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

[Minor][Models] Pass partial_rotary_factor parameter to rope (#17266)

Browse Source 
Signed-off-by: evian <eviantai@u.nus.edu>
Co-authored-by: evian <eviantai@u.nus.edu>
This commit is contained in:
Wanrui Dai
2025-04-28 12:28:59 +08:00
committed by GitHub
parent 8262a3e23b
commit 7fcc4223dc
3 changed files with 10 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/model_executor/models/stablelm.py
View File

@@ -104,9 +104,8 @@ class StablelmAttention(nn.Module):
1, self.total_num_key_value_heads // tp_size)
self.head_dim = self.hidden_size // self.total_num_heads
self.max_position_embeddings = config.max_position_embeddings
rope_pct = getattr(config, "rope_pct",
getattr(config, "partial_rotary_factor", 1))
self.rotary_ndims = int(self.head_dim * rope_pct)
self.partial_rotary_factor = getattr(
config, "rope_pct", getattr(config, "partial_rotary_factor", 1))
self.scaling = self.head_dim**-0.5
self.q_size = self.num_heads * self.head_dim
self.kv_size = self.num_key_value_heads * self.head_dim
@@ -130,9 +129,10 @@ class StablelmAttention(nn.Module):
prefix=f"{prefix}.o_proj")
self.rotary_emb = get_rope(
self.head_dim,
rotary_dim=self.rotary_ndims,
rotary_dim=self.head_dim,
max_position=self.config.max_position_embeddings,
base=self.config.rope_theta,
partial_rotary_factor=self.partial_rotary_factor,
)
self.attn = Attention(self.num_heads,
self.head_dim,