About nn.Functional.upsample function

I want to resize my image tensor using the function as:
torch.nn.functional.upsample(input, size=None, scale_factor=None, mode=‘nearest’, align_corners=None)

where my statement is as follows:
image =image.view(1,3,h,w)
resizedimg = F.upsample(image, size=(nw,nh),mode = ‘bilinear’)

when I print (resizedimg.shape), it shows that resizedimg is single channle, but I input a 3 channel tensor, I do not understand why output become single channel.

and I read in help document that size could be Tuple[int, int, int]), but when I set size as : size=(3,nw,nh), It prompt that size only could be 2 dimesion.

Could anybody help me?

The issue is strange, as image should have 3 channels if the previous line of code didn’t throw an error.
Here is a small code snippet working as you would expect:

x = torch.randn(1, 3, 24, 24)
output = F.upsample(x, size=(30, 30), mode='bilinear')
> torch.Size([1, 3, 30, 30])

Could you print the image shape right before passing it to F.upsample?

sorry, I have make mistake. That function is ok