Open SeucheAchat9115 opened 6 months ago
https://github.com/MichalGeyer/pnp-diffusers/blob/main/pnp_utils.py
...
def register_conv_control_efficient(model, injection_schedule):
def conv_forward(self):
def forward(input_tensor, temb, scale=None): # https://github.com/omerbt/TokenFlow/issues/29 workaround with diffusers 0.21+
hidden_states = input_tensor
...
Thank you very much!
Hi, the Code does not work anymore with diffusers version e.g. 0.25.0 because the unet expects another input nowadays. Is there a quick fix how to handle such?