--- a
+++ b/local_test/local_grade.py
@@ -0,0 +1,40 @@
+import random
+from local_test.baselines.common.vec_env.vec_normalize import VecNormalize
+from local_test.baselines.common.vec_env.dummy_vec_env import DummyVecEnv
+import local_test.baselines.ppo2.ppo2 as ppo2
+import local_test.baselines.ppo2.policies as policies
+from local_test.local_grade_env import LocalGradeEnv, LocalGradeRepeatActionEnv
+from local_test.baselines.logger import configure
+
+import tensorflow as tf
+
+
+def make_local_grade_env():
+    random_seeds = [random.randint(0, 2 ** 32 - 1) for _ in range(10)]
+    local_env = LocalGradeEnv(random_seeds, visualization=False)
+    repeated_local_env = LocalGradeRepeatActionEnv(env=local_env, repeat=2)
+    return repeated_local_env
+
+
+configure(dir="./logs")
+config = tf.ConfigProto()
+with tf.Session(config=config):
+    env = DummyVecEnv([make_local_grade_env])
+    env = VecNormalize(env)
+    ppo2.learn(
+        policy=policies.MlpPolicy,
+        env=env,
+        nsteps=100000,
+        nminibatches=1,
+        lam=0.95,
+        gamma=0.99,
+        noptepochs=5,
+        log_interval=1,
+        vf_coef=0.5,
+        ent_coef=0.001,
+        lr=3e-4,
+        cliprange=0.2,
+        save_interval=10,
+        load_path="../checkpoints/00160",
+        total_timesteps=100000
+    )