Update Edward to version 1.3.0 (#623)

* version 1.3.0

* remove call to _ref now that tf.gradients uses as_ref=True in implementation

* make miscellaneous revisions

docs/tex/getting-started.tex

@@ -5,6 +5,11 @@ \subsection{Getting Started}
 
 \subsubsection{Installation}
 
+\textbf{Note:
+TensorFlow 1.1.0 made breaking API changes.
+Edward's latest stable version requires at least TensorFlow 1.1.0.
+}
+
 To install the latest stable version, run
 
 \begin{lstlisting}[language=Java]

docs/tex/iclr2017.tex

@@ -12,9 +12,9 @@ \subsection{Deep Probabilistic Programming}
 The code snippets assume the following versions.
 
 \begin{lstlisting}[language=bash]
-pip install -e "git+https://github.com/blei-lab/edward.git#egg=edward"
-pip install tensorflow==1.0.0  # alternatively, tensorflow-gpu==1.0.0
-pip install keras==1.0.0
+pip install edward==1.3.0
+pip install tensorflow==1.1.0  # alternatively, tensorflow-gpu==1.1.0
+pip install keras==2.0.0
 \end{lstlisting}
 
 \subsubsection{Section 3. Compositional Representations for Probabilistic Models}

edward/inferences/conjugacy/conjugate_log_probs.py

@@ -29,7 +29,7 @@ def beta_log_prob(self, val):
   conc0 = self.parameters['concentration0']
   conc1 = self.parameters['concentration1']
   result = (conc1 - 1.0) * tf.log(val)
-  result += (conc0 - 1.0) * tf.log(tf.constant(1.0) - val)
+  result += (conc0 - 1.0) * tf.log(1.0 - val)
   result += -tf.lgamma(conc1) - tf.lgamma(conc0) + tf.lgamma(conc1 + conc0)
   return result
 

edward/inferences/klpq.py

@@ -136,6 +136,6 @@ def build_loss_and_gradients(self, var_list):
     loss = tf.reduce_mean(w_norm * log_w)
     grads = tf.gradients(
         -tf.reduce_mean(q_log_prob * tf.stop_gradient(w_norm)),
-        [v._ref() for v in var_list])
+        var_list)
     grads_and_vars = list(zip(grads, var_list))
     return loss, grads_and_vars

edward/inferences/klqp.py

@@ -392,7 +392,7 @@ def build_reparam_loss_and_gradients(inference, var_list):
   q_log_prob = tf.stack(q_log_prob)
   loss = -tf.reduce_mean(p_log_prob - q_log_prob)
 
-  grads = tf.gradients(loss, [v._ref() for v in var_list])
+  grads = tf.gradients(loss, var_list)
   grads_and_vars = list(zip(grads, var_list))
   return loss, grads_and_vars
 
@@ -446,7 +446,7 @@ def build_reparam_kl_loss_and_gradients(inference, var_list):
 
   loss = -(tf.reduce_mean(p_log_lik) - kl)
 
-  grads = tf.gradients(loss, [v._ref() for v in var_list])
+  grads = tf.gradients(loss, var_list)
   grads_and_vars = list(zip(grads, var_list))
   return loss, grads_and_vars
 
@@ -504,7 +504,7 @@ def build_reparam_entropy_loss_and_gradients(inference, var_list):
 
   loss = -(tf.reduce_mean(p_log_prob) + q_entropy)
 
-  grads = tf.gradients(loss, [v._ref() for v in var_list])
+  grads = tf.gradients(loss, var_list)
   grads_and_vars = list(zip(grads, var_list))
   return loss, grads_and_vars
 
@@ -558,7 +558,7 @@ def build_score_loss_and_gradients(inference, var_list):
 
   grads = tf.gradients(
       -tf.reduce_mean(q_log_prob * tf.stop_gradient(losses)),
-      [v._ref() for v in var_list])
+      var_list)
   grads_and_vars = list(zip(grads, var_list))
   return loss, grads_and_vars
 
@@ -611,7 +611,7 @@ def build_score_kl_loss_and_gradients(inference, var_list):
   loss = -(tf.reduce_mean(p_log_lik) - kl)
   grads = tf.gradients(
       -(tf.reduce_mean(q_log_prob * tf.stop_gradient(p_log_lik)) - kl),
-      [v._ref() for v in var_list])
+      var_list)
   grads_and_vars = list(zip(grads, var_list))
   return loss, grads_and_vars
 
@@ -669,6 +669,6 @@ def build_score_entropy_loss_and_gradients(inference, var_list):
   grads = tf.gradients(
       -(tf.reduce_mean(q_log_prob * tf.stop_gradient(p_log_prob)) +
           q_entropy),
-      [v._ref() for v in var_list])
+      var_list)
   grads_and_vars = list(zip(grads, var_list))
   return loss, grads_and_vars

edward/inferences/map.py

@@ -122,6 +122,6 @@ def build_loss_and_gradients(self, var_list):
 
     loss = -p_log_prob
 
-    grads = tf.gradients(loss, [v._ref() for v in var_list])
+    grads = tf.gradients(loss, var_list)
     grads_and_vars = list(zip(grads, var_list))
     return loss, grads_and_vars

edward/version.py

@@ -1 +1 @@
-__version__ = '1.2.4'
+__version__ = '1.3.0'

notebooks/iclr2017.ipynb

@@ -8,9 +8,9 @@
     "\n",
     "The code snippets assume the following versions.\n",
     "```bash\n",
-    "pip install -e \"git+https://github.com/blei-lab/edward.git#egg=edward\"\n",
-    "pip install tensorflow==1.0.0  # alternatively, tensorflow-gpu==1.0.0\n",
-    "pip install keras==1.0.0\n",
+    "pip install edward==1.3.0\n",
+    "pip install tensorflow==1.1.0  # alternatively, tensorflow-gpu==1.1.0\n",
+    "pip install keras==2.0.0\n",
     "```"
    ]
   },