| 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 | && VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s tests/quantization/test_auto_awq.py' |
| 29 |