moved polciy to own model to allow it getting persisted properly

docs/tutorial_supervised.rst

@@ -151,9 +151,9 @@ For this bot, we came up with our own policy.
 This policy extends the Keras policy modifying the ML architecture of the
 underlying neural network.
 Check out the
-``RestaurantPolicy`` class in ``bot.py`` for the gory details:
+``RestaurantPolicy`` class in ``policy.py`` for the glory details:
 
-.. literalinclude:: ../examples/restaurantbot/bot.py
+.. literalinclude:: ../examples/restaurantbot/policy.py
    :linenos:
    :pyobject: RestaurantPolicy
 

examples/restaurantbot/bot.py

@@ -7,19 +7,19 @@
 import logging
 import warnings
 
+from policy import RestaurantPolicy
 from rasa_core import utils
 from rasa_core.actions import Action
 from rasa_core.agent import Agent
 from rasa_core.channels.console import ConsoleInputChannel
+from rasa_core.events import SlotSet
 from rasa_core.interpreter import RasaNLUInterpreter
-from rasa_core.policies.keras_policy import KerasPolicy
 from rasa_core.policies.memoization import MemoizationPolicy
-from rasa_core.events import SlotSet
 
 logger = logging.getLogger(__name__)
 
 
-class RestaurantAPI:
+class RestaurantAPI(object):
     def search(self, info):
         return "papi's pizza place"
 
@@ -48,30 +48,6 @@ def run(self, dispatcher, tracker, domain):
         return []
 
 
-class RestaurantPolicy(KerasPolicy):
-    def model_architecture(self, num_features, num_actions, max_history_len):
-        """Build a Keras model and return a compiled model."""
-        from keras.layers import LSTM, Activation, Masking, Dense
-        from keras.models import Sequential
-
-        n_hidden = 32  # size of hidden layer in LSTM
-        # Build Model
-        batch_shape = (None, max_history_len, num_features)
-
-        model = Sequential()
-        model.add(Masking(-1, batch_input_shape=batch_shape))
-        model.add(LSTM(n_hidden, batch_input_shape=batch_shape))
-        model.add(Dense(input_dim=n_hidden, output_dim=num_actions))
-        model.add(Activation('softmax'))
-
-        model.compile(loss='categorical_crossentropy',
-                      optimizer='adam',
-                      metrics=['accuracy'])
-
-        logger.debug(model.summary())
-        return model
-
-
 def train_dialogue(domain_file="restaurant_domain.yml",
                    model_path="models/dialogue",
                    training_data_file="data/babi_stories.md"):

examples/restaurantbot/policy.py

@@ -0,0 +1,34 @@
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+from __future__ import unicode_literals
+
+import logging
+
+from rasa_core.policies.keras_policy import KerasPolicy
+
+logger = logging.getLogger(__name__)
+
+
+class RestaurantPolicy(KerasPolicy):
+    def model_architecture(self, num_features, num_actions, max_history_len):
+        """Build a Keras model and return a compiled model."""
+        from keras.layers import LSTM, Activation, Masking, Dense
+        from keras.models import Sequential
+
+        n_hidden = 32  # size of hidden layer in LSTM
+        # Build Model
+        batch_shape = (None, max_history_len, num_features)
+
+        model = Sequential()
+        model.add(Masking(-1, batch_input_shape=batch_shape))
+        model.add(LSTM(n_hidden, batch_input_shape=batch_shape))
+        model.add(Dense(input_dim=n_hidden, output_dim=num_actions))
+        model.add(Activation('softmax'))
+
+        model.compile(loss='categorical_crossentropy',
+                      optimizer='adam',
+                      metrics=['accuracy'])
+
+        logger.debug(model.summary())
+        return model