How should we handle gated models when testing

[KennethEnevoldsen]

          We have quite a few gated models. Should we add a token for these for the test or ignore gated errors?

Originally posted by @KennethEnevoldsen in #1698 (comment)

[isaac-chung]

@KennethEnevoldsen or @Muennighoff do we already have an HF_TOKEN (read) in repo secrets that can be used? I don't mind adding mine but I don't have access to repo settings.

[Samoed]

Can we just run these models locally? It seems we need to manually apply settings for each model (most cases would likely involve resolving access to Llama-3).