Resolved merge conflicts

RL_Framework/pipeline.py

@@ -184,7 +184,7 @@ def evaluate_policy(self, folder_path, num_evals, env, is_baseline=False, baseli
             pathlib.Path(folder_path + '/Returns').mkdir(parents=True, exist_ok=True)
             filename = folder_path + '/Returns' + '/predict_' + str(i) + '.json'
             f = open(filename, 'w')
-            f.write(json.dumps(e))
+            f.write(json.dumps(e, indent=4))
             f.close()
 
     def single_run(self, folder_path, num_evals, policy_kwargs=None, is_baseline=False, baseline_policy=None):
@@ -240,7 +240,7 @@ def single_run(self, folder_path, num_evals, policy_kwargs=None, is_baseline=Fal
             # Instantiate the agent
 #            model = PPO2(CustomCnnPolicy, env, policy_kwargs=policy_kwargs, ent_coef=ent_coef, n_steps=n_steps, nminibatches=nminibatches, noptepochs=noptepochs, learning_rate=learning_rate, verbose=1, tensorboard_log=folder_path + '/ppo_v2_tensorboard/')
 
-            model = PPO2(MlpPolicy, env, verbose=1, tensorboard_log=folder_path + '/ppo_v2_tensorboard/')
+            model = PPO2(MlpPolicy, env, ent_coef=ent_coef, n_steps=n_steps, nminibatches=nminibatches, noptepochs=noptepochs, learning_rate=learning_rate, verbose=1, tensorboard_log=folder_path + '/ppo_v2_tensorboard/')
             # Train the agent
             model.learn(total_timesteps=rl_time_steps)  # this will crash explaining that the invalid value originated from the env
 
@@ -336,7 +336,7 @@ def batch_run(self, n, rl_config, garden_x, garden_y, num_plant_types, num_plant
     rl_config = [
         {
             'rl_algorithm': 'MLP', 
-            'time_steps': 100000,
+            'time_steps': 7000000,
             'ent_coef': 0.0,
             'n_steps': 40000,
             'nminibatches': 4,
@@ -360,8 +360,8 @@ def batch_run(self, n, rl_config, garden_x, garden_y, num_plant_types, num_plant
             "NUM_CONVS": 1,
             "FILTER_SIZE": 1,
             "STRIDE": 1,
-            'CC_COEF': 0.6,
-            'WATER_COEF': 0.4
+            'CC_COEF': 10,
+            'WATER_COEF': 100
         }
     ]
     num_evals = 50

Simulator_v2/simulatorv2/SimAlphaGardenWrapper.py

@@ -40,7 +40,8 @@ def reward(self, state):
         entropy = -np.sum(prob*np.log(prob))
         water_coef = self.config.getfloat('cnn', 'water_coef')
         cc_coef = self.config.getfloat('cnn', 'cc_coef')
-        return (cc_coef * total_cc) + (0 * entropy) - (water_coef * np.sum(self.curr_action))
+        action_sum = self.N * self.M 
+        return (cc_coef * total_cc) + (0 * entropy) + water_coef * np.sum(-1 * self.curr_action/action_sum + 1)
 
     '''
     Method called by the gym environment to execute an action.

Simulator_v2/simulatorv2/garden.py

@@ -1,8 +1,13 @@
 import numpy as np
 from simulatorv2.logger import Logger, Event
 from simulatorv2.plant import Plant
+<<<<<<< HEAD
 from simulatorv2.visualization import setup_animation
 from simulatorv2.sim_globals import MAX_WATER_LEVEL
+=======
+from visualization import setup_animation
+from sim_globals import MAX_WATER_LEVEL
+>>>>>>> 7ad33658520da88f64bf99fb211839a1ef43230e
 
 class Garden:
     def __init__(self, plants=[], N=50, M=50, step=1, drainage_rate=0.4, irr_threshold=5, plant_types=[], skip_initial_germination=True, animate=False):
@@ -290,4 +295,4 @@ def show_animation(self):
         if self.animate:
            self.anim_show() 
         else:
-            print("[Garden] No animation to show. Set animate=True when initializing to allow animating history of garden!")
+            print("[Garden] No animation to show. Set animate=True when initializing to allow animating history of garden!")