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

Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-11-19 18:06:36 +01:00
committed by GitHub
parent d44e9df7d4
commit a8b70304d6
104 changed files with 542 additions and 910 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/model_executor/models/gemma2.py
View File

@@ -107,7 +107,6 @@ class Gemma2Attention(nn.Module):
num_kv_heads: int,
head_dim: int,
max_position_embeddings: int,
rope_theta: float,
cache_config: CacheConfig | None = None,
quant_config: QuantizationConfig | None = None,
attn_logits_soft_cap: float | None = None,
@@ -134,7 +133,6 @@ class Gemma2Attention(nn.Module):
self.q_size = self.num_heads * self.head_dim
self.kv_size = self.num_kv_heads * self.head_dim
self.scaling = config.query_pre_attn_scalar**-0.5
self.rope_theta = rope_theta
self.qkv_proj = QKVParallelLinear(
hidden_size,
@@ -156,7 +154,7 @@ class Gemma2Attention(nn.Module):
self.head_dim,
rotary_dim=self.head_dim,
max_position=max_position_embeddings,
base=self.rope_theta,
rope_parameters=config.rope_parameters,
is_neox_style=True,
)
@@ -206,7 +204,6 @@ class Gemma2DecoderLayer(nn.Module):
num_kv_heads=config.num_key_value_heads,
head_dim=config.head_dim,
max_position_embeddings=config.max_position_embeddings,
rope_theta=config.rope_theta,
cache_config=cache_config,
quant_config=quant_config,
attn_logits_soft_cap=config.attn_logit_softcapping,