[XPU][CI]Add more cases in intel GPU CI and reorganize to align non-xpu part (#51630)
Signed-off-by: zengxian <xiangdong.zeng@intel.com> Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
xiangdong committed
Aug 21, 2026 at 17:31 UTC
36bad1b90cb2958b6de4abdb5a9249b136f39720
7 files changed
+175
-30
.buildkite/hardware_tests/intel_xpu_ci/test-intel.yaml
+87
@@ -58,6 +58,28 @@ steps:
58
bash .buildkite/scripts/hardware_ci/run-intel-test.sh
59
'bash .buildkite/scripts/hardware_ci/run-intel-ci-test.sh v1'
60
parallelism: 2
61
+ - label: "XPU W8A8 FP8 Linear Examples"
62
+ depends_on:
63
+ - image-build-xpu
64
+ timeout_in_minutes: 60
65
+ optional: true
66
+ device: intel_gpu
67
+ agent_tags:
68
+ label: production
69
+ gpu: 1+
70
+ mem: 24+
71
+ no_plugin: true
72
+ env:
73
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
74
+ REPO: "vllm-ci-test-repo"
75
+ VLLM_TEST_DEVICE: "xpu"
76
+ source_file_dependencies:
77
+ - .buildkite/hardware_tests/intel_xpu_ci/test-intel.yaml
78
+ - .buildkite/scripts/hardware_ci/run-intel-ci-test.sh
79
+ commands:
80
+ - >-
81
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
82
+ 'bash .buildkite/scripts/hardware_ci/run-intel-ci-test.sh w8a8-fp8-linear'
83
- label: "XPU server test"
84
depends_on:
85
- image-build-xpu
@@ -79,3 +101,68 @@ steps:
101
- >-
102
bash .buildkite/scripts/hardware_ci/run-intel-test.sh
103
'bash .buildkite/scripts/hardware_ci/run-intel-ci-test.sh server'
104
+ - label: "XPU quantization test"
105
+ depends_on:
106
+ - image-build-xpu
107
+ timeout_in_minutes: 60
108
+ optional: true
109
+ device: intel_gpu
110
+ agent_tags:
111
+ label: production
112
+ gpu: 1+
113
+ mem: 16+
114
+ no_plugin: true
115
+ env:
116
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
117
+ REPO: "vllm-ci-test-repo"
118
+ source_file_dependencies:
119
+ - .buildkite/hardware_tests/intel_xpu_ci/test-intel.yaml
120
+ - .buildkite/scripts/hardware_ci/run-intel-ci-test.sh
121
+ commands:
122
+ - >-
123
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
124
+ 'bash .buildkite/scripts/hardware_ci/run-intel-ci-test.sh quantization'
125
+ - label: "XPU compressed tensors FP8 test"
126
+ depends_on:
127
+ - image-build-xpu
128
+ timeout_in_minutes: 60
129
+ optional: true
130
+ device: intel_gpu
131
+ agent_tags:
132
+ label: production
133
+ gpu: 1+
134
+ mem: 16+
135
+ no_plugin: true
136
+ env:
137
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
138
+ REPO: "vllm-ci-test-repo"
139
+ VLLM_TEST_DEVICE: "xpu"
140
+ source_file_dependencies:
141
+ - .buildkite/hardware_tests/intel_xpu_ci/test-intel.yaml
142
+ - .buildkite/scripts/hardware_ci/run-intel-ci-test.sh
143
+ commands:
144
+ - >-
145
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
146
+ 'bash .buildkite/scripts/hardware_ci/run-intel-ci-test.sh compressed-tensors-fp8'
147
+ - label: "XPU Graph"
148
+ depends_on:
149
+ - image-build-xpu
150
+ timeout_in_minutes: 60
151
+ optional: true
152
+ device: intel_gpu
153
+ agent_tags:
154
+ label: production
155
+ gpu: 1+
156
+ mem: 32+
157
+ no_plugin: true
158
+ env:
159
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
160
+ REPO: "vllm-ci-test-repo"
161
+ VLLM_TEST_DEVICE: "xpu"
162
+ source_file_dependencies:
163
+ - .buildkite/hardware_tests/intel_xpu_ci/test-intel.yaml
164
+ - .buildkite/scripts/hardware_ci/run-intel-ci-test.sh
165
+ commands:
166
+ - >-
167
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
168
+ 'bash .buildkite/scripts/hardware_ci/run-intel-ci-test.sh graph'
.buildkite/intel_jobs/basic_correctness_intel.yaml
renamed
+2
-2
@@ -7,7 +7,7 @@ steps:
7
device: intel_gpu
8
agent_tags:
9
label: production
10
- gpu: 1+
10
+ gpu: 2+
11
mem: 16+
12
no_plugin: true
13
working_dir: "."
@@ -24,4 +24,4 @@ steps:
24
'cd tests &&
25
export VLLM_WORKER_MULTIPROC_METHOD=spawn &&
26
pytest -v -s basic_correctness/test_cpu_offload.py &&
27
- pytest -v -s basic_correctness/test_mem.py::test_end_to_end'
27
+ pytest -v -s basic_correctness/test_mem.py'
.buildkite/intel_jobs/expert_parallelism_intel.yaml
+2
-1
@@ -24,4 +24,5 @@ steps:
24
- >-
25
bash .buildkite/scripts/hardware_ci/run-intel-test.sh
26
'cd tests &&
27
- pytest -v -s distributed/test_eplb_algo.py'
27
+ pytest -v -s distributed/test_eplb_algo.py &&
28
+ pytest -v -s distributed/test_eplb_utils.py'
.buildkite/intel_jobs/lm_eval_intel.yaml
new
+28
@@ -0,0 +1,28 @@
1
+group: LM Eval
2
+depends_on:
3
+ - image-build-xpu
4
+steps:
5
+- label: "XPU GPQA Eval (GPT-OSS)"
6
+ key: xpu-gpqa-eval-gpt-oss
7
+ timeout_in_minutes: 60
8
+ device: intel_gpu
9
+ agent_tags:
10
+ label: production
11
+ gpu: 1+
12
+ mem: 24+
13
+ no_plugin: true
14
+ working_dir: "."
15
+ env:
16
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
17
+ REPO: "vllm-ci-test-repo"
18
+ VLLM_TEST_DEVICE: "xpu"
19
+ source_file_dependencies:
20
+ - csrc/
21
+ - vllm/model_executor/layers/quantization
22
+ - tests/evals/gpt_oss/
23
+ commands:
24
+ - >-
25
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
26
+ 'pip install "gpt-oss[eval]==0.0.5" &&
27
+ cd tests &&
28
+ pytest -s -v evals/gpt_oss/test_gpqa_correctness.py --config-list-file=configs/models-xpu.txt'
.buildkite/intel_jobs/models_multimodal_intel.yaml
+30
-2
@@ -29,12 +29,12 @@ steps:
29
30
- label: "Multi-Modal Models (Standard) 2: qwen3 + gemma"
31
key: multi-modal-models-standard-2-qwen3-gemma
32
- timeout_in_minutes: 70
32
+ timeout_in_minutes: 90
33
device: intel_gpu
34
agent_tags:
35
label: production
36
gpu: 1+
37
- mem: 16+
37
+ mem: 24+
38
no_plugin: true
39
working_dir: "."
40
env:
@@ -48,6 +48,8 @@ steps:
48
- >-
49
bash .buildkite/scripts/hardware_ci/run-intel-test.sh
50
'cd tests &&
51
+ pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "qwen3 or gemma" &&
52
+ pytest -v -s models/multimodal/generation/test_mm_prefix_lm.py -m core_model &&
53
pytest -v -s models/multimodal/generation/test_qwen2_5_vl.py -m core_model'
54
55
- label: "Multi-Modal Models (Standard) 3: llava + qwen2_vl"
@@ -125,3 +127,29 @@ steps:
127
pytest -v -s models/multimodal/processing/test_tensor_schema.py
128
--num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB'
129
parallelism: 4
130
+
131
+- label: Multi-Modal Models (Extended Generation 1)
132
+ key: multi-modal-models-extended-generation-1
133
+ timeout_in_minutes: 60
134
+ device: intel_gpu
135
+ agent_tags:
136
+ label: production
137
+ gpu: 1+
138
+ mem: 24+
139
+ no_plugin: true
140
+ working_dir: "."
141
+ env:
142
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
143
+ REPO: "vllm-ci-test-repo"
144
+ VLLM_TEST_DEVICE: "xpu"
145
+ optional: true
146
+ source_file_dependencies:
147
+ - vllm/
148
+ - "!vllm/distributed/kv_transfer/"
149
+ - tests/models/multimodal/generation
150
+ - tests/models/multimodal/test_mapping.py
151
+ commands:
152
+ - >-
153
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
154
+ 'cd tests &&
155
+ pytest -v -s models/multimodal/test_mapping.py'
.buildkite/intel_jobs/test-intel.yaml
-24
@@ -151,30 +151,6 @@ steps:
151
'cd tests &&
152
pytest -v -s quantization/test_auto_round.py &&
153
pytest -v -s quantization/test_online.py'
154
- - label: "XPU GPQA Eval (GPT-OSS)"
155
- depends_on:
156
- - image-build-xpu
157
- timeout_in_minutes: 60
158
- device: intel_gpu
159
- agent_tags:
160
- label: production
161
- gpu: 1+
162
- mem: 24+
163
- no_plugin: true
164
- env:
165
- REGISTRY: "public.ecr.aws/q9t5s3a7"
166
- REPO: "vllm-ci-test-repo"
167
- VLLM_TEST_DEVICE: "xpu"
168
- source_file_dependencies:
169
- - vllm/
170
- - tests/evals/gpt_oss/
171
- - .buildkite/intel_jobs/test-intel.yaml
172
- commands:
173
- - >-
174
- bash .buildkite/scripts/hardware_ci/run-intel-test.sh
175
- 'pip install "gpt-oss[eval]==0.0.5" &&
176
- cd tests &&
177
- pytest -s -v evals/gpt_oss/test_gpqa_correctness.py --config-list-file=configs/models-xpu.txt'
154
- label: "XPU compressed tensors FP8 test"
155
depends_on:
156
- image-build-xpu
.buildkite/scripts/hardware_ci/run-intel-ci-test.sh
+26
-1
@@ -5,7 +5,7 @@ set -euo pipefail
5
test_suite="${1:-}"
6
7
if [[ -z "${test_suite}" ]]; then
8
- echo "Usage: $0 <example|v1|server>" >&2
8
+ echo "Usage: $0 <example|w8a8-fp8-linear|v1|server|quantization|compressed-tensors-fp8|graph>" >&2
9
exit 1
10
fi
11
@@ -26,6 +26,13 @@ case "${test_suite}" in
26
python3 examples/basic/offline_inference/generate.py --model ibm-research/PowerMoE-3b --block-size 64 --enforce-eager -tp 2 --enable-expert-parallel
27
python3 examples/basic/offline_inference/generate.py --model superjob/Qwen3-4B-Instruct-2507-GPTQ-Int4 --max-model-len 8192
28
;;
29
+ w8a8-fp8-linear)
30
+ python3 examples/basic/offline_inference/generate.py --linear-backend xpu --model RedHatAI/Meta-Llama-3.1-8B-Instruct-FP8 --enforce-eager --max-model-len 4096
31
+ python3 examples/basic/offline_inference/generate.py --linear-backend torch --model RedHatAI/Meta-Llama-3.1-8B-Instruct-FP8 --enforce-eager --max-model-len 4096
32
+ python3 examples/basic/offline_inference/generate.py --linear-backend xpu --model neuralmagic/Llama-3.2-1B-Instruct-FP8-dynamic --enforce-eager --max-model-len 4096
33
+ python3 examples/basic/offline_inference/generate.py --linear-backend xpu --model meta-llama/Llama-3.2-1B-Instruct --quantization fp8 --enforce-eager --max-model-len 4096
34
+ python3 examples/basic/offline_inference/generate.py --linear-backend torch --model meta-llama/Llama-3.2-1B-Instruct --quantization fp8 --enforce-eager --max-model-len 4096
35
+ ;;
36
v1)
37
cd tests
38
@@ -46,6 +53,24 @@ case "${test_suite}" in
53
pytest -v -s entrypoints/multimodal/openai/chat_completion/test_audio_in_video.py
54
pytest -v -s benchmarks/test_serve_cli.py
55
;;
56
+ quantization)
57
+ cd tests
58
+
59
+ pytest -v -s quantization/test_auto_round.py
60
+ pytest -v -s quantization/test_online.py
61
+ ;;
62
+ compressed-tensors-fp8)
63
+ cd tests
64
+
65
+ pytest -v -s quantization/test_compressed_tensors.py::test_compressed_tensors_fp8
66
+ ;;
67
+ graph)
68
+ export VLLM_XPU_ENABLE_XPU_GRAPH=1
69
+
70
+ python3 examples/basic/offline_inference/generate.py --model Qwen/Qwen3-0.6B
71
+ python3 examples/basic/offline_inference/generate.py --model Qwen/Qwen3-0.6B --kv-cache-dtype fp8
72
+ python3 examples/basic/offline_inference/generate.py --model INCModel/Qwen3-30B-A3B-Instruct-2507-MXFP4-CT-AutoRound --max-model-len 4096
73
+ ;;
74
*)
75
echo "Unknown Intel test suite: ${test_suite}" >&2
76
exit 1