[V0 Deprecation] Refactor kv cache from list to element (#37487)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
@@ -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
|
||||
|
||||
Reference in New Issue
Block a user