xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-13 20:04:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Minor] Zero-initialize attn output buffer (#19784)

Browse Source 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Woosuk Kwon 2025-06-17 23:59:27 -07:00 committed by GitHub
parent 19a53b2783
commit f04d604567
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/attention/layer.py
View File

@ -209,7 +209,7 @@ class Attention(nn.Module):
if self.use_output: if self.use_output:
output_shape = (output_shape output_shape = (output_shape
if output_shape is not None else query.shape) if output_shape is not None else query.shape)
output = torch.empty(output_shape, output = torch.zeros(output_shape,
dtype=query.dtype, dtype=query.dtype,
device=query.device) device=query.device)
hidden_size = output_shape[-1] hidden_size = output_shape[-1]