A high-throughput and memory-efficient inference and serving engine for LLMs
[XPU] skip UT test_with_ngram_gpu_spec_decoding (#44423)
Signed-off-by: Lai, Yejing <yejing.lai@intel.com>
Yejing Lai committed
Jun 15, 2026 at 16:46 UTC
9872921c5f5e733a4f46943562c0a31c9ff69493
1 file changed
+4
tests/v1/e2e/general/test_async_scheduling.py
+4
@@ -158,6 +158,10 @@ def test_with_eagle3_spec_decoding(sample_json_schema, monkeypatch: pytest.Monke
158
159
160
@pytest.mark.flaky(reruns=2, only_on=current_platform.is_rocm())
161
+@pytest.mark.skipif(
162
+ current_platform.is_xpu(),
163
+ reason=("XPU matmul/attention kernels are not batch-invariant"),
164
+)
165
def test_with_ngram_gpu_spec_decoding(monkeypatch: pytest.MonkeyPatch):
166
"""Test ngram_gpu speculative decoding with different configurations.
167