[XPU][UT] skip fp8_per_channel test on XPU (#54861)
Signed-off-by: mayuyuace <qiming1.zhang@intel.com>
This commit is contained in:
@@ -202,8 +202,11 @@ def test_online_quantization(
|
||||
monkeypatch.setenv("VLLM_ROCM_USE_AITER", "1" if use_rocm_aiter else "0")
|
||||
rocm_aiter_ops.refresh_env_variables()
|
||||
|
||||
if current_platform.is_xpu() and quant_scheme == "fp8_per_block":
|
||||
pytest.skip("Skip test for online fp8_per_block on XPU platform.")
|
||||
if current_platform.is_xpu() and quant_scheme in (
|
||||
"fp8_per_block",
|
||||
"fp8_per_channel",
|
||||
):
|
||||
pytest.skip(f"Skip test for online {quant_scheme} on XPU platform.")
|
||||
|
||||
model_name = "ibm-granite/granite-3.0-1b-a400m-base"
|
||||
model, vllm_config = load_model_without_vllm_runner(
|
||||
|
||||
Reference in New Issue
Block a user