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 14, 05:30 UTC
[Log] Use Debug Once for DeepGEMM E8M0 When not Enabled (#23858)
Wentao Ye committed
Aug 28, 2025 at 18:18 UTC
d3d2aad5a2a06b0ea22ae09cb0c6fb6912fa64d8
1 file changed
+1
-1
vllm/utils/deep_gemm.py
+1
-1
@@ -36,7 +36,7 @@ def is_deep_gemm_e8m0_used() -> bool:
36
"E8M0 scale on a Hopper or Blackwell-class GPU.
37
"""
38
if not is_deep_gemm_supported():
39
- logger.info_once(
39
+ logger.debug_once(
40
"DeepGEMM E8M0 disabled: DeepGEMM not supported on this system.")
41
return False
42