pymc.vectorize_over_posterior#
- pymc.vectorize_over_posterior(outputs, posterior, input_rvs, allow_rvs_in_graph=True, sample_dims=('chain', 'draw'))[source]#
Vectorize outputs over posterior samples of subset of input rvs.
This function creates a new graph for the supplied outputs, where the required subset of input rvs are replaced by their posterior samples (chain and draw dimensions are flattened). The other input tensors are kept as is.
- Parameters:
- outputs
list[Variable] The list of variables to vectorize over the posterior samples.
- posterior
xr.Dataset The posterior samples to use as replacements for the input_rvs.
- input_rvs
list[Variable] The list of random variables to replace with their posterior samples.
- allow_rvs_in_graphbool
Whether to allow random variables to be present in the graph. If False, an error will be raised if any random variables are found in the graph. If True, the remaining random variables will be resized to match the number of draws from the posterior.
- sample_dims
tuple[str, …] The dimensions of the posterior samples to use for vectorizing the input_rvs.
- outputs
- Returns:
- vectorized_outputs
list[Variable] The vectorized variables
- vectorized_outputs
- Raises:
RuntimeErrorIf random variables are found in the graph and allow_rvs_in_graph is False