Merge pull request openai#106 from colllin/patch-1

[Exercise 1.3] Flatten actions in policy rollout
MKorablyov · Jun 28, 2019 · affcea4 · affcea4
2 parents 71495c1 + 617dd56
commit affcea4
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/spinup/exercises/problem_set_1/exercise1_3.py b/spinup/exercises/problem_set_1/exercise1_3.py
@@ -258,7 +258,7 @@ def td3(env_fn, actor_critic=core.mlp_actor_critic, ac_kwargs=dict(), seed=0,
     logger.setup_tf_saver(sess, inputs={'x': x_ph, 'a': a_ph}, outputs={'pi': pi, 'q1': q1, 'q2': q2})
 
     def get_action(o, noise_scale):
-        a = sess.run(pi, feed_dict={x_ph: o.reshape(1,-1)})
+        a = sess.run(pi, feed_dict={x_ph: o.reshape(1,-1)})[0]
         a += noise_scale * np.random.randn(act_dim)
         return np.clip(a, -act_limit, act_limit)
 
@@ -382,4 +382,4 @@ def test_agent(n=10):
     if args.use_soln:
         true_td3(**all_kwargs)
     else:
-        td3(**all_kwargs)
+        td3(**all_kwargs)