start_timer = time.time()
for i_episode in itertools.count(1):
# Initialise episode rewards
episode_reward = 0
episode_peak_reward = 0
episode_day_reward = 0
episode_night_reward = 0
episode_smooth_reward = 0
episode_steps = 0
done = False
state = env.reset()
temporal_state = state[:3]
state = state[3:]
state = encoder.encode_min(state)
state = temporal_state.tolist() + state
grads_G1_daily = []
grads_G1_weekly = []
grads_G2_daily = []
grads_G2_weekly = []
# For every step
while not done:
# If learning hasn't started yet, sample random action
if args.start_steps > total_numsteps:
#
action = env.action_space.sample()
agent.action_tracker.append(action)
