Max Balandat

> I'm happy to try to some variant of this up as a PR as well over the coming weeks if that'd be useful. @wjmaddox that would be awesome! Did...

> we were using genetic algorithms to optimize everything b/c the tasks we considered were discrete cc @sdaulton this could be a good real-world test case for some of your...

Are you able to successfully use the `MultiDeviceKernel` in gpytorch? I just tried running the [`Simple_MultiGPU_GP_Regression` tutorial](https://github.com/cornellius-gp/gpytorch/blob/master/examples/02_Scalable_Exact_GPs/Simple_MultiGPU_GP_Regression.ipynb), and I'm getting a `RuntimeError` during `torch.autograd.backward`. Not sure if an issue with...

Could you upstream raise an issue on the gpytorch github about this? I am not very familiar with this code, hopefully someone else can take a look there to make...

> Though originally the problem was that there is no Multidevicekernel implementation in BoTorch. We don't necessarily aim to re-implement / re-wrap every model/kernel combination in Botorch, that would be...

I don't see anything blatantly wrong with this. I'm not very familiar with `MultiDeviceKernel` and whether it fully supports generic batching etc. But let's figure out the upstream issue first...

Can we do a `from .common ..` import instead of a `from common` import in the `test_` files?

Hmm this looks like a bug. @esantorella would you be able to take a look at this next week?

I tried this out but wasn't able to reproduce this issue. I had to make some changes since I don't have the `problem_function` module in your code example. The following...

Hmm looks like your example doesn't do the `.to('cuda')` on the instantiated `KroneckerMultiTaskGP`? If you do that does this fix the issue? (note that this was part of your original...