beanmachine

beanmachine copied to clipboard

Elementwise multiplication

2

Reviewed By: yucenli Differential Revision: D36650612

shannyang

Add stronger preconditions to devectorizer

2

Summary: Many Bean Machine models are vectorized -- a model for instance might sample from a `Normal(tensor([0., 0]), tensor([1., 1]))` distribution to get a two-element vector of samples, and then...

ericlippert

Start implementing tensor sum operator support in compiler

2

Summary: We wish to compile Bean Machine models containing tensor sum operators into BMG graphs; this diff gets a start on that feature by detecting usages of the sum operator...

ericlippert

Compiler now requires matmul operands to be real matrices

2

Summary: This diff completes my efforts to finish the implementation of matrix multiplication in the compiler. We now ensure that the inputs to any matrix multiplication are double-valued matrices, as...

ericlippert

Diagnostic tool with summary tab

1

### Feature Description Add a tool that calculates the standard stats from ArviZ in a data grid for the tool.

ndmlny-qs

Marginal 2D labels

1

### Feature Description Add point statistic labels to the marginal 2D tool, just like the marginal 1D tool has. See image below as reference. 

ndmlny-qs

### Feature Description Allow the user to choose which point statistic (mean, media, mode?) for display in the marginal diagnostic tools

ndmlny-qs

### Feature Description Investigate using `__repr_html__` for displaying diagnostic tools directly from a mcs object.

ndmlny-qs

### Issue Description In Bean Machine right now there is some code for converting MonteCarloSamples objects into arviz-compatible InferenceData objects. It would be better if that functionality lived in arviz.

zaxtax

Accessing params in SVI

1

Right now, there isn't a great way to access the params directly associated with a given guide program. Is this a deliberate design choice, or do we want to expose...

zaxtax