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

2
vllm/model_executor/models/persimmon.py
View File

@@ -106,7 +106,6 @@ class PersimmonAttention(nn.Module):
self.num_heads = self.total_num_heads // tensor_parallel_world_size
self.head_dim = self.hidden_size // self.total_num_heads
self.max_position_embeddings = config.max_position_embeddings
self.partial_rotary_factor = config.partial_rotary_factor
self.is_causal = True
assert (self.head_dim * self.total_num_heads) == self.hidden_size
@@ -138,7 +137,6 @@ class PersimmonAttention(nn.Module):
rotary_dim=self.head_dim,
max_position=self.max_position_embeddings,
rope_parameters=config.rope_parameters,
partial_rotary_factor=self.partial_rotary_factor,
)
self.scaling = self.head_dim**-0.5
self.attn = Attention(