initial workaround to fix #2510

logp fails if value is not a TensorVariable. Also the SplitWrapper (self.interp_op) does not do broadcasting. This workaround is probably not the fastest though.

Author: michaelosthege

pymc3/distributions/continuous.py

@@ -2004,4 +2004,11 @@ def random(self, point=None, size=None, repeat=None):
                                 size=size)
 
     def logp(self, value):
-        return tt.log(self.interp_op(value) / self.Z)
+        # TODO: this implementation is suboptimal
+        #       the SplineWrapper should get proper broadcasting behavior
+        value = tt.as_tensor_variable(value)
+        if hasattr(value, "__iter__"):
+            iv = tt.stack([self.interp_op(v) for v in value])
+        else:
+            iv = self.interp_op(value)
+        return tt.log(iv / self.Z)