bug 🐛

[johndpope]

https://github.com/eccv2024tcan/TCAN/blob/4a1d228dafbf55920b074d8b2e36fa7acc15ec26/train.py#L561

UPDATE

im seeing this error

File "/media/oem/12TB/TCAN/train.py", line 38, in from magicanimate.models.channel_fixer import ChannelFixer ModuleNotFoundError: No module named 'magicanimate.models.channel_fixer'

UPDATE i just comment that import line out and it works.

custom_attention_processor

if you can share these or how to create this file - openpose_dict/director_kim_0001_01.pkl parser.add_argument("--src_img", type=str, default="./inputs/webtoon_characters/source_images/director_kim_0001_01.png") parser.add_argument("--src_posedict", type=str, default="./inputs/webtoon_characters/openpose_dict/director_kim_0001_01.pkl") parser.add_argument("--driving_video", type=str, default="./inputs/webtoon_characters/driving_videos/orig_videos/running.mp4")

[ZivKidd]

hi, how do you find the control_v11p_sd15_openpose_RenamedForMA.pth?

I have tried with https://huggingface.co/lllyasviel/ControlNet-v1-1/blob/main/control_v11p_sd15_openpose.pth, and find 340 missing keys and 340 unexpected keys

assert len(m) == 42 or len(m) == 0, f"missing {len(m)} : there are additional missing layers except of initial layer (named controlnet_cond_embedding.) and zero conv"

[johndpope]

There’s a similar file slightly different name from hugging face weights from memory

[ZivKidd]

There’s a similar file slightly different name from hugging face weights from memory

thanks for your reply, could you write the url here? I have searched it for two hours, lol

[ZivKidd]

There’s a similar file slightly different name from hugging face weights from memory

https://huggingface.co/lllyasviel/control_v11p_sd15_openpose/blob/main/diffusion_pytorch_model.bin is it this one?

[johndpope]

hmmmm - idk now if i ever got it working will check again tomorrow - i switched in an openpose model from either here or realishuman maybe thing https://github.com/Boese0601/MagicDance

[ZivKidd]

hmmmm - idk now if i ever got it working will check again tomorrow - i switched in an openpose model from either here or realishuman maybe thing https://github.com/Boese0601/MagicDance

thx so much!