def main():
"""Route 0 entry point.
"""
parser = argparse.ArgumentParser(
description='Launch an FRR network experiment in Mininet.')
parser.add_argument('-t',
'--topology',
required=True,
help='the topology of the network')
parser.add_argument('-s',
'--scenario',
required=True,
help='the scenario to set up in the network')
args = parser.parse_args()
experiment = Experiment(args.topology, args.scenario)
run(experiment)
from metrics.dmetric_reinforce_rbuf_over_time import *
from util.experiment import Experiment
for base in range(0, 10):
try:
exp = Experiment(
'diversity_over_time_rein_full_150steps_base{}'.format(base),
dry=False)
env_list = []
for i in [
base * 5, base * 5 + 1, base * 5 + 2, base * 5 + 3,
base * 5 + 4
]:
env_list.append(Gridworld(width=4, height=4, cell_size=32, seed=i))
exp.run(get_diversity,
params={
'env_list': env_list,
'training_iterations': 100,
'steps': 150,
'verbose': True
},
k=2)
except:
print('crash base=', base)
from metrics.dmetric_over_time_greed import *
from util.experiment import Experiment
exp = Experiment('diversity_greed_test_setC')
env_list = []
for j in [10, 11, 12, 13, 14]:
env_list.append(Gridworld(width=4, height=4, cell_size=32, seed=j))
exp.run(get_diversity,
params={
'env_list': env_list,
'training_iterations': 100,
'steps': 50,
'verbose': True
},
k=5)
from metrics.dmetric import *
from util.experiment import Experiment
for base in range(1, 100):
exp = Experiment('diversity_set_size_10_offset_{}_3k'.format(base))
env_list = []
for i in range(10):
env_list.append(Gridworld(width=4, height=4, cell_size=32, seed=(10 * base + i)))
exp.run(get_diversity, params={ 'env_list' : env_list, 'training_iterations' : 3000 }, k=4)
from metrics.dmetric_reinforce import *
from util.experiment import Experiment
exp = Experiment('diversity_set_b_reinforce_10k', dry=False)
env_list = []
for i in range(10, 15):
env_list.append(Gridworld(width=4, height=4, cell_size=32, seed=i))
exp.run(get_diversity,
params={
'env_list': env_list,
'training_iterations': 10000
},
k=10)
from metrics.dmetric_over_time_rbuf import *
from util.experiment import Experiment
for i in range(0, 10):
exp = Experiment('identity_test_{}_all_states_25k'.format(i))
env_list = [
Gridworld(width=4, height=4, cell_size=32, seed=i),
Gridworld(width=4, height=4, cell_size=32, seed=i)
]
exp.run(get_diversity,
params={
'env_list': env_list,
'training_iterations': 1000,
'steps': 25,
'verbose': True
},
k=1)
parser = argparse.ArgumentParser()
parser.add_argument('--model', type=absolute_path, required=True, help="Path to the model to verify (.pth)")
parser.add_argument('--num_points', type=int, default=1024, help="Number of points per point cloud")
parser.add_argument('--experiment', type=str, help="Unique identifier name for the experiment. Will be generated by default")
parser.add_argument('--theta', type=parse_theta, nargs='+', required=True, help="List of transformation parameters to certify. Either number or number followed by 'deg' for degree.")
parser.add_argument('--intervals', type=int, default=1, help="Number of parameter splits per dimension")
parser.add_argument('--transformation', type=str, default='RotationZ', help="The transformation function, or composition of multiple transformations separated by '+'")
parser.add_argument('--pooling', choices=['improved_max'], default='improved_max', help='The pooling function to use')
parser.add_argument('--relaxation', default='taylor', choices=['taylor', 'interval'], help="The relaxation to use")
parser.add_argument('--seed', type=int, default=DEFAULT_SEED, help='Seed for random number generators')
parser.add_argument('--implicit-intervals', type=int, default=1, help="Number of implicit splits for interval bound refinement")
settings = parser.parse_args()
experiment = Experiment(settings)
logger = experiment.logger
checkpoints = experiment.load_checkpoints()
transformation = parse_transformation(settings.transformation)
parameters = settings.theta
if len(parameters) != transformation.num_params:
raise ValueError(f"The number of parameters does no match those required by the transformation. Expected {transformation.num_params}, received {parameters}")
test_data = datasets.shapenet(num_points=settings.num_points, split='test', rotate='none')
num_total_classes = 50
torch_model = PointNetSegmentation(
number_points=settings.num_points,
num_seg_classes=num_total_classes,
encode_onnx=True
import numpy as np
from gw_collect import Gridworld
from util.experiment import Experiment
from common.nn import LinRegNet, LogRegNet
from vicero.algorithms.deepqlearning import DQN
from vicero.algorithms.reinforce import Reinforce
exp = Experiment('gw_collect_rein_benchmark_all2', dry=False)
def random_benchmark(env, iterations):
all_durations = []
for _ in range(iterations):
env.reset()
done = False
duration = 0
while not done:
_, _, done, _ = env.step(env.action_space.sample())
duration += 1
all_durations.append(duration)
return all_durations
def dqn_benchmark(env, iterations):
histories = []
for _ in range(iterations):
dqn = DQN(env,
qnet=LinRegNet(64, 4).double(),
choices=['improved_max', 'max', 'avg'],
default='improved_max',
help='The pooling function to use')
parser.add_argument('--max_features',
type=int,
default=1024,
help='The number of global features')
parser.add_argument('--seed',
type=int,
default=DEFAULT_SEED,
help='seed for random number generator')
parser.add_argument('--experiment', type=str, help='name of the experiment')
settings = parser.parse_args()
experiment = Experiment(settings)
logger = experiment.logger
set_random_seed(settings.seed)
test_data = datasets.modelnet40(num_points=settings.num_points,
split='test',
rotate='none')
torch_model = PointNet(number_points=settings.num_points,
num_classes=test_data.num_classes,
max_features=settings.max_features,
pool_function=settings.pooling,
disable_assertions=True,
transposed_input=True)
torch_model.load_state_dict(
torch.load(settings.model, map_location=torch.device('cpu')))
from metrics.diversity import get_diversity
from util.experiment import Experiment
from gw_collect import Gridworld
for base in range(1, 6):
env_list = []
for i in [
base * 5, base * 5 + 1, base * 5 + 2, base * 5 + 3, base * 5 + 4
]:
env_list.append(Gridworld(width=4, height=4, cell_size=32, seed=i))
# dqn full
try:
exp = Experiment('set{}_div_dqn_full'.format(base), dry=False)
exp.run(get_diversity,
params={
'env_list': env_list,
'training_iterations': 100,
'steps': 100,
'verbose': True,
'learning_alg': 'dqn',
'state_dist': 'full'
},
k=3)
except Exception as e:
print(e)
# dqn memory
try:
exp = Experiment('set{}_div_dqn_memory'.format(base), dry=False)
from metrics.dmetric_over_time import *
from util.experiment import Experiment
exp = Experiment('set_size__diversity', dry=True)
env_list = []
for j in [0, 1, 2, 3, 4]:
env_list.append(Gridworld(width=4, height=4, cell_size=32, seed=j))
exp.run(get_diversity, params={ 'env_list' : env_list, 'training_iterations' : 100, 'steps' : 50, 'verbose' : True }, k=5)