[CI] intel CI: add quantization and awq case for xpu (#46456)
Signed-off-by: wenjun.liu <wenjun.liu@intel.com> Signed-off-by: zengxian <xiangdong.zeng@intel.com> Co-authored-by: zengxian <xiangdong.zeng@intel.com>
wenjun liu committed
Jul 3, 2026 at 09:51 UTC
34bf7b45a0776d46a9516d8c0821e43f35b80cf5
3 files changed
+30
.buildkite/intel_jobs/quantization.yaml
new
+28
@@ -0,0 +1,28 @@
1
+group: Quantization
2
+depends_on:
3
+ - image-build-xpu
4
+steps:
5
+- label: Quantization
6
+ key: quantization
7
+ timeout_in_minutes: 30
8
+ env:
9
+ REGISTRY: "public.ecr.aws/q9t5s3a7"
10
+ REPO: "vllm-ci-test-repo"
11
+ VLLM_TEST_DEVICE: "xpu"
12
+ no_plugin: true
13
+ working_dir: "."
14
+ device: intel_gpu
15
+ agent_tags:
16
+ label: production
17
+ gpu: 1+
18
+ mem: 16+
19
+ source_file_dependencies:
20
+ - csrc/
21
+ - vllm/model_executor/layers/quantization
22
+ - tests/quantization
23
+ commands:
24
+ # - VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s quantization/ --ignore quantization/test_blackwell_moe.py
25
+ - >-
26
+ bash .buildkite/scripts/hardware_ci/run-intel-test.sh
27
+ 'VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s tests/quantization/test_per_token_kv_cache.py --deselect="tests/quantization/test_per_token_kv_cache.py::test_triton_unified_attention_per_token_head_scale[int4-16-128-num_heads0-seq_lens1]"'
28
+
.buildkite/intel_jobs/test-intel.yaml
+1
@@ -42,6 +42,7 @@ steps:
42
python3 examples/basic/offline_inference/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager --kv-cache-dtype fp8 &&
43
python3 examples/basic/offline_inference/generate.py --model nvidia/Llama-3.1-8B-Instruct-FP8 --block-size 64 --enforce-eager --quantization modelopt --kv-cache-dtype fp8 --attention-backend TRITON_ATTN --max-model-len 4096 &&
44
python3 examples/basic/offline_inference/generate.py --model superjob/Qwen3-4B-Instruct-2507-GPTQ-Int4 --block-size 64 --enforce-eager --max-model-len 8192 &&
45
+ python3 examples/basic/offline_inference/generate.py --model TheBloke/TinyLlama-1.1B-Chat-v0.3-AWQ --block-size 64 --enforce-eager &&
46
python3 examples/basic/offline_inference/generate.py --model ibm-research/PowerMoE-3b --block-size 64 --enforce-eager -tp 2 &&
47
python3 examples/basic/offline_inference/generate.py --model ibm-research/PowerMoE-3b --block-size 64 --enforce-eager -tp 2 --enable-expert-parallel &&
48
python3 examples/basic/offline_inference/generate.py --model superjob/Qwen3-4B-Instruct-2507-GPTQ-Int4 --max-model-len 8192 &&
.buildkite/scripts/hardware_ci/run-intel-ci-test.sh
+1
@@ -21,6 +21,7 @@ case "${test_suite}" in
21
python3 examples/basic/offline_inference/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager --kv-cache-dtype fp8
22
python3 examples/basic/offline_inference/generate.py --model nvidia/Llama-3.1-8B-Instruct-FP8 --block-size 64 --enforce-eager --quantization modelopt --kv-cache-dtype fp8 --attention-backend TRITON_ATTN --max-model-len 4096
23
python3 examples/basic/offline_inference/generate.py --model superjob/Qwen3-4B-Instruct-2507-GPTQ-Int4 --block-size 64 --enforce-eager --max-model-len 8192
24
+ python3 examples/basic/offline_inference/generate.py --model TheBloke/TinyLlama-1.1B-Chat-v0.3-AWQ --block-size 64 --enforce-eager
25
python3 examples/basic/offline_inference/generate.py --model ibm-research/PowerMoE-3b --block-size 64 --enforce-eager -tp 2
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