output with shape [1, 28, 28] doesn't match the broadcast shape [3, 28, 28]

[ywc1026]

When I run GAN code, I got a runtime error output with shape [1, 28, 28] doesn't match the broadcast shape [3, 28, 28] how fix it

[yuanzhoulvpi2017]

Maybe your picture is a gray image, you should use a color picture

[ywc1026]

Maybe your picture is a gray image, you should use a color picture 都是调用MNIST那个接口，从哪里可以下载到彩色的呢？

[ywc1026]

之前改过了，transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize(mean=0.5, std=0.5)]) 又报错 too many indices for tensor of dimension 0 我把后面那个 for i, (images, _) in enumerate(data_loader): 改成 for i, images in enumerate(data_loader): 也不对

[ywc1026]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

[QingyuGuo]

感谢😂

[laohur]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

改成楼上确实好用

[VinayMatcha]

If you are still having the problem please use this code instead of above in place of tranform. transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,)) ])

[ladycatusa]

@VinayMatcha transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

This is not work for me ? I still get an error on this line .I think the parentheses is wrong

[ahlusar1989]

@ladycatusa The snippet provided by @VinayMatcha does indeed correctly produce the correct output shape. I would recommend referencing this issue: https://github.com/fungtion/DANN/issues/8 for as to why this occurring.

[Loerbri]

I just ran into the same error message in a completely unrelated context, and changing the version of torchvision to 0.2.1 fixed it for me. Maybe this helps :)

[bitiniuer]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])]) 好使！感谢！请问大家都在做udacity吗 有没有学习群之类的？

[RenMao1314]

Let me clarify, if the img has three channels, you should have three number for mean, for example, img is RGB, mean is [0.5, 0.5, 0.5], the normalize result is R 0.5, G 0.5, B 0.5. If img is grey type that only one channel, so mean should be [0.5], the normalize result is R 0.5

[windyoo7]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

改成这样之后，又报错了 IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number 定位在 d_losses[epoch] = d_losses[epoch](i/(i+1.)) + d_loss.data[0](1./(i+1.)) how to fix it? thx

[RenMao1314]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

改成这样之后，又报错了 IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number 定位在 d_losses[epoch] = dlosses[epoch](i/(i+1.)) + dloss.data[0](1./(i+1.)) how to fix it? thx

If you're using pytorch>=1.0 or not? In pytorch 1.0, loss.item() replace loss.data[0] but it would just show user warning and why you got an error? Maybe a further issue.

Anyway please change your code firstly d_losses[epoch] = d_losses [ epoch ] (i/(i+1.)) + d_loss.item()(1./(i+1.))

if any issue happen please show more detail log to us.

[windyoo7]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

改成这样之后，又报错了 IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number 定位在 d_losses[epoch] = d_lossesepoch + d_loss.data0 how to fix it? thx

If you're using pytorch>=1.0 or not? In pytorch 1.0, loss.item() replace loss.data[0] but it would just show user warning and why you got an error? Maybe a further issue. Anyway please change your code firstly d_losses[epoch] = d_losses [ epoch ] (i/(i+1.)) + d_loss.item()(1./(i+1.)) if any issue happen please show more detail log to us.

yes. I'm using pytorch 1.0. you means that I should change the code d_losses[epoch] = d_lossesepoch + d_loss.data0 to be code below d_losses[epoch] = d_lossesepoch + d_loss.item()(1./(i+1.)) ? thx you very much! I will try it .

[windyoo7]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

改成这样之后，又报错了 IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number 定位在 d_losses[epoch] = d_lossesepoch + d_loss.data0 how to fix it? thx

If you're using pytorch>=1.0 or not? In pytorch 1.0, loss.item() replace loss.data[0] but it would just show user warning and why you got an error? Maybe a further issue. Anyway please change your code firstly d_losses[epoch] = d_losses [ epoch ] (i/(i+1.)) + d_loss.item()(1./(i+1.)) if any issue happen please show more detail log to us.

nice! it works! I replace the ".data[0]" with ".item()" and the code begins to work.

[ShuuTsubaki]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

改成这样之后，又报错了 IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number 定位在 d_losses[epoch] = dlosses[epoch](i/(i+1.)) + dloss.data[0](1./(i+1.)) how to fix it? thx

If you're using pytorch>=1.0 or not? In pytorch 1.0, loss.item() replace loss.data[0] but it would just show user warning and why you got an error? Maybe a further issue.

Anyway please change your code firstly d_losses[epoch] = d_losses [ epoch ] (i/(i+1.)) + d_loss.item()(1./(i+1.))

if any issue happen please show more detail log to us.

I got following error:RuntimeError: Given groups=1, weight of size 16 3 3 3, expected input[128, 1, 28, 28] to have 3 channels, but got 1 channels instead

[swg209]

@ShuuTsubaki I also encounter to the problem. Do you find the way to fit it?

[mvivekc]

Downgrading torch and torchvision to 0.2.0 and 0.2.1 solved this issue for me.

[israrbacha]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

改成这样之后，又报错了 IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number 定位在 d_losses[epoch] = dlosses[epoch](i/(i+1.)) + dloss.data[0](1./(i+1.)) how to fix it? thx

just remove d._loss.data[0] and write d_loss.data it works for me.

[immu0001]

This Fixed the error for me:

transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,)) ])

[Hongxueting]

transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])]) 为什么我还是不行

[16xder]

I just ran into the same error message in a completely unrelated context, and changing the version of torchvision to 0.2.1 fixed it for me. Maybe this helps :)

it is useful!

[prateek0635]

transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,)) ]) Try this I was also facing the same problem but now its done.

[Mehmetnurbostan]

its working i wrote in Normalize ((0.5) ,(0.5)) as func not as matrix transform=transforms.Compose([transforms.ToTensor(), transforms.Normalize([0.5], [0.5])]) and transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,)) ]) it works also

[sourcecode369]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

worked for me. Thanks.

[zhunipingan]

这样改就好了 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])

改成这样之后，又报错了 IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number 定位在 d_losses[epoch] = dlosses[epoch](i/(i+1.)) + dloss.data[0](1./(i+1.)) how to fix it? thx

If you're using pytorch>=1.0 or not? In pytorch 1.0, loss.item() replace loss.data[0] but it would just show user warning and why you got an error? Maybe a further issue. Anyway please change your code firstly d_losses[epoch] = d_losses [ epoch ] (i/(i+1.)) + d_loss.item()(1./(i+1.)) if any issue happen please show more detail log to us.

I got following error:RuntimeError: Given groups=1, weight of size 16 3 3 3, expected input[128, 1, 28, 28] to have 3 channels, but got 1 channels instead

Do you meet the problem that the d_loss will near to 0 gradually, which is not what we expect. who know how to fix it.

[chintanckg]

If you are still having the problem please use this code instead of above in place of tranform. transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,)) ])

Can you please explain in short? What went wrong and how did this work?

[liuguicen]

the model have three channels， only change the data loading way is useless，how to change the image to three channels?

[syedjameel]

I just transformed all the images to Grayscale and boom, it worked like charm.

This is my code: transform = transforms.Compose([transforms.Grayscale(), transforms.Resize((28,28)), transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,)), ])