Open MianWang123 opened 1 day ago
Update!! For cogvideox-5b-controlnet-canny-v1 model, transformer_num_layers=18 and out_proj_dim=3072, which differs from cogvideox-2b-controlnet-canny-v1 model. But the base model has to be CogVideoX-2b for either of the above 2 checkpoints.
Hey :) 5b and 2b models have the same base model (2b), but 5b transformer have different inner dimention and because of this controlnet for 5b was created with additional output layer which is mapping 2b dim to 5b.
Hi, I was trying to finetune controlnet based on your checkpoint, but there's always a mismatch between the current model and your checkpoint. The checkpoint I used was "cogvideox-5b-controlnet-canny-v1", base model was "CogVideoX-5b", do you know where the problem could be?
![Uploading error.jpg…]()