Add modified Bayesian regression tutorial with more direct PyTorch usage #2996
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR is an attempt at making a couple small changes to the Pyro and
PyroModuleAPI to makePyroModulemore compatible with vanilla PyTorch programming idioms. The API changes are simple, although the implementations insidePyroModuleare a bit hacky and may not yet be correct.Changes:
pyro.enable_module_local_param()forPyroModuleparameters to be stored locally, rather than the global parameter store. Currently implemented by associating a newParamStoreDictobject with eachPyroModuleinstance, which may not be ideal.__call__method topyro.infer.ELBOthat returns atorch.nn.Modulebound to a specific model and guide, allowing direct use of the PyTorch JIT API (e.g.torch.jit.trace)PyroModule