[Bugfix] Generalize KV block zeroing to AttentionSpec (#51749)
Signed-off-by: mgoin <mgoin64@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
co-authored by
mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent
513f83e7ec
commit
e3fe212eaf
@@ -29,7 +29,6 @@ from vllm.v1.core.kv_cache_utils import KVCacheBlockCopy
|
||||
from vllm.v1.kv_cache_interface import (
|
||||
AttentionSpec,
|
||||
EncoderOnlyAttentionSpec,
|
||||
FullAttentionSpec,
|
||||
KVCacheConfig,
|
||||
KVCacheGroupSpec,
|
||||
KVCacheSpec,
|
||||
@@ -147,7 +146,7 @@ class KVBlockZeroer:
|
||||
|
||||
for group in attn_groups_iter:
|
||||
spec = group.kv_cache_spec
|
||||
if not isinstance(spec, FullAttentionSpec):
|
||||
if not isinstance(spec, AttentionSpec):
|
||||
continue
|
||||
if group.kv_cache_group_id >= len(kernel_block_sizes):
|
||||
continue
|
||||
|
||||
Reference in New Issue
Block a user