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

[Models]: Use MMEncoderAttention for MoonViT (#31738)

Browse Source 
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: Roger Wang <hey@rogerw.io>
Signed-off-by: h100 <h100@inferact.ai>
Co-authored-by: Roger Wang <hey@rogerw.io>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: h100 <h100@inferact.ai>
This commit is contained in:
Isotr0py
2026-01-06 16:00:25 +08:00
committed by GitHub
parent e9717801bd
commit 7101e0851f
2 changed files with 72 additions and 158 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/model_executor/models/kimi_vl.py
View File

@@ -325,7 +325,7 @@ class KimiVLForConditionalGeneration(nn.Module, SupportsMultiModal, SupportsPP):
self.hidden_size = config.text_config.hidden_size
self.vision_tower = MoonVitPretrainedModel(
config.vision_config,
self.use_data_parallel,
multimodal_config=model_config.multimodal_config,
prefix=maybe_prefix(prefix, "vision_tower"),
)