Error calling CasualForestDML.dowhy.fit()

Wanting to use the DoWhy module during the fit call so I can do refutation later. Getting the following error. Can you help? For context, this is being run inside Databricks in conjunction with MLFlow. EconML version 0.11.1. DoWhy version 0.6.

TypeError                                 Traceback (most recent call last)
<command-1717976837619280> in <module>
      7 with mlflow.start_run(run_name=run_name) as run:
      8 
----> 9   causal_forest.dowhy.fit(y, t, X=x, W=w)
     10 
     11   # log model params

/local_disk0/pythonVirtualEnvDirs/virtualEnv-cea1c30c-4db1-4650-a59d-8736863da592/lib/python3.7/site-packages/econml/dowhy.py in fit(self, Y, T, X, W, Z, outcome_names, treatment_names, feature_names, confounder_names, instrument_names, graph, estimand_type, proceed_when_unidentifiable, missing_nodes_as_confounders, control_value, treatment_value, target_units, **kwargs)
    158             missing_nodes_as_confounders=missing_nodes_as_confounders
    159         )
--> 160         self.identified_estimand_ = self.dowhy_.identify_effect(proceed_when_unidentifiable=True)
    161         method_name = "backdoor." + self._cate_estimator.__module__ + "." + self._cate_estimator.__class__.__name__
    162         init_params = {}

/local_disk0/pythonVirtualEnvDirs/virtualEnv-cea1c30c-4db1-4650-a59d-8736863da592/lib/python3.7/site-packages/dowhy/causal_model.py in identify_effect(self, estimand_type, method_name, proceed_when_unidentifiable)
    137                                            method_name,
    138                                            proceed_when_unidentifiable=proceed_when_unidentifiable)
--> 139         identified_estimand = self.identifier.identify_effect()
    140 
    141         return identified_estimand

/local_disk0/pythonVirtualEnvDirs/virtualEnv-cea1c30c-4db1-4650-a59d-8736863da592/lib/python3.7/site-packages/dowhy/causal_identifier.py in identify_effect(self)
     46         """
     47         if self.estimand_type == CausalIdentifier.NONPARAMETRIC_ATE:
---> 48             return self.identify_ate_effect()
     49         elif self.estimand_type == CausalIdentifier.NONPARAMETRIC_NDE:
     50             return self.identify_nde_effect()

/local_disk0/pythonVirtualEnvDirs/virtualEnv-cea1c30c-4db1-4650-a59d-8736863da592/lib/python3.7/site-packages/dowhy/causal_identifier.py in identify_ate_effect(self)
     68                 self.outcome_name,
     69                 backdoor_sets,
---> 70                 estimands_dict)
     71         # Setting default "backdoor" identification adjustment set
     72         default_backdoor_id = self.get_default_backdoor_set_id(backdoor_variables_dict)

/local_disk0/pythonVirtualEnvDirs/virtualEnv-cea1c30c-4db1-4650-a59d-8736863da592/lib/python3.7/site-packages/dowhy/causal_identifier.py in build_backdoor_estimands_dict(self, treatment_name, outcome_name, backdoor_sets, estimands_dict, proceed_when_unidentifiable)
    335             backdoor_estimand_expr = self.construct_backdoor_estimand(
    336                 self.estimand_type, treatment_name,
--> 337                 outcome_name, backdoor_sets_arr[i])
    338             self.logger.debug("Identified expression = " + str(backdoor_estimand_expr))
    339             estimands_dict["backdoor"+str(i+1)] = backdoor_estimand_expr

/local_disk0/pythonVirtualEnvDirs/virtualEnv-cea1c30c-4db1-4650-a59d-8736863da592/lib/python3.7/site-packages/dowhy/causal_identifier.py in construct_backdoor_estimand(self, estimand_type, treatment_name, outcome_name, common_causes)
    437         num_expr_str = outcome_name
    438         if len(common_causes)>0:
--> 439             num_expr_str += "|" + ",".join(common_causes)
    440         expr = "d(" + num_expr_str + ")/d" + ",".join(treatment_name)
    441         sym_mu = sp.Symbol("mu")

TypeError: sequence item 0: expected str instance, int found

Start at CasualForestDML.dowhy.fit and follow the traceback into dowhy_.identify_effect(), especially causal_identifier.py's construct_backdoor_estimand. Reproduce with EconML 0.11.1, DoWhy 0.6, and the Databricks/MLFlow context described; done means the fit call completes without the reported TypeError and permits later refutation.