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

Access partial_rotary_factor from rope_parameters (#29966)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-12-04 18:42:49 +00:00
committed by GitHub
parent ece2825a29
commit e10c84e06a
21 changed files with 43 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/model_executor/models/llama.py
View File

@@ -149,8 +149,6 @@ class LlamaAttention(nn.Module):
if head_dim is None:
head_dim = self.hidden_size // self.total_num_heads
self.head_dim = head_dim
# Phi models introduced a partial_rotary_factor parameter in the config
self.partial_rotary_factor = getattr(config, "partial_rotary_factor", 1)
self.q_size = self.num_heads * self.head_dim
self.kv_size = self.num_kv_heads * self.head_dim
self.scaling = self.head_dim**-0.5
@@ -265,7 +263,6 @@ class LlamaAttention(nn.Module):
max_position=self.max_position_embeddings,
rope_parameters=getattr(config, "rope_parameters", None),
is_neox_style=is_neox_style,
partial_rotary_factor=self.partial_rotary_factor,
)