def test(test_network):
network = network_parser.parse(test_network)
net = Net(network.state_dim, network.action_dim)
net.load_state_dict(torch.load(os.path.join('saved_model', 'subnet-10-machine-30-service-6-v0-ddqn-277.pt')))
net.eval()
print('state_dim:', network.state_dim, 'action_dim: ', network.action_dim)
num_episodes = 50
interval = 5
score_history = list()
for i in range(num_episodes):
done = False
score = 0
state = network.reset()
while not done:
state = torch.tensor(state)
state = state.unsqueeze(0)
action_values = net(state, model='online')
action = torch.argmax(action_values, axis=1).item()
# print(action)
next_state, reward, done, info = network.step(action)
state = next_state
score += reward
score_history.append(score)
avg_score = np.mean(score_history[-100:])
if i % interval == 0:
print('eipode ', i, 'score %.2f average score %.2f' % (score, avg_score))
def test(test_network):
network = network_parser.parse(test_network)
actor = Actor(network.state_dim, network.action_dim)
actor.load_state_dict(
torch.load(
os.path.join('saved_model',
'subnet-10-machine-30-service-6-v0-actor-187.pt')))
actor.eval()
print('state_dim:', network.state_dim, 'action_dim: ', network.action_dim)
num_episodes = 50
interval = 5
score_history = list()
for i in range(num_episodes):
done = False
score = 0
state = network.reset()
while not done:
state = torch.tensor(state)
state = state.unsqueeze(0)
prob = actor(state)
prob = torch.distributions.Categorical(prob)
action = prob.sample().item()
# action = torch.argmax(prob).item()
# print(action)
next_state, reward, done, info = network.step(action)
state = next_state
score += reward
score_history.append(score)
avg_score = np.mean(score_history[-100:])
if i % interval == 0:
print('eipode ', i,
'score %.2f average score %.2f' % (score, avg_score))
try:
args = parser.parse_args()
except:
parser.print_help()
exit(0)
if args.config is None:
raise Exception('Unrecognized config file.')
else:
config_path = args.config
logging.basicConfig(filename='result.log', level=logging.INFO)
logging.info("start parsing settings")
params = parse(config_path)
logging.info("finish parsing settings")
dtype = torch.float32
# Check whether a GPU is available
if torch.cuda.is_available():
device = 2#torch.device("cuda")
cuda.init()
c_device = aboutCudaDevices()
print(c_device.info())
print("selected device: ", device)
else:
device = torch.device("cpu")
print("No GPU is found")
def __init__(self, network_name, agents, plot_dir):
super().__init__()
self.network = network_parser.parse(network_name)
self.plot_dir = plot_dir
self.agents = agents
self.agent_dict = {'a2c': A2C, 'ddqn': DDQN}
print('state_dim:', network.state_dim, 'action_dim: ', network.action_dim)
num_episodes = 50
interval = 5
score_history = list()
for i in range(num_episodes):
done = False
score = 0
state = network.reset()
while not done:
state = torch.tensor(state)
state = state.unsqueeze(0)
action_values = net(state, model='online')
action = torch.argmax(action_values, axis=1).item()
# print(action)
next_state, reward, done, info = network.step(action)
state = next_state
score += reward
score_history.append(score)
avg_score = np.mean(score_history[-100:])
if i % interval == 0:
print('eipode ', i, 'score %.2f average score %.2f' % (score, avg_score))
if __name__ == '__main__':
network = network_parser.parse('subnet-10-machine-30-service-6-v0.json')
agent = DDQN(network, './plots', './saved_model')
agent.run(2000, 0)
agent.plot()
# test('subnet-10-machine-30-service-6-v0.json')
def parse_all(self, network_list):
networks = list()
for n in network_list:
network = network_parser.parse(n)
networks.append(network)
return networks