Patrick Labatut

@xhochy Thanks! Sure, I had actually signed the CLA shortly after seeing this being required. It looks to me the matching [PR](https://github.com/conda/infrastructure/pull/823) in the infrastructure repo did acknowledge that (with...

The volume renderer should return "features" (i.e. color) and opacity values. A first thing to try would thus be to leverage the returned opacity to blend the returned color with...

I will try to add patch matching to this work-in-progress PR: #187

Releasing LVD-142M is not something that we are considering releasing, I am afraid. Re: open-sourcing the data curation code, this could depend on feedback and interest from the community.

Using #56 instead to keep track of requests about data curation code. Also happy to provide clarifications on the procedure.

Thanks @oylz for helping out! @dnth We have never tried to export the models to ONNX. If there is enough feedback / signal, we could considered providing model weights in...

Thanks @alihassanijr for helping out! To be clear loading a model (via `torch.hub.load()`) should not actually require xFormers (but can optionally use it if available). If you have xFormers already...

With #180 it's now possible to force not trying to use `xFormers` (even when it is installed and can be imported) by setting the `XFORMERS_DISABLED` environment variable. That's obviously a...

No, but that's something we will definitely consider documenting better. Meanwhile I would point you the evaluation code available under `dinov2/eval/` which does extract (and combine) features to feed them...

Closing this specific issue and keeping #2 open until the documentation is improved.