[INC][ARK] Direct Register Custom Op for ARK (#46361)

Signed-off-by: Zhenzhong1 <zhenzhong.xu@intel.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Zhenzhong Xu
2026-07-06 13:45:50 +08:00
committed by GitHub
co-authored by mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent e9cc1fd093
commit d9c1767cd4
6 changed files with 152 additions and 43 deletions
+5 -5
View File
@@ -61,7 +61,7 @@ MODELS = [
)
@pytest.mark.parametrize("model", MODELS)
def test_auto_round_model(vllm_runner, model):
with vllm_runner(model, enforce_eager=True) as llm:
with vllm_runner(model) as llm:
output = llm.generate_greedy(["The capital of France is"], max_tokens=8)
assert output
@@ -336,7 +336,7 @@ def test_wna16_xpu_prefers_ark_when_available(monkeypatch) -> None:
monkeypatch.setattr(current_platform, "is_xpu", lambda: True)
monkeypatch.setattr(current_platform, "is_cpu", lambda: False)
monkeypatch.setattr(
"vllm.model_executor.layers.quantization.inc.schemes.inc_wna16_linear.get_ark_state",
"vllm.model_executor.layers.quantization.inc.schemes.inc_ark_ops.get_ark_state",
lambda: (True, None, object(), DummyQuantLinear),
)
@@ -355,7 +355,7 @@ def test_wna16_xpu_falls_back_when_ark_unavailable(monkeypatch) -> None:
monkeypatch.setattr(current_platform, "is_xpu", lambda: True)
monkeypatch.setattr(current_platform, "is_cpu", lambda: False)
monkeypatch.setattr(
"vllm.model_executor.layers.quantization.inc.schemes.inc_wna16_linear.get_ark_state",
"vllm.model_executor.layers.quantization.inc.schemes.inc_ark_ops.get_ark_state",
lambda: (False, "missing", None, None),
)
@@ -377,7 +377,7 @@ def test_wna16_cpu_gptq_prefers_ark_when_available(monkeypatch) -> None:
monkeypatch.setattr(current_platform, "is_xpu", lambda: False)
monkeypatch.setattr(current_platform, "is_cpu", lambda: True)
monkeypatch.setattr(
"vllm.model_executor.layers.quantization.inc.schemes.inc_wna16_linear.get_ark_state",
"vllm.model_executor.layers.quantization.inc.schemes.inc_ark_ops.get_ark_state",
lambda: (True, None, object(), DummyQuantLinear),
)
@@ -398,7 +398,7 @@ def test_wna16_cpu_gptq_raises_when_ark_and_marlin_unavailable(
monkeypatch.setattr(current_platform, "is_xpu", lambda: False)
monkeypatch.setattr(current_platform, "is_cpu", lambda: True)
monkeypatch.setattr(
"vllm.model_executor.layers.quantization.inc.schemes.inc_wna16_linear.get_ark_state",
"vllm.model_executor.layers.quantization.inc.schemes.inc_ark_ops.get_ark_state",
lambda: (False, "missing", None, None),
)
monkeypatch.setattr(