JLNTV_IPD/CogVideo
4
0
Fork 0
mirror of https://github.com/THUDM/CogVideo.git synced 2025-04-06 03:57:56 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix for CogVideoX-{2B,5B}

Browse Source 
When loading CogVideX-{2B,5B}, `patch_size_t` is None,
which results in the `prepare_rotary_position_embeddings` function.
This commit is contained in:
OleehyO 2024-12-12 14:18:55 +00:00
parent 36f1333788
commit 7b4c9db6d9
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/train_cogvideox_lora.py
View File

@ -1375,7 +1375,7 @@ def main(args):
num_frames=num_frames,
vae_scale_factor_spatial=vae_scale_factor_spatial,
patch_size=model_config.patch_size,
patch_size_t=model_config.patch_size_t,
patch_size_t=model_config.patch_size_t if model_config.patch_size_t is not None else 1,
attention_head_dim=model_config.attention_head_dim,
device=accelerator.device,
)