Normalize should map image tensors in the range -1 to 1 to be compatible with Meta-Dataset

[jfb54]

Instead of the ImageNet friendly normalize transform:

normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])

use the following:

normalize = transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])

as Meta-Dataset guarantees that all images will be normalized in the range -1 to 1.

[mboudiaf]

Indeed, the transforms that I've put in my example don't scrupulously follow the ones from the original dataset. This should not make a big difference though. Still, I will modify to follow those at https://github.com/google-research/meta-dataset/blob/a38e8cea0f60e3aa1edb4424f5574c10e355f3d4/meta_dataset/data/decoder.py#L74

[jfb54]

Awesome. Thanks!