# evaluate
# load the weights from file
#agent.load_state_dict(torch.load('./cem_cartpole.pth'))
#agent.load_state_dict(torch.load('./cem_cartpole_5.pth')) # Path to load model from
#agent.load_state_dict(torch.load('./mimic_cav_90_.pth'))
num_episodes = num_eps
rewards = []
for i in range(num_episodes):
#
data_t = []
data_d = []
start_disp = None
#
state = env.reset()
# For Graph
#env.verbose = True
start_disp = env.center_state(env.current_states[0])
#
reward = None
t = 0
while True:
with torch.no_grad():
#env.render()
window.appendleft(torch.Tensor(state))
action_probs = agent(deque2state(env)).detach().numpy()
action = np.argmax(action_probs)
a = (env.a_max - env.a_min) * (
(action) / (agent.action_size - 1)) + env.a_min
# CAV Simulator (Generates Fake Data now)
env = Simulator(num_leading_vehicle, num_following_vehicle)
env.normalize = False
#env.verbose = True
num_episodes = num_eps
rewards = []
for i in range(num_episodes):
#
data_t = []
data_d = []
start_disp = None
#
s = env.reset()
#
env.normalize = True
start_disp = env.center_state(env.current_states[0])
env.normalize = False
#
done = 0
i = 0
reward = None
while not done:
#print(env.t)
#env.render()
# For graph
add2loc_map(env)
#print(s)
plt.xlabel("Time")
plt.show()
env = Simulator(num_leading_vehicle,num_following_vehicle)
env.normalize = False
#env.verbose = True
num_episodes = num_eps
results = []
for i in range(num_episodes):
#
data_t = []
data_d = []
start_disp = None
#
s = env.reset()
#
env.normalize = True
start_disp = env.center_state(env.current_states[0])
env.normalize = False
#
done = 0
i = 0
reward = None
while not done:
#print(env.t)
#env.render()
# For graph
add2loc_map(env)
#print(s)