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

[Bugfix][Qwen3VL] fix deepstack in qwen3vl (#26626)

Browse Source 
Signed-off-by: liuye.hj <liuye.hj@alibaba-inc.com>
Signed-off-by: JJJYmmm <92386084+JJJYmmm@users.noreply.github.com>
Co-authored-by: liuye.hj <liuye.hj@alibaba-inc.com>
This commit is contained in:
JJJYmmm
2025-10-11 20:58:33 +08:00
committed by GitHub
parent a25f2adee9
commit 9d6cff3ede
1 changed files with 0 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/model_executor/models/qwen3_vl.py
View File

@@ -1702,12 +1702,6 @@ class Qwen3VLForConditionalGeneration(
)
if deepstack_input_embeds is not None:
deepstack_input_embeds = (
torch.zeros_like(inputs_embeds)
.unsqueeze(0)
.repeat(self.deepstack_num_level, 1, 1)
.contiguous()
)
self._set_deepstack_input_embeds(deepstack_input_embeds)
return inputs_embeds