Update run_dagger.py

run_dagger.py

@@ -77,14 +77,10 @@ def main():
     # architecture of the MLP policy function
     x = tf.placeholder(tf.float32, shape=[None, obs_dim])
     yhot = tf.placeholder(tf.float32, shape=[None, act_dim])
-    is_train = tf.placeholder(tf.bool)
 
     h1 = tf.layers.dense(inputs=x, units=128, activation=tf.nn.relu)
-    # drop1=tf.layers.dropout(inputs=h1,rate=0.5,training=is_train)
     h2 = tf.layers.dense(inputs=h1, units=64, activation=tf.nn.relu)
-    # drop2=tf.layers.dropout(inputs=h2,rate=0.5,training=is_train)
     h3 = tf.layers.dense(inputs=h2, units=32, activation=tf.nn.relu)
-    # drop3=tf.layers.dropout(inputs=h3,rate=0.5,training=is_train)
     yhat = tf.layers.dense(inputs=h3, units=act_dim, activation=None)
 
     loss_l2 = tf.reduce_mean(tf.square(yhot - yhat))
@@ -106,10 +102,10 @@ def main():
             batch_size = 25
             for step in range(10000):
                 batch_i = np.random.randint(0, obs_data.shape[0], size=batch_size)
-                train_step.run(feed_dict={x: obs_data[batch_i, ], yhot: act_data[batch_i, ], is_train: 1})
+                train_step.run(feed_dict={x: obs_data[batch_i, ], yhot: act_data[batch_i, ]})
                 if (step % 1000 == 0):
                     print 'opmization step ', step
-                    print 'obj value is ', loss_l2.eval(feed_dict={x:obs_data, yhot:act_data, is_train: 0})
+                    print 'obj value is ', loss_l2.eval(feed_dict={x:obs_data, yhot:act_data})
             print 'Optimization Finished!'
             # use trained MLP to perform
             max_steps = env.spec.timestep_limit
@@ -124,7 +120,7 @@ def main():
                 totalr = 0.
                 steps = 0
                 while not done:
-                    action = yhat.eval(feed_dict={x:obs[None, :], is_train:0})
+                    action = yhat.eval(feed_dict={x:obs[None, :]})
                     observations.append(obs)
                     actions.append(action)
                     obs, r, done, _ = env.step(action)