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

[Bugfix] Fix num video tokens calculation for Qwen2-VL (#13148)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-02-12 19:55:23 +08:00
committed by GitHub
parent f4d97e4fc2
commit 985b4a2b19
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/model_executor/models/qwen2_vl.py
View File

@ -800,7 +800,11 @@ class Qwen2VLProcessingInfo(BaseProcessingInfo):
preprocessed_size = ImageSize(width=image_width,
height=image_height)
grid_t = max(num_frames // temporal_patch_size, 1)
# NOTE: Frames are padded to be divisible by `temporal_patch_size`
# https://github.com/huggingface/transformers/blob/v4.48.3/src/transformers/models/qwen2_vl/image_processing_qwen2_vl.py#L294
padded_num_frames = num_frames + num_frames % temporal_patch_size
grid_t = max(padded_num_frames // temporal_patch_size, 1)
grid_h = preprocessed_size.height // patch_size
grid_w = preprocessed_size.width // patch_size