[1/N][Cleanup] Standardize on use of is_quantized_kv_cache (#38659)

Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
Matthew Bonanni
2026-04-01 04:08:01 +00:00
committed by GitHub
parent 7b01d97a22
commit 116f4be405
28 changed files with 90 additions and 75 deletions
+2 -2
View File
@@ -16,7 +16,7 @@ import torch
from vllm import envs
from vllm.logger import init_logger
from vllm.v1.attention.backend import is_quantized_kv_cache
from vllm.utils.torch_utils import is_quantized_kv_cache
from vllm.v1.attention.backends.registry import AttentionBackendEnum
from .interface import CpuArchEnum, Platform, PlatformEnum
@@ -183,7 +183,7 @@ class CpuPlatform(Platform):
"backend is not compatible with FP8 KV cache."
)
if cache_config.cache_dtype.startswith("fp8"):
if is_quantized_kv_cache(cache_config.cache_dtype):
logger.warning(
"CPU backend doesn't support KV cache quantization fallback to auto."
)