Austin Tripp
Austin Tripp
@Ryan-Rhys can you link me to the PR where this was merged? Happy to close this PR if the functionality has been added.
Try setting a custom PYTHONPATH in your call to it, such as `PYTHONPATH="some/dir:$PYTHONPATH" hyperopt-mongo-worker `
I think this was also introduced in #60
> Thanks for the super detailed example! Will try to get a fix in within the next week! Will add you as a PR reviewer if that's OK? Yes, happy...
> FWIW I'd prefer to reset caches if parameters change/on `initialize`. We have a library wrapping a gpytorch model for fully bayesian inference, and often I load multiple sets of...
> @jpchen @AustinT when you call initialize to load parameters for the kernel, are you typically calling initialize on the kernel (e.g., `model.covar_module.initialize(lengthscale=...)`) or on the model (e.g. `model.initialize(**{'covar_module.lengthscale': ...})`)?...
> @AustinT @jacobrgardner @jpchen one possible (but maybe overly-complicated) solution would be to give each module a UUID that is updated every time some sort of state is changed (e.g....
> I think changing `Module`'s `setattr` to give each module a parent reference would fix this in the majority of cases. The only edge case I can think of is...
A bit late, but I solved this issue in #10
Hi @john-bradshaw , @yuxuanou623 is an MPhil student working with Miguel and I. We discussed this issue in our meeting so I thought I would comment. @yuxuanou623 , pytorch versions...