print("Started learning")
rewards = []
time_learned = []
n_episode = 100
time_per_n_episode = datetime.now()
total_cleared = 0
for i_episode in range(NUM_EPISODES):
stepcount = 0
state = env.reset()
done = False
current_rewards = 0
current_time_learned = []
while not done:
action = agent.choose_action(state, i_episode, train=True)
next_state, reward, done, info = env.step(action)
if reward == 1: total_cleared += 1
agent.store_transition(state, action, reward, next_state, done)
state = next_state
time_start = datetime.now()
agent.learn(BATCH_SIZE, i_episode)
time_difference = datetime.now() - time_start
current_time_learned.append(time_difference.total_seconds())
stepcount += 1
current_rewards += reward
if stepcount % 10000 == 0:
print("Started learning")
rewards = []
time_learned = []
n_episode = 100
time_per_n_episode = datetime.now()
for i_episode in range(NUM_EPISODES):
stepcount = 0
state = env.reset()
state_d = downsample(state)
done = False
current_rewards = 0
current_time_learned = []
while not done:
action = agent.choose_action(state_d, i_episode, train=True)
next_state, reward, done, info = env.step(action)
next_state_d = downsample(next_state)
agent.store_transition(state_d, action, reward, next_state_d, done)
state_d = next_state_d
time_start = datetime.now()
agent.learn(BATCH_SIZE, i_episode)
time_difference = datetime.now() - time_start
current_time_learned.append(time_difference.total_seconds())
stepcount += 1
current_rewards += reward