frozenleaves/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-10-20 14:53:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Llama4] Update attn_temperature_tuning (#19997)

Browse Source 
Signed-off-by: Brayden Zhong <b8zhong@uwaterloo.ca>
This commit is contained in:
Brayden Zhong
2025-06-24 22:42:53 -04:00
committed by GitHub
parent 0d06b533a0
commit 1afa9948f5
1 changed files with 1 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/model_executor/models/llama4.py
View File

@ -148,9 +148,8 @@ class Llama4Attention(nn.Module):
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
# TODO: attn_temperature_tuning should be a bool in huggingface
self.attn_temperature_tuning = self.nope and \
config.attn_temperature_tuning > 0
config.attn_temperature_tuning
self.floor_scale = getattr(config, "floor_scale", 8192.0)
self.attn_scale = getattr(config, "attn_scale", 0.1)