memory = Memory(100000)
if render_map:
root, canvas = init_map()
steps = 0
scores = []
epsilon = 1.0
for episode in range(hp.num_episodes):
state = env.reset()
state = pre_process(state)
history = np.stack((state, state, state, state), axis=2)
history = np.reshape([history], (84, 84, 4))
for i in range(3):
action = env.action_space.sample()
next_state, reward, done, info = env.step(action)
next_state = pre_process(next_state)
next_state = np.reshape(next_state, (84, 84, 1))
history = np.append(next_state, history[:, :, :3], axis=2)
score = 0
prev_life = 20
episode_len = 0
while True:
env.render(mode='rgb_array')
steps += 1
qvalue = model(to_tensor(history).unsqueeze(0))
action = get_action(epsilon, qvalue, num_actions)
next_state, reward, done, info = env.step(action)
video = cv2.VideoWriter('record/mob-fun.avi', cv2.VideoWriter_fourcc('M','J','P','G'), 20, (800, 600))
done = False
write = False
batch_size = 1
for i in range(1):
score = 0
count = 0
env.reset()
while True:
count += 1
env.render(mode="rgb_array")
action = env.action_space.sample()
obs, reward, done, info = env.step(action)
num_states = info['number_of_observations_since_last_state']
num_rewards = info['number_of_rewards_since_last_state']
observation = info['observation']
# print(num_states, num_rewards)
score += reward
obs = np.reshape(obs, (600, 800, 3))
img, origin_img, dim = prep_image(obs, inp_dim)
with torch.no_grad():
prediction = model(img, CUDA)
prediction = write_results(prediction, confidence, num_classes,
nms = True, nms_conf = nms_thesh)