Skip to content

GitLab

Unverified Commit a0f8a796 authored by Avshalom Manevich's avatar Avshalom Manevich Committed by GitHub
Browse files

[fix] fix qwen image_embeds input (#21049) 


Signed-off-by: default avatarh-avsha <avshalom.manevich@hcompany.ai>
parent 18bdcf41
Hide whitespace changes
Inline Side-by-side
Showing with 2 additions and 2 deletions
vllm/model_executor/models/qwen2_5_vl.py
View file @ a0f8a796
...@@ -974,7 +974,7 @@ class Qwen2_5_VLForConditionalGeneration(nn.Module, SupportsMultiModal, ...@@ -974,7 +974,7 @@ class Qwen2_5_VLForConditionalGeneration(nn.Module, SupportsMultiModal,
grid_thw_list = grid_thw.tolist() grid_thw_list = grid_thw.tolist()
if image_input["type"] == "image_embeds": if image_input["type"] == "image_embeds":
image_embeds = image_input["image_embeds"] image_embeds = image_input["image_embeds"].type(self.visual.dtype)
else: else:
pixel_values = image_input["pixel_values"] pixel_values = image_input["pixel_values"]
image_embeds = self.visual(pixel_values, grid_thw=grid_thw_list) image_embeds = self.visual(pixel_values, grid_thw=grid_thw_list)
...@@ -994,7 +994,7 @@ class Qwen2_5_VLForConditionalGeneration(nn.Module, SupportsMultiModal, ...@@ -994,7 +994,7 @@ class Qwen2_5_VLForConditionalGeneration(nn.Module, SupportsMultiModal,
grid_thw_list = grid_thw.tolist() grid_thw_list = grid_thw.tolist()
if video_input["type"] == "video_embeds": if video_input["type"] == "video_embeds":
video_embeds = video_input["video_embeds"] video_embeds = video_input["video_embeds"].type(self.visual.dtype)
else: else:
pixel_values_videos = video_input["pixel_values_videos"] pixel_values_videos = video_input["pixel_values_videos"]
video_embeds = self.visual(pixel_values_videos, video_embeds = self.visual(pixel_values_videos,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment