What if the pretrained models are not a part of either library, but trained on a standard publically available dataset, eg. ImageNet. And similarly many of the transformer models, BERT, GPT2, etc.
And who decides what a popular dataset is.
What if the pretrained models are not a part of either library, but trained on a standard publically available dataset, eg. ImageNet. And similarly many of the transformer models, BERT, GPT2, etc.
And who decides what a popular dataset is.