Would a GPU make this code run faster?

[brandon-holt]

Would a GPU make this code run faster for making recommendations, training model, etc.?

[AdrianSosic]

Hi @brandon-holt, this is a duplicate of #356, hence closing the issue. As you can see there, the discussion has already started with a list of open TODOs. But yes, the code execution can largely benefit from GPU support!

[brandon-holt]

@AdrianSosic Thank you!

So all I need to do is set torch.set_default_device("cuda") before i add_measurements() and recommend()?

And what version of baybe starts supporting GPU?

[AdrianSosic]

Hi @brandon-holt, I'm afraid you misunderstood me 😬 The support is not yet there. I actually don't expect that much needs to be adjusted to get it running (ideally, this should only require setting some device configuration in a few places) but it's also not a super high priority at the moment, I have to say. Perhaps @Hrovatin can look into this as her next work item? Can discuss this with the team early next week