biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[V0 Deprecation] Refactor kv cache from list to element (#37487)

Browse Source 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wentao Ye
2026-03-23 23:10:11 -04:00
committed by GitHub
parent de99d91ece
commit c59a132f96
27 changed files with 70 additions and 85 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/model_executor/layers/mamba/short_conv.py
View File

@@ -117,7 +117,7 @@ class ShortConv(MambaBase, CustomOp):
assert isinstance(attn_metadata, dict)
attn_metadata = attn_metadata[self.prefix]
assert isinstance(attn_metadata, ShortConvAttentionMetadata)
self_kv_cache = self.kv_cache[0]
self_kv_cache = self.kv_cache
conv_state = self_kv_cache[0].transpose(-1, -2)
state_indices_tensor_p = attn_metadata.state_indices_tensor_p
state_indices_tensor_d = attn_metadata.state_indices_tensor_d