def run_cubic_decomp():
for n in [5, 6, 11, 35, 1021, 11111]:
print(n)
with ContextTimer():
print(cubic_decomp(n)[0])
with ContextTimer():
print(cubic_decomp_dp(n))
print()
def perf_ngsim_env_step():
basedir = os.path.expanduser('~/.julia/packages/NGSIM/9OYUa/data')
filename = 'trajdata_i101_trajectories-0750am-0805am.txt'
filepaths = [os.path.join(basedir, filename)]
env = JuliaEnv(env_id='NGSIMEnv',
env_params=dict(trajectory_filepaths=filepaths, ),
using='AutoEnvs')
n_steps = 10000
action = np.array([1., 0.])
env.reset()
with ContextTimer():
for _ in range(n_steps):
_, _, terminal, _ = env.step(action)
if terminal:
env.reset()
def step(self, action):
self.steps_since_restart += 1
self.total_steps_ever += 1
# Send a state and get a response
with ContextTimer(post_print=False) as timer:
self.server.send_state(action)
is_over, image, new_score = self._get_state()
# # Print FPS?
# if self.total_steps_ever % 1000 == 0:
# print("FPS", 1 / (timer.elapsed + .00001))
# Update the score and log info
reward = new_score - self.latest_total_score
self.latest_total_score = new_score
return image, reward, is_over, None
def perf_vectorized_ngsim_env_step():
basedir = os.path.expanduser('~/.julia/packages/NGSIM/9OYUa/data')
filename = 'trajdata_i101_trajectories-0750am-0805am.txt'
filepaths = [os.path.join(basedir, filename)]
n_envs = 100
env = JuliaEnv(env_id='VectorizedNGSIMEnv',
env_params=dict(
n_envs=n_envs,
trajectory_filepaths=filepaths,
),
using='AutoEnvs')
n_steps = 10000
action = np.zeros((n_envs, 2))
action[:, 0] = 1.
env.reset()
with ContextTimer():
for _ in range(0, n_steps, n_envs):
_, _, terminal, _ = env.step(action)
env.reset(terminal)
def collect(egoids,
starts,
args,
exp_dir,
use_hgail,
params_filename,
n_proc,
max_steps=200,
collect_fn=parallel_collect_trajectories):
'''
Description:
- prepare for running collection in parallel
- multiagent note: egoids and starts are not currently used when running
this with args.env_multiagent == True
'''
# load information relevant to the experiment
params_filepath = os.path.join(exp_dir,
'imitate/log/{}'.format(params_filename))
params = hgail.misc.utils.load_params(params_filepath)
# validation setup
validation_dir = os.path.join(exp_dir, 'imitate', 'validation')
utils.maybe_mkdir(validation_dir)
output_filepath = os.path.join(
validation_dir,
'{}_trajectories.npz'.format(args.ngsim_filename.split('.')[0]))
with ContextTimer():
trajs = collect_fn(args,
params,
egoids,
starts,
n_proc,
max_steps=max_steps,
use_hgail=use_hgail)
utils.write_trajectories(output_filepath, trajs)
for j in range(m):
for k in range(m):
if a[j, k] > a[j, i] + a[i, k]:
a[j, k] = a[j, i] + a[i, k]
return a
if __name__ == '__main__':
side = 200
random_adj_mat = np.random.rand(side * side) * side
mask = np.random.rand(side**2) < .8
random_adj_mat[mask] = np.inf
random_adj_mat = (random_adj_mat).reshape(side, side)
inputs = [[[0, np.inf, 1], [1, 0, np.inf], [np.inf, 1, 0]],
[[0, 5, np.inf, 10], [np.inf, 0, 3, np.inf],
[np.inf, np.inf, 0, 1], [np.inf, np.inf, np.inf, 0]],
random_adj_mat]
inputs = [np.array(a) for a in inputs]
expect = [[[0, 2, 1], [1, 0, 2], [2, 1, 0]],
[[0, 5, 8, 9], [np.inf, 0, 3, 4], [np.inf, np.inf, 0, 1],
[np.inf, np.inf, np.inf, 0]], 'no idea']
expect = [np.array(a) for a in expect]
for (i, e) in zip(inputs, expect):
print(e)
with ContextTimer():
print(floyd_warshall(i.copy()))
with ContextTimer():
print(floyd_warshall_fast(i.copy()))