Need local pre-trained models

[RendongZhang]

Dear authors,

Thank you for your great work. I am trying to run the training process. But it seems like a local stable-diffusion-v1-5 is required. Would you please share the download link for that ?

Thank you.

final using

    **ddim_scheduler_name_or_path: str = "/home/super/Desktop/8TDisk/huggingface/CompVis/stable-diffusion-v1-5"**
    ip2p_name_or_path: str = "/home/super/Desktop/8TDisk/huggingface/timbrooks/instruct-pix2pix"

[Lxiangyue]

Thanks for your attention. You could download it from here https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5/tree/main.

[RendongZhang]

Thank you for your response! I have a follow up question about the training process. How can i train from scratch ? I see all the training command you provide is starting from a checkpoint. I wonder how could i train a checkpoint like that.

If i simply delete the --start_checkpoint , a bug will be like this:
mouth_mask = gaussians.mouth_langsam_mask_dicts[motionID][time_idx, cam_idx,].cuda()
TypeError: 'NoneType' object is not subscriptable

What command should i use for training from scratch?

Sincerely thank you again!

[Lxiangyue]

If you don't have the mouth mask of that avatar, you cannot use the mouth mask loss. You could just delete the following three lines of code: lines 285-287 in the train.py. And it will be ok to run.

GaussianAvatar-Editor/train.py

Line 285 in 0817cd8

mouth_mask = gaussians.mouth_langsam_mask_dicts[motionID][time_idx, cam_idx,].cuda()