Disable sampling of transformed variables from prior_predictive

[ricardoV94]

This was added to facilitate sampling from the prior as a starting point for SMC, but with the new functions added in #4983 this is no longer needed.

pymc/pymc3/sampling.py

Lines 1983 to 1999 in 6d2aa5d

	# Any variables from var_names that are missing must be transformed variables.
	# Misspelled variables would have raised a KeyError above.
	missing_names = vars_.difference(names)
	for name in missing_names:
	transformed_value_var = model[name]
	rv_var = model.values_to_rvs[transformed_value_var]
	transform = transformed_value_var.tag.transform
	transformed_rv_var = transform.forward(rv_var, rv_var)
	names.append(name)
	vars_to_sample.append(transformed_rv_var)
	# If the user asked for the transformed variable in var_names, but not the
	# original RV, we add it manually here
	if rv_var.name not in names:
	names.append(rv_var.name)
	vars_to_sample.append(rv_var)

[markvrma]

I'd like to take up this issue.

[purna135]

Is it okay if I work on this?

[ricardoV94]

Is it okay if I work on this?

Sure

[purna135]

I need some assistance in replacing the following code with make_initial_point_fn

pymc/pymc/smc/smc.py

Lines 179 to 184 in e6fc2ec

	return sample_prior_predictive(
	self.draws,
	var_names=[v.name for v in self.model.unobserved_value_vars],
	model=self.model,
	return_inferencedata=False,
	)

pymc/pymc/initial_point.py

Line 130 in d84c718

def make_initial_point_fn(

[ricardoV94]

You will have to change the parameters of make_initial_point_fn so that it is returning draws from the transformed prior. You can do that by setting the overrides dictionary so that every value variable strategy is "prior", and also switching the default from "moment" to "prior". Then you will need to make sure you call the returned function with a different seed self.draws times.