JLNTV_IPD/CogVideo
4
0
Fork 0
mirror of https://github.com/THUDM/CogVideo.git synced 2025-04-05 19:41:59 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix potential bug of i2v

Browse Source 
Image value is in [0, 255] and should be transformed into [-1, 1], similar to video.
This commit is contained in:
Zheng Guang Cong 2025-01-11 17:08:25 +08:00 committed by GitHub
parent 8f1829f1cd
commit 35383e2db3
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
finetune/datasets/i2v_dataset.py
View File

@ -153,7 +153,7 @@ class BaseI2VDataset(Dataset):
# Add image into the first frame.
# Note, **this operation maybe model-specific**, and maybe change in the future.
frames = torch.cat([image.unsqueeze(0), frames], dim=0)
frames = torch.cat([self.image_transform(image).unsqueeze(0), frames], dim=0)
# Convert to [B, C, F, H, W]
frames = frames.unsqueeze(0)