[Bugfix] Fix speculative decoding for short_conv (LFM2) models (#50272)
Signed-off-by: jens.luecke <jens.luecke@liquid.ai> Co-authored-by: Tonoken3 <sakamakismile@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
co-authored by
Tonoken3
mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent
704f12aa1e
commit
da329cc303
@@ -148,6 +148,7 @@ from vllm.v1.attention.backends.linear_attn import (
|
||||
BailingLinearAttentionMetadataBuilder,
|
||||
)
|
||||
from vllm.v1.attention.backends.mamba2_attn import Mamba2AttentionMetadataBuilder
|
||||
from vllm.v1.attention.backends.short_conv_attn import ShortConvAttentionMetadataBuilder
|
||||
from vllm.v1.attention.backends.utils import (
|
||||
NULL_BLOCK_ID,
|
||||
create_fast_prefill_custom_backend,
|
||||
@@ -2572,6 +2573,7 @@ class GPUModelRunner(
|
||||
Mamba2AttentionMetadataBuilder,
|
||||
GDNAttentionMetadataBuilder,
|
||||
BailingLinearAttentionMetadataBuilder,
|
||||
ShortConvAttentionMetadataBuilder,
|
||||
),
|
||||
):
|
||||
assert ubid is None, (
|
||||
|
||||
Reference in New Issue
Block a user