mirror of
https://git.datalinker.icu/vllm-project/vllm.git
synced 2026-05-08 20:45:45 +08:00
Merge e85f4a368652d94272bcbeeb3ee201978f7d0d00 into 254f6b986720c92ddf97fbb1a6a6465da8e87e29
This commit is contained in:
commit
e771d2d57a
@ -116,6 +116,7 @@ class RocmAttentionMetadataBuilder(AttentionMetadataBuilder[RocmAttentionMetadat
|
|||||||
slot_mapping = common_attn_metadata.slot_mapping
|
slot_mapping = common_attn_metadata.slot_mapping
|
||||||
|
|
||||||
use_cascade = common_prefix_len > 0
|
use_cascade = common_prefix_len > 0
|
||||||
|
prefix_scheduler_metadata = None
|
||||||
|
|
||||||
if use_cascade:
|
if use_cascade:
|
||||||
cu_prefix_query_lens = torch.tensor(
|
cu_prefix_query_lens = torch.tensor(
|
||||||
@ -130,7 +131,6 @@ class RocmAttentionMetadataBuilder(AttentionMetadataBuilder[RocmAttentionMetadat
|
|||||||
cu_prefix_query_lens = None
|
cu_prefix_query_lens = None
|
||||||
prefix_kv_lens = None
|
prefix_kv_lens = None
|
||||||
suffix_kv_lens = None
|
suffix_kv_lens = None
|
||||||
prefix_scheduler_metadata = None
|
|
||||||
|
|
||||||
attn_metadata = RocmAttentionMetadata(
|
attn_metadata = RocmAttentionMetadata(
|
||||||
num_actual_tokens=num_actual_tokens,
|
num_actual_tokens=num_actual_tokens,
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user