Srinivasan Sivanandan

`extra_tokens["channels"]` should contain channel indices per batch and should be of shape batch_size x n_channels. For example, in the ImageNet [dataset](https://github.com/insitro/ChannelViT/blob/1077a103e30cf20b6223b64935666962d0e2e836/channelvit/data/imagenet.py#L60), we return a dictionary containing channels per sample which...

@shuaijun-36 [README](https://github.com/insitro/ChannelViT?tab=readme-ov-file#channel-vision-transformer-an-image-is-worth-c-x-16-x-16-words) provides all the details for reproducing the experiments with instructions/links to download datasets from public repositories for ImageNet, Camelyon17, So2Sat and JUMP-CP