[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
+4
View File
@@ -61,6 +61,10 @@ MODELOPT_TO_VLLM_KV_CACHE_DTYPE_MAP = {
T = TypeVar("T")
def is_quantized_kv_cache(kv_cache_dtype: str) -> bool:
return kv_cache_dtype.startswith("fp8")
def is_strictly_contiguous(t: torch.Tensor) -> bool:
"""
Check if tensor is contiguous AND has no degenerate strides.