Some confusion about the torchvison.transorms.ToTensor class

The tutorial shows that Converts a PIL Image or numpy.ndarray (H x W x C) in the range [0, 255] to a torch.FloatTensor of shape (C x H x W) in the range [0.0, 1.0].So I wanna know why do we have to limit the value to [0, 1] rather than keep the origial value, just to change the tensor shape. This transformation can help it converge faster or any other reason?

Yes, you usually want to normalize your input because it helps with training.

You can find a nice short explanation on normalizing inputs in this video from the Deep Learning course on Coursera.