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 15, 11:00 UTC
fix broken test vllm:test_kernels - test_attention_selector.py::test_flash_attn (#17873)
Co-authored-by: Stephen Chen <tracelog@meta.com>
tracelogfb committed
May 9, 2025 at 19:46 UTC
246e3e0a36fdffdc22a31fbcb575d80dd59682d1
1 file changed
+3
-2
tests/kernels/attention/test_attention_selector.py
+3
-2
@@ -188,8 +188,9 @@ def test_flash_attn(monkeypatch: pytest.MonkeyPatch):
188
m.setenv(STR_BACKEND_ENV_VAR, STR_FLASH_ATTN_VAL)
189
190
# Unsupported CUDA arch
191
- monkeypatch.setattr(torch.cuda, "get_device_capability", lambda:
192
- (7, 5))
191
+ monkeypatch.setattr(torch.cuda,
192
+ "get_device_capability",
193
+ lambda _=None: (7, 5))
194
backend = get_attn_backend(16, torch.float16, None, 16, False)
195
assert backend.get_name() != STR_FLASH_ATTN_VAL
196