Version 1.3.3

docs/tex/troubleshooting.tex

@@ -46,14 +46,16 @@ \subsubsection{Full Installation}
   \item Neural networks are supported through three libraries:
   \href{http://keras.io}{Keras} (>=1.0)
 \begin{lstlisting}[language=JSON]
-pip install keras
+pip install keras==2.0.4
 \end{lstlisting}
   \href{https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/slim}{TensorFlow Slim}
   (native in TensorFlow), and
   \href{https://github.com/google/prettytensor}{PrettyTensor} (>=0.7.4)
 \begin{lstlisting}[language=JSON]
 pip install prettytensor
 \end{lstlisting}
+Note that for Keras 2.0.5 and beyond, all neural net layer transformations cannot be directly applied on random variables anymore. For example, if \texttt{x} is a \texttt{ed.RandomVariable} object, one must call \texttt{tf.convert_to_tensor} before applying it to a layer transformation, \texttt{Dense(256)(tf.convert_to_tensor(x))}.
+See \href{https://github.com/fchollet/keras/issues/6979}{here} for more details.
   \item Notebooks require
   \href{http://jupyter.org}{Jupyter} (>=1.0.0)
 \begin{lstlisting}[language=JSON]

edward/inferences/gan_inference.py

@@ -224,7 +224,6 @@ def _build_optimizer(optimizer, global_step):
                                                100, 0.9, staircase=True)
   else:
     learning_rate = 0.01
-    global_step = None
 
   # Build optimizer.
   if optimizer is None:
@@ -247,6 +246,6 @@ def _build_optimizer(optimizer, global_step):
     else:
       raise ValueError('Optimizer class not found:', optimizer)
   elif not isinstance(optimizer, tf.train.Optimizer):
-    raise TypeError("Optimizer must be a str, tf.train.Optimizer, or None.")
+    raise TypeError("Optimizer must be str, tf.train.Optimizer, or None.")
 
   return optimizer, global_step

edward/inferences/inference.py

@@ -163,7 +163,10 @@ def initialize(self, n_iter=1000, n_print=None, scale=None, logdir=None,
     Parameters
     ----------
     n_iter : int, optional
-      Number of iterations for algorithm.
+      Number of iterations for algorithm when calling ``run()``.
+      Alternatively if controlling inference manually, it is the
+      expected number of calls to ``update()``; this number determines
+      tracking information during the print progress.
     n_print : int, optional
       Number of iterations for each print progress. To suppress print
       progress, then specify 0. Default is ``int(n_iter / 100)``.

edward/inferences/variational_inference.py

@@ -99,7 +99,6 @@ def initialize(self, optimizer=None, var_list=None, use_prettytensor=False,
                                                  100, 0.9, staircase=True)
     else:
       learning_rate = 0.01
-      global_step = None
 
     # Build optimizer.
     if optimizer is None:

edward/version.py

@@ -1 +1 @@
-__version__ = '1.3.2'
+__version__ = '1.3.3'