iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[CI/Build] Fix CPU platform pre-commit formatting (#52981)

Browse Source 
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
Misha Goin
2026-08-19 10:32:37 -07:00
committed by GitHub
co-authored by OpenAI Codex
parent 9b5f3454f2
commit f76d71d7ce
1 changed files with 5 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/platforms/cpu.py
+5 -2
View File
@@ -181,13 +181,16 @@ class CpuPlatform(Platform):
"otherwise the performance is not optimized."
)
# Accelerated GDN (AMX tiles or AVX-512BF16 VDPBF16PS) requires float32 SSM state.
# Accelerated GDN (AMX tiles or AVX-512BF16 VDPBF16PS) requires
# float32 SSM state.
if (
torch.cpu._is_avx512_bf16_supported()
and cache_config.mamba_ssm_cache_dtype != "float32"
):
cache_config.mamba_ssm_cache_dtype = "float32"
logger.warning("Reset SSM cache type to float32 for accelerated GDN mamba attention.")
logger.warning(
"Reset SSM cache type to float32 for accelerated GDN mamba attention."
)
# Lagecy setting
env_key = "VLLM_CPU_KVCACHE_SPACE"