CUDA out of memory.

[Jamie-Cheung]

I try to train this code on 23090, but it still replys that CUDA out of memory. In your paper, you say that your model is trained on 2 Tesla V100 and 2 GeForce RTX 3090 GPUs spending about one day. I want to know whether this code can be trained on 23090. Than you for your contribution.

[diyiiyiii]

You can reduce the content size and style size in the training process. But I am not sure if it will affect the stylization effect.

[yuchen202]

can you test it now? can we communicate?

[Jamie-Cheung]

can you test it now? can we communicate?

I can't test successfully and i don't know why.

[yuchen202]

can you test it now? can we communicate?

I can't test successfully and i don't know why.

my wechat is c319186779.Can we communicate more?

[55jun]

Hallo, can you test successfully now? How to slove the error tuple object has no attribute cpu

[Jamie-Cheung]

Hallo, can you test successfully now? How to slove the error tuple object has no attribute cpu

for content_path in content_paths:
for style_path in style_paths:
print(content_path)

    content_tf1 = content_transform()
    content = content_tf(Image.open(content_path).convert("RGB"))

    h, w, c = np.shape(content)
    style_tf1 = style_transform(h, w)
    style = style_tf(Image.open(style_path).convert("RGB"))

    style = style.to(device).unsqueeze(0)
    content = content.to(device).unsqueeze(0)

    with torch.no_grad():
        output = network(content, style)
        # output = output.cuda()
        output = output

    output_name = '{:s}/{:s}_stylized_{:s}{:s}'.format(
        output_path, splitext(basename(content_path))[0],
        splitext(basename(style_path))[0], save_ext
    )
    save_image(output[0], output_name)

You can copy this code to replace corresponding code on test.py

[phhandong]

Hallo, can you test successfully now? How to slove the error tuple object has no attribute cpu

for content_path in content_paths: for style_path in style_paths: print(content_path)

    content_tf1 = content_transform()
    content = content_tf(Image.open(content_path).convert("RGB"))

    h, w, c = np.shape(content)
    style_tf1 = style_transform(h, w)
    style = style_tf(Image.open(style_path).convert("RGB"))

    style = style.to(device).unsqueeze(0)
    content = content.to(device).unsqueeze(0)

    with torch.no_grad():
        output = network(content, style)
        # output = output.cuda()
        output = output

    output_name = '{:s}/{:s}_stylized_{:s}{:s}'.format(
        output_path, splitext(basename(content_path))[0],
        splitext(basename(style_path))[0], save_ext
    )
    save_image(output[0], output_name)

You can copy this code to replace corresponding code on test.py

Did you successfully test it?I editted the code as yours but in vain, may you share your environment?

[Shenyve0101]

Hallo, can you test successfully now? How to slove the error tuple object has no attribute cpu

for content_path in content_paths: for style_path in style_paths: print(content_path)

    content_tf1 = content_transform()
    content = content_tf(Image.open(content_path).convert("RGB"))

    h, w, c = np.shape(content)
    style_tf1 = style_transform(h, w)
    style = style_tf(Image.open(style_path).convert("RGB"))

    style = style.to(device).unsqueeze(0)
    content = content.to(device).unsqueeze(0)

    with torch.no_grad():
        output = network(content, style)
        # output = output.cuda()
        output = output

    output_name = '{:s}/{:s}_stylized_{:s}{:s}'.format(
        output_path, splitext(basename(content_path))[0],
        splitext(basename(style_path))[0], save_ext
    )
    save_image(output[0], output_name)

You can copy this code to replace corresponding code on test.py

Did you successfully test it?I editted the code as yours but in vain, may you share your environment?

Maybe you can try the following code：

        style = style.to(device).unsqueeze(0)
        content = content.to(device).unsqueeze(0)

        #with torch.no_grad():
        #    output= network(content,style)       
        #output = output.cpu()

        with torch.no_grad():
            output = network(content, style)
        output = output[0].cpu()
                
        output_name = '{:s}/{:s}_stylized_{:s}{:s}'.format(
            output_path, splitext(basename(content_path))[0],
            splitext(basename(style_path))[0], save_ext
        )
 
        save_image(output, output_name)

[phhandong]

Hallo, can you test successfully now? How to slove the error tuple object has no attribute cpu

for content_path in content_paths: for style_path in style_paths: print(content_path)

    content_tf1 = content_transform()
    content = content_tf(Image.open(content_path).convert("RGB"))

    h, w, c = np.shape(content)
    style_tf1 = style_transform(h, w)
    style = style_tf(Image.open(style_path).convert("RGB"))

    style = style.to(device).unsqueeze(0)
    content = content.to(device).unsqueeze(0)

    with torch.no_grad():
        output = network(content, style)
        # output = output.cuda()
        output = output

    output_name = '{:s}/{:s}_stylized_{:s}{:s}'.format(
        output_path, splitext(basename(content_path))[0],
        splitext(basename(style_path))[0], save_ext
    )
    save_image(output[0], output_name)

You can copy this code to replace corresponding code on test.py

Did you successfully test it?I editted the code as yours but in vain, may you share your environment?

Maybe you can try the following code：

        style = style.to(device).unsqueeze(0)
        content = content.to(device).unsqueeze(0)

        #with torch.no_grad():
        #    output= network(content,style)       
        #output = output.cpu()

        with torch.no_grad():
            output = network(content, style)
        output = output[0].cpu()
                
        output_name = '{:s}/{:s}_stylized_{:s}{:s}'.format(
            output_path, splitext(basename(content_path))[0],
            splitext(basename(style_path))[0], save_ext
        )
 
        save_image(output, output_name)

Thank u so much！It works perfectly，and I found a mistake in the version of cuda，the code can only run on the cuda 10.x or lower.

[Link-32]

Have you solved your problem? I tried to run this code in RTX3080 with 10G memory and got "CUDA OUT OF MEMORY". I can only run this code when the batch size is 1. Once the batch size is greater than 1, even if iters is 1, it will report an error.

[yuchen202]

收到了，谢谢