[more training information] Is the training dataset(s), used for pretraining each model available information?

[LITZY0619]

Hi, I think someone will have the same question as I do. When using these pre-trained models, I am always curious which dataset, which training paradigm (e.g. fully supervised or self-supervised), and which loss functions are used to generate these pre-trained weights. Or are all models trained with supervised loss on ImageNet-1k? I would appreciate if this information could be provided in model.default_cfg or elsewhere.