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

[Bugfix] Fix glm4.1v video inference issue (#22067)

Browse Source 
Signed-off-by: Isotr0py <2037008807@qq.com>
This commit is contained in:
Isotr0py
2025-08-02 00:33:30 +08:00
committed by GitHub
parent 326a1b001d
commit 3f8e952179
2 changed files with 53 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/model_executor/models/glm4_1v.py
View File

@@ -937,7 +937,7 @@ class Glm4vProcessingInfo(BaseProcessingInfo):
total_frames: int) -> list[int]:
video_processor = self.get_video_processor()
video_fps = metadata.get("fps", 2.0)
video_fps = metadata.get("fps", video_processor.fps)
meta_frames = metadata.get("total_num_frames", total_frames)
max_frame_idx = meta_frames - 1
duration = metadata.get("duration",
@@ -1120,11 +1120,7 @@ class Glm4vMultiModalProcessor(BaseMultiModalProcessor[Glm4vProcessingInfo]):
video_placeholder,
)
grid_t = len(video_outputs["video_grid_thw"])
_, grid_h, grid_w = video_outputs["video_grid_thw"][0]
grid_thw = torch.tensor([[grid_t, grid_h, grid_w]])
video_grid_thw_lst.append(grid_thw)
video_grid_thw_lst.append(video_outputs["video_grid_thw"])
pixel_values_videos_lst.append(
video_outputs["pixel_values_videos"])
video_outputs = dict(