gui_diff.py -> ValueError: [x] Unknown Model: DiffusionNew

[Sawaumi]

ran on python 3.8 (windows11) + cuda 11.8 + torch 2.0.0 + torchaudio 2.0.1 gui_diff.py would close itself on "start conversion", got it ran on pycharm and see what happened:

Traceback (most recent call last): File "[...file name...]\DDSP-SVC-4.0\gui_diff.py", line 580, in gui = GUI() File "[...file name...]\DDSP-SVC-4.0\gui_diff.py", line 229, in init self.launcher() # start File "[...file name...]\DDSP-SVC-4.0\gui_diff.py", line 311, in launcher self.event_handler() File "[...file name...]\DDSP-SVC-4.0\gui_diff.py", line 335, in event_handler self.start_vc() File "[...file name...]\DDSP-SVC-4.0\gui_diff.py", line 447, in start_vc self.svc_model.update_model(self.config.checkpoint_path) File "[...file name...]\DDSP-SVC-4.0\gui_diff.py", line 52, in update_model self.model, self.args = load_model(model_path, device=self.device) File "[...file name...]\DDSP-SVC-4.0\ddsp\vocoder.py", line 486, in load_model raise ValueError(f" [x] Unknown Model: {args.model.type}") ValueError: [x] Unknown Model: DiffusionNew

model name "DiffusionNew" comes from this https://github.com/yxlllc/DDSP-SVC/blob/master/train_diff.py while certainly shouldnt be read from this https://github.com/yxlllc/DDSP-SVC/blob/master/ddsp/vocoder.py

Does this mean I loaded the wrong model? model downloaded from https://github.com/yxlllc/DDSP-SVC/releases/download/4.0/opencpop+kiritan.zip and had .../DDSP-SVC-4.0/exp/diffusion-new-demo/model_200000.pt loaded (i suppose thats done by chosing it) chosen. Also the demo itself works fine (command lines) but gui does close itself.

[Sawaumi]

wait there wasnt a DiffusionNew type model was it? Is that model provided should I be loading that diffusion model?

Also: should i be training that another model seperately if I trained an 4.0 model by: python train_diff.py -c configs/diffusion-new.yaml?

[Sawaumi]

update: tried https://github.com/CNChTu/Diffusion-SVC and the sample combined model gui still kills itself by the time i start conversion

[yxlllc]

You need to load the diffusion model in the right hand side of the gui. The DDSP model in the left hand side can be empty if this is a 4.0 model.

[thedoctorbeat]

Hello! Mine also crashes when process starts.
Is there somewhere where I can find a step by step on how to use, train and all that??? All things I see are in Chinese, ahahah... Even the gui

I have ValueError: [x] Unknown Model: Diffusion

[KillauraHacks]

It seems Diffusion and DiffusionNew are not implemented and are missing from the vocoder.py