[Attention] Refactor CUDA attention backend selection logic (#24794)
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com> Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com> Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
This commit is contained in:
co-authored by
Luka Govedič
parent
2e78150d24
commit
b30dfa03c5
@@ -21,7 +21,15 @@ else:
|
||||
logger = init_logger(__name__)
|
||||
|
||||
BlockSize = Literal[1, 8, 16, 32, 64, 128, 256]
|
||||
CacheDType = Literal["auto", "bfloat16", "fp8", "fp8_e4m3", "fp8_e5m2", "fp8_inc"]
|
||||
CacheDType = Literal[
|
||||
"auto",
|
||||
"bfloat16",
|
||||
"fp8",
|
||||
"fp8_e4m3",
|
||||
"fp8_e5m2",
|
||||
"fp8_inc",
|
||||
"fp8_ds_mla",
|
||||
]
|
||||
MambaDType = Literal["auto", "float32"]
|
||||
PrefixCachingHashAlgo = Literal["sha256", "sha256_cbor"]
|
||||
KVOffloadingBackend = Literal["native", "lmcache"]
|
||||
|
||||
Reference in New Issue
Block a user