update_times += 1
### copy value net parameters to target net ###
if update_times % NET_COPY_STEP == 0:
target_net.load_state_dict(value_net.state_dict())
torch.save(value_net.state_dict(), PATH)
if __name__ == '__main__':
# set up matplotlib
is_ipython = 'inline' in matplotlib.get_backend()
if is_ipython:
from IPython import display
plt.ion()
learn(env=env.Env(),
MAX_EPISODE=2000000,
EPS_START=0.9,
EPS_END=0.05,
EPS_DECAY=200,
ACTION_NUM=6,
REPLAY_MEMORY_CAPACITY=10000,
BATCH_SIZE=32,
LOSS_FUNCTION=nn.SmoothL1Loss,
OPTIM_METHOD=optim.Adam,
LEARNING_RATE=1e-4,
GAMMA=0.99,
NET_COPY_STEP=1000,
OBSERVE=10000,
TRAIN_FREQ=4,
PATH='net_param.pt')
# -*- coding: utf-8 -*-
from data import env
import cv2
env = env.Env()
dis = 0.9
REPLAY_MEMORY = 50000
def main():
max_episodes = 1500
for episode in range(max_episodes):
done = False
step_count = 0
env.reset()
obs, _, _, _, _, _, _ = env.step(0)
cv2.imshow('mario', obs)
cv2.waitKey(0)
cv2.destroyAllWindows()
while not done:
state, reward, done, s4, s5, s6, s7 = env.step(11)
# 0 next_state:{ndarray} shape (90,90)
# 1 reward:{int}
# 2 done:{bool}
# 3 state_clear:{bool}
# 4 max_x:{int}
# 5 time_out:{bool}
# 6 now_x:{int}