ready for experiment

docs/index.md

@@ -7,25 +7,36 @@ In this first and thrid video, you can see the robot learning how to interact wi
 Below there are four videos please fill out your responses in the survey below each video. Thanks for participating!
 
 ## Algorithm 1 (Learning)
-link youtube vid here ...
+
+<iframe width="420" height="315"
+src="https://www.youtube.com/embed/tgbNymZ7vqY">
+</iframe>
 
 
 <iframe src="https://docs.google.com/forms/d/e/1FAIpQLSfcW9IFqY0mSJinQuvx8mZV1KMFScGUH_wVxpi6KV4SIMtUVw/viewform?embedded=true" width="640" height="890" frameborder="0" marginheight="0" marginwidth="0">Loading…</iframe>
 
 ## Algorithm 1
 
+<iframe width="420" height="315"
+src="https://www.youtube.com/embed/tgbNymZ7vqY">
+</iframe>
 
 <iframe src="https://docs.google.com/forms/d/e/1FAIpQLSfeYxbYgFo_Dyy5n0Ap_MxWxOJbfXomKOAY5SuRk9BhM2fafA/viewform?embedded=true" width="640" height="390" frameborder="0" marginheight="0" marginwidth="0">Loading…</iframe>
 
 ## Algorithm 2 (Learning)
 
-<iframe width="640" height="1057"
+<iframe width="420" height="315"
 src="https://www.youtube.com/embed/tgbNymZ7vqY">
 </iframe>
 
 <iframe src="https://docs.google.com/forms/d/e/1FAIpQLSf1ESyx6VQF0gWw-EU1O__RvmL7Gak7DjO1yrS67ZFGFokp4w/viewform?embedded=true" width="640" height="1057" frameborder="0" marginheight="0" marginwidth="0">Loading…</iframe>
 
 ## Algorithm 2
 
+<iframe width="420" height="315"
+src="https://www.youtube.com/embed/tgbNymZ7vqY">
+</iframe>
+
+
 <iframe src="https://docs.google.com/forms/d/e/1FAIpQLSfR9pqg4Wj6w175omNOWPtDBD2C81l8kkPNz6zcQwWXKTzvJA/viewform?embedded=true" width="640" height="1057" frameborder="0" marginheight="0" marginwidth="0">Loading…</iframe>
 

main.py

@@ -13,9 +13,10 @@
     IS_PI = True
     # from src.DRL.run_session import train_and_test_bot
     from src.DRL.train import train
+    from src.DRL.test import test
 elif hostname == settings.PC:
     IS_PI = False
-   # from src.MainNode.Utils.OverHead import OverHead
+    # from src.MainNode.Utils.OverHead import OverHead
     # from src.MainNode.Utils.stream_oh_inference import run
 else:
     raise Exception("Must configure this device in settings.py")
@@ -30,8 +31,9 @@ def main_train_and_test():
     else:
         pibot = PiBot2()
         steps = 200
-        train.train_session(pibot, steps)
+        train.train_session(pibot, steps, continue_prompt=True)
         # publish_data(actions)
+        test.run_policy(pibot, steps, continue_prompt=True)
 
 
 

src/DRL/test/test.py

@@ -9,24 +9,22 @@
 with open(POLICYF, "r") as f:
     POLICIES = json.load(f)
 
-def run_policy(num_steps,pibot, policy_dict=POLICIES) -> str:
+def run_policy(pibot,num_steps, policy_dict=POLICIES, continue_prompt=False) -> None:
+    # If the environment don't follow the interface, an error will be thrown
     env = PiBotEnv2(pibot)
+    # record actions
+    env._RECORD_ACTION = True
     check_env(env, warn=True)
-    # The algorithms require a vectorized environment to run
-    env = DummyVecEnv([partial(PiBotEnv2, PiBot=pibot)])
-    results = {"DEVICE_NAME":HOSTNAME,
-               "POLICIES":[]} 
+    d_env = DummyVecEnv([partial(PiBotEnv2, PiBot=pibot)])
     for k, v in policy_dict.items():
-        results["POLICIES"].append(k)
         if "PPO2" in k:
             model = PPO2.load(v)
         if "A2C" in k:
             model = A2C.load(v)
-        obs = env.reset()
-        results[k] = []
+        obs = d_env.reset()
         for i in range(num_steps):
           action, _states = model.predict(obs)
-          results[k].append(list(list(int(i) for i in a) for a in action))
           obs, rewards, done, info = env.step(action)
-        print(results)
-    return json.dumps(results)
+        env._record_actions(k+"_INFERENCE")
+        if continue_prompt:
+            input("hit enter to continue: ")

src/DRL/train/train.py

@@ -15,19 +15,21 @@
     #"ACER": partial(ACER, policy=MlpPolicy, verbose=1),
     # "ACKTR": partial(ACKTR, policy=MlpPolicy, verbose=1),
     # "DQN": partial(DQN, policy=MlpPolicy, verbose=1),
-   # "PPO2": partial(PPO2, policy=MlpPolicy, verbose=1),
+    "PPO2": partial(PPO2, policy=MlpPolicy, verbose=1),
     "A2C": partial(A2C, policy=MlpPolicy, verbose=1),
 
 
 }
 
-def train_session(pibot, steps, train_dict=TRAIN_DICT):
+def train_session(pibot, steps, train_dict=TRAIN_DICT, continue_prompt=False):
     """
     wrap train dict with training function
     """
     for model, func in train_dict.items():
         print(f"TRAINING: {model}")
         train(steps, pibot, func, model)
+        if continue_prompt:
+            input("hit enter to continue: ")
 
 def train(steps, pibot, model, model_name) -> str:
     # If the environment don't follow the interface, an error will be thrown
@@ -46,6 +48,6 @@ def train(steps, pibot, model, model_name) -> str:
       policies[model_name] = fpath
     with open(POLICYF, "w") as f:
         json.dump(policies, f)
-    env._record_actions(model_name)
+    env._record_actions(model_name+"_TRAINING")
 
     return fpath