Resolve "Introduce conditional prior sets"

Closes #270 (closed)

PP test show that the new capabilities do not interfere with the results we are getting so far. This test is done without any conditional priors, it just shows that everything still works as expected.

https://ldas-jobs.ligo.caltech.edu/~moritz.huebner/conditional_prior_pp/outdir_pp_test_4s/pp.png

bilby/core/sampler/base_sampler.py

@@ -39,6 +39,11 @@ class Sampler(object):
         The result class to use. By default, `bilby.core.result.Result` is used,
         but objects which inherit from this class can be given providing
         additional methods.
+    soft_init: bool, optional
+        Switch to enable a soft initialization that prevents the likelihood
+        from being tested before running the sampler. This is relevant when
+        using custom likelihoods that must NOT be initialized on the main thread
+        when using multiprocessing, e.g. when using tensorflow in the likelihood.
     **kwargs: dict
         Additional keyword arguments
 
@@ -86,7 +91,7 @@ class Sampler(object):
             self, likelihood, priors, outdir='outdir', label='label',
             use_ratio=False, plot=False, skip_import_verification=False,
             injection_parameters=None, meta_data=None, result_class=None,
-            likelihood_benchmark=False,
+            likelihood_benchmark=False, soft_init=False,
             **kwargs):
         self.likelihood = likelihood
         if isinstance(priors, PriorDict):
@@ -108,9 +113,12 @@ class Sampler(object):
         self._fixed_parameter_keys = list()
         self._constraint_parameter_keys = list()
         self._initialise_parameters()
-        self._verify_parameters()
-        self._time_likelihood()
-        self._verify_use_ratio()
+
+        if not soft_init:
+            self._verify_parameters()
+            self._time_likelihood()
+            self._verify_use_ratio()
+
         self.kwargs = kwargs
 
         self._check_cached_result()