From 8a031e6626b0b75abd29162977153e92d6f273ef Mon Sep 17 00:00:00 2001
From: Emile Mathieu <emilem@bamlab.fr>
Date: Fri, 11 Aug 2017 16:17:48 +0100
Subject: [PATCH] typo, scope_iter and scope fixes

---
 edward/inferences/hmcda.py | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/edward/inferences/hmcda.py b/edward/inferences/hmcda.py
index 219ed3e62..5c3c3aa0d 100644
--- a/edward/inferences/hmcda.py
+++ b/edward/inferences/hmcda.py
@@ -60,7 +60,8 @@ def initialize(self, n_adapt, delta=0.65, Lambda=0.15, *args, **kwargs):
     Lambda : float, optional
       Target leapfrog length
     """
-    self.scope_iter = 0  # a convenient counter for log joint calculations
+    # store global scope for log joint calculations
+    self._scope = tf.get_default_graph().unique_name("inference") + '/'
 
     # Find initial epsilon
     step_size = self.find_good_eps()
@@ -137,7 +138,7 @@ def build_update(self):
     should_adapt = self.t <= self.n_adapt
     assign_ops = tf.cond(should_adapt,
                          lambda: self._adapt_step_size(alpha),
-                         lambda: self._do_not__adapt_step_size(alpha))
+                         lambda: self._do_not_adapt_step_size(alpha))
 
     # Update Empirical random variables.
     for z, qz in six.iteritems(self.latent_vars):
@@ -148,7 +149,7 @@ def build_update(self):
     assign_ops.append(self.n_accept.assign_add(tf.where(accept, 1, 0)))
     return tf.group(*assign_ops)
 
-  def _do_not__adapt_step_size(self, alpha):
+  def _do_not_adapt_step_size(self, alpha):
     # Do not adapt step size but assign last running averaged epsilon to epsilon
     assign_ops = []
     assign_ops.append(tf.assign(self.H_B, self.H_B).op)
@@ -188,7 +189,7 @@ def find_good_eps(self):
       old_r[z] = normal.sample()
 
     # Initialize espilon at 1.0
-    epsilon = tf.Variable(1.0, trainable=False)
+    epsilon = tf.constant(1.0)
 
     # Calculate log joint probability
     old_z = {z: tf.gather(qz.params, 0)
@@ -257,8 +258,7 @@ def _log_joint(self, z_sample):
     z_sample : dict
       Latent variable keys to samples.
     """
-    self.scope_iter += 1
-    scope = 'inference_' + str(id(self)) + '/' + str(self.scope_iter)
+    scope = self._scope + tf.get_default_graph().unique_name("sample")
     # Form dictionary in order to replace conditioning on prior or
     # observed variable with conditioning on a specific value.
     dict_swap = z_sample.copy()