import gym
import numpy as np
from q_learner_agent import QLearnerAgent
from replay_buffer import ReplayBuffer
env = gym.make('Pendulum-v0')
agent = QLearnerAgent(env.observation_space.shape[0], 2)
buffer = ReplayBuffer(100000)
batch_size = 32
s = env.reset()
epsilon = 0.1
episode_reward = 0
print(env.action_space)
while True:
# take action
if buffer.length() < batch_size or np.random.uniform(0, 1) < epsilon:
a = np.random.randint(0, 2)#env.action_space.n)
else:
a = agent.get_action([s])[0]
processed_action = [2.0] if a == 0 else [-2.0]
sp, r, t, _ = env.step(processed_action)
episode_reward += r
env.render()
buffer.append(s, a, r, sp, t)
if t:
s = env.reset()
print(f'Episode Reward: {episode_reward}')
episode_reward = 0
else:
s = sp
