpu"
77
+ source_file_dependencies:
78
+ - vllm/
79
+ - tests/models/multimodal
80
+ commands:
81
+ - >-
82
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
83
+ 'pip install av git+https://github.com/TIGER-AI-Lab/Mantis.git &&
84
+ cd tests &&
85
+ pytest -v -s models/multimodal -m core_model --ignore models/multimodal/generation/test_common.py --ignore models/multimodal/generation/test_ultravox.py --ignore models/multimodal/generation/test_qwen2_5_vl.py --ignore models/multimodal/generation/test_qwen2_vl.py --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/generation/test_memory_leak.py --ignore models/multimodal/processing'
86
+
87
+- label: Multi-Modal Processor # 44min
88
+ key: multi-modal-processor
89
+ timeout_in_minutes: 45
90
+ device: intel_gpu
91
+ no_plugin: true
92
+ working_dir: "."
93
+ env:
94
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
95
+ REPO: "vllm-ci-test-repo"
96
+ VLLM_TEST_DEVICE: "xpu"
97
+ source_file_dependencies:
98
+ - vllm/
99
+ - tests/models/multimodal
100
+ - tests/models/registry.py
101
+ commands:
102
+ - >-
103
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
104
+ 'pip install av matplotlib ftfy git+https://github.com/TIGER-AI-Lab/Mantis.git &&
105
+ pip install open-clip-torch --no-deps &&
106
+ cd tests &&
107
+ pytest -v -s models/multimodal/processing/test_tensor_schema.py
108
+ --deselect "tests/models/multimodal/processing/test_tensor_schema.py::test_model_tensor_schema[mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4]"
109
+ --deselect "tests/models/multimodal/processing/test_tensor_schema.py::test_model_tensor_schema[Qwen/Qwen2.5-Omni-7B-AWQ]"
110
+ --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB'
111
+ parallelism: 4
.buildkite/scripts/hardware_ci/run-intel-test.sh
+2
@@ -372,6 +372,8 @@ export HF_TOKEN ZE_AFFINITY_MASK
372
--entrypoint='' \
373
-e HF_TOKEN \
374
-e ZE_AFFINITY_MASK \
375
+ -e BUILDKITE_PARALLEL_JOB \
376
+ -e BUILDKITE_PARALLEL_JOB_COUNT \
377
-e CMDS \
378
--name "${container_name}" \
379
"${IMAGE}" \