Make it possible to pickle simple kernels and likelihoods #1876
Conversation
| @@ -172,7 +172,7 @@ def __init__( | |||
| if not isinstance(lengthscale_prior, Prior): | |||
| raise TypeError("Expected gpytorch.priors.Prior but got " + type(lengthscale_prior).__name__) | |||
| self.register_prior( | |||
| "lengthscale_prior", lengthscale_prior, lambda m: m.lengthscale, lambda m, v: m._set_lengthscale(v) | |||
| "lengthscale_prior", lengthscale_prior, self._lengthscale_param, self._lengthscale_closure | |||
There was a problem hiding this comment.
I guess this should also work here?
| "lengthscale_prior", lengthscale_prior, self._lengthscale_param, self._lengthscale_closure | |
| "lengthscale_prior", lengthscale_prior, "lengthscale", self._lengthscale_closure |
There was a problem hiding this comment.
I don't think this works (making this change causes 31 tests to fail). Note that _lengthscale_param returns m.lengthscale and not self.lengthscale.
There was a problem hiding this comment.
Hmm not sure I understand why it doesn’t, seems like the closure definition here should do the same?
Line 235 in 32cde57
Either way, we can figure that out later (or I’m missing sth obvious). Accepting this as is.
There was a problem hiding this comment.
Sounds good! Looks like merging is blocked though until the docs can be built successfully. Looking at https://readthedocs.org/projects/gpytorch/builds/15684237/ it looks like the issue is a warning that is unrelated to this PR.
|
Ok, fixed the doc building issue by pinning |
We sometimes need to pickle (simple) GPyTorch models, but this doesn't work when using priors since
register_priorcreates local lambdas. This PR makes it possible to pickle simple kernels such as Matern, RBF, Scale as well as the Gaussian likelihood.