Aborted (core dumped) about real-time-voice-cloning HOT 10 CLOSED

It looks like the encoder failed to be loaded. What is your version of PyTorch? Have you managed to run other PyTorch models with your setup before?

Also, if you could try to find which of these lines is causing the problem, that would help.

from real-time-voice-cloning.

The models were trained and executed on pytorch 1.10.

At what point does the code I gave you fail?

from real-time-voice-cloning.

It looks like the encoder failed to be loaded. What is your version of PyTorch? Have you managed to run other PyTorch models with your setup before?

Also, if you could try to find which of these lines is causing the problem, that would help.

Thanks for your answer!
I tried with Pytorch 0.41 and Pytorch 1.1.0.
I change device with CPU on Line 39.
After I change lines, to CPU and can load model encode .
However, it not work.
I debug and discovered line About core dump

from real-time-voice-cloning.

Can you put this script at the root of the project and run it, try to see at what point it fails? You will need a torchvision version that matches your pytorch version.

from torchvision.models import resnet18
from encoder import inference as encoder
from pathlib import Path
import numpy as np
import torch


if __name__ == '__main__':
    # Test to see if your pytorch install works on CPU
    print("Forwarding resnet on CPU")
    inputs = np.zeros((4, 3, 224, 224), dtype=np.float32)
    model = resnet18()
    model.forward(torch.from_numpy(inputs))
    
    # Same but on GPU
    print("Forwarding resnet on GPU")
    model = model.cuda()
    model.forward(torch.from_numpy(inputs).cuda())

    # Test to see if the encoder works at all (on CPU) with a small input
    print("Forwarding encoder on CPU")
    weights_fpath = Path("encoder/saved_models/pretrained.pt")
    device = torch.device("cpu")
    encoder.load_model(weights_fpath, device)
    inputs = np.zeros((1, 100, 40), dtype=np.float32)
    encoder.embed_frames_batch(inputs)
    
    # Same but on GPU
    print("Forwarding encoder on GPU")
    device = torch.device("cuda")
    encoder.load_model(weights_fpath, device)
    encoder.embed_frames_batch(inputs)
    

from real-time-voice-cloning.

Do you think its because you have PyQt5 and not 4?

from real-time-voice-cloning.

Hey @hung96ad, any update?

from real-time-voice-cloning.

Hey @hung96ad, any update?

I tried new code but it not work.
Can you tell me your pytorch version?

from real-time-voice-cloning.

Do you think its because you have PyQt5 and not 4?

I don't think so. Because PyQt5 is frontend, it doesn't affect load models

from real-time-voice-cloning.

The models were trained and executed on pytorch 1.10.

At what point does the code I gave you fail?

Thank you!
I tried install pytorch and torchvision. It's work.
I think you should add requirements torchvision.

from real-time-voice-cloning.

I don't think torchvision has any role to play in the repo. It's a source for common datasets and models and I've only used it in the script I gave you. I suspect something else that you did fixed your setup. I'll still keep this issue in mind if someone else has the same problem. Glad this is fixed for you.

from real-time-voice-cloning.