Model type (LoRa, Cheheckpoint, etc) should be in the version settings, not in the global settings of the model page.

[ostap667inbox]

Look at this page

Here is one model, but two versions and two kinds of files to download - Checkpoint and LoRa. It would be logically incorrect to separate them into different pages, since both models were created within the same training on the same dataset.

Now I can attach a checkpoint file to one version and a LoRa file to the other, which is fine. But I can't specify for each version separately: 'this is LoRa, and this is Checkpoint'.

I can only select one type for the whole model page, globally. This is inconvenient for users who use the search. And, as you can see, they also write in the comments that the file type is not specified correctly.

Is it possible to implement this idea or will we have to make a separate model page for each file type?