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

Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wentao Ye
2026-03-23 20:10:11 -07:00
committed by GitHub
parent de99d91ece
commit c59a132f96
27 changed files with 70 additions and 85 deletions
+2 -2
View File
@@ -858,7 +858,7 @@ class Qwen3NextGatedDeltaNet(nn.Module, MambaBase):
non_spec_token_indx = attn_metadata.non_spec_token_indx
spec_state_indices_tensor = attn_metadata.spec_state_indices_tensor # noqa: E501
non_spec_state_indices_tensor = attn_metadata.non_spec_state_indices_tensor # noqa: E501
self_kv_cache = self.kv_cache[0]
self_kv_cache = self.kv_cache
conv_state = self_kv_cache[0].transpose(-1, -2)
ssm_state = self_kv_cache[1]
num_actual_tokens = attn_metadata.num_actual_tokens
@@ -1046,7 +1046,7 @@ class Qwen3NextGatedDeltaNet(nn.Module, MambaBase):
Core attention computation with a packed non-spec decode fast path.
"""
non_spec_state_indices_tensor = attn_metadata.non_spec_state_indices_tensor # noqa: E501
self_kv_cache = self.kv_cache[0]
self_kv_cache = self.kv_cache
conv_state = self_kv_cache[0].transpose(-1, -2)
ssm_state = self_kv_cache[1]
num_actual_tokens = attn_metadata.num_actual_tokens