Merge pull request #1905 from pymc-devs/revert-1810-master

Revert "small fix for multivariate mixture models"

Author: AustinRochford

pymc3/distributions/mixture.py

@@ -61,9 +61,7 @@ def __init__(self, w, comp_dists, *args, **kwargs):
 
         try:
             comp_modes = self._comp_modes()
-            # for logp the different modes are like different observations, i.e. rows, hence use comp_modes.T
-            # logPs is a vector, hence self.logp(..).T == self.logp(..)
-            comp_mode_logps = self.logp(comp_modes.T)
+            comp_mode_logps = self.logp(comp_modes)
             self.mode = comp_modes[tt.argmax(w * comp_mode_logps, axis=-1)]
 
             if 'mode' not in defaults: