[Misc] Fix Qwen3-VL video_grid_thw typing (#25646)
Signed-off-by: Roger Wang <hey@rogerw.io>
This commit is contained in:
@@ -1249,7 +1249,7 @@ class Qwen3VLForConditionalGeneration(nn.Module, SupportsMultiModal,
|
||||
rope_type="rope_3d")
|
||||
else:
|
||||
video_embeds = self.visual(pixel_values_videos,
|
||||
grid_thw=grid_thw)
|
||||
grid_thw=grid_thw_list)
|
||||
|
||||
# Split concatenated embeddings for each video item.
|
||||
# Using prod on grid_thw_list instead of grid_thw.prod avoids CUDA sync
|
||||
|
||||
Reference in New Issue
Block a user