state_size = 8
agent = PolicyGradientAgent(state_size, action_size)
print("Training...")
train_episodes = 5000
avg_score = 0
loss = 0
for episode in range(train_episodes):
state = env.reset()
state = np.reshape(state, [1, state_size])
cum_reward = 0
for i in range(1000):
action = agent.act(state, is_training=True)
next_state, reward, done, _ = env.step(action)
next_state = np.reshape(next_state, [1, state_size])
agent.remember(state, action, reward)
state = next_state
cum_reward += reward
if done:
avg_score += cum_reward
break
current_loss = agent.update()[0]
loss += current_loss
