Some(?) models seem to have their discriminators posted as separate models

[joeyballentine]

Example: https://openmodeldb.info/models/2x-VimeoScale-Unet

[ferrosa]

As per chaiNNer-org/chaiNNer#2228 (comment) and also the model tag, this doesn't appear to be a discriminator, but a SOFVSR model.
I guess the issue is that not all .pth models are equal and users are left unguided.
In particular: chaiNNer can't run this one.
Maybe a "compatible with chaiNNer 0.X.Y+" tag/filter option (or a tools compatibility checklist... hellish to maintain), but that's off issue.

[joeyballentine]

I guess the issue is that not all .pth models are equal and users are left unguided.

.pth files are just model weights. They contain no information about what the model is supposed to be, what kind of input it needs, etc. It's just a bunch of numbers meant to be fed into the model structure (technically it's a bunch of key/value pairs of tensors). This means that to use a model, you need to support its architecture, and you cant just use random models with random programs. chaiNNer is a bit special in that I've come up with a way to detect model arch by how it's pth file looks, but it's not perfect and still means I need to add manual support for new architectures.

And yeah, showing a list of programs that support each model is something we wanted to have but haven't gotten around to implementating yet