A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 12, 02:15 UTC
[CI/Build][AMD] Skip test on test_hybrid_attention_mamba_tensor_shapes on ROCm, requires FLASHINFER (#29995)
Signed-off-by: Randall Smith <ransmith@amd.com> Co-authored-by: Randall Smith <ransmith@amd.com>
rasmith committed
Dec 4, 2025 at 03:30 UTC
f2f4cea6ccaad20becb6f02e253ae673f8a249ae
1 file changed
+4
tests/v1/worker/test_gpu_model_runner.py
+4
@@ -761,6 +761,10 @@ def test_init_kv_cache_with_kv_sharing_valid():
761
assert kv_cache_config_after_init.kv_cache_groups[0].layer_names[1] == layer_1
762
763
764
+@pytest.mark.skipif(
765
+ current_platform.is_rocm(),
766
+ reason="Attention backend FLASHINFER is not supported on ROCm.",
767
+)
768
def test_hybrid_attention_mamba_tensor_shapes(monkeypatch):
769
"""
770
The GPU model runner creates different views into the