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
Support block size of 256 used by Intel HPU (#26883)
Signed-off-by: mandy-li <mandy.j.li@intel.com>
Mandy Li committed
Oct 16, 2025 at 12:10 UTC
ac3ed5a8159ed014d2fc975782c1276119f690f9
1 file changed
+1
-1
vllm/config/cache.py
+1
-1
@@ -19,7 +19,7 @@ else:
19
20
logger = init_logger(__name__)
21
22
-BlockSize = Literal[1, 8, 16, 32, 64, 128]
22
+BlockSize = Literal[1, 8, 16, 32, 64, 128, 256]
23
CacheDType = Literal["auto", "bfloat16", "fp8", "fp8_e4m3", "fp8_e5m2", "fp8_inc"]
24
MambaDType = Literal["auto", "float32"]
25
PrefixCachingHashAlgo = Literal["sha256", "sha256_cbor"]