def launch_server(host, port):
# to avoid cluttering the terminal...
if rank == 0:
print("launch with a lot of output, this is rank 0")
tensorforce_environment = resume_env(plot=False, step=100, dump=100)
else:
print("Launch with less output, this is higher rank")
# tensorforce_environment = resume_env(plot=False, step=100, dump=100)
# still dump to ensure the CSVs are present, this will look messy in command line...
tensorforce_environment = resume_env(plot=False, step=100, dump=100)
RemoteEnvironmentServer(tensorforce_environment, host=host, port=port)
def launch_server(host, port, verbose=0):
tensorforce_environment = resume_env(rank)
RemoteEnvironmentServer(tensorforce_environment,
host=host,
port=port,
verbose=verbose)
def launch_server(host, port):
# to avoid cluttering the terminal...
if rank == 0:
print("launch with a lot of output, this is rank 0")
tensorforce_environment = resume_env(plot=False, dump_debug=1)
RemoteEnvironmentServer(tensorforce_environment, host=host, port=port)
else:
print("Launch with less output, this is higher rank")
tensorforce_environment = resume_env(plot=False,
dump_debug=1,
dump_CL=False) # No CL output
RemoteEnvironmentServer(tensorforce_environment,
host=host,
port=port,
verbose=0)
'''
Perform a single run of the flow without control
'''
import os
import socket
import numpy as np
import csv
from tensorforce.agents import Agent
from tensorforce.execution import ParallelRunner
from env import resume_env, nb_actuations, simulation_duration
example_environment = resume_env(plot=False,
dump_CL=100,
dump_debug=1,
dump_vtu=50,
single_run=True)
deterministic = True
network = [dict(type='dense', size=512), dict(type='dense', size=512)]
if (os.path.exists("saved_models/test_strategy.csv")):
os.remove("saved_models/test_strategy.csv")
if (os.path.exists("saved_models/test_strategy_avg.csv")):
os.remove("saved_models/test_strategy_avg.csv")
def one_run():
import env
import os
import csv
"""
import sys
import os
cwd = os.getcwd()
sys.path.append(cwd + "/../Simulation/")
from Env2DCylinder import Env2DCylinder
"""
printi("resume env")
environment = resume_env(plot=500, dump=10, single_run=True)
deterministic=True
printi("define network specs")
network_spec = [
dict(type='dense', size=512),
dict(type='dense', size=512),
]
printi("define agent")
printiv(environment.states)
printiv(environment.actions)
printiv(network_spec)
import env env.resume_env(plot=500, remesh=True)
import env env.resume_env(plot=False, remesh=True)
import env env.resume_env(plot=False, remesh=True, dump_debug=1, dump_vtu=50)
import numpy as np
from tensorforce.agents import PPOAgent
from tensorforce.execution import Runner
"""
import sys
import os
cwd = os.getcwd()
sys.path.append(cwd + "/../Simulation/")
from Env2DCylinder import Env2DCylinder
"""
printi("resume env")
environment = resume_env(plot=False, step=50, dump=100)
printi("define network specs")
network_spec = [
dict(type='dense', size=512),
dict(type='dense', size=512),
]
printi("define agent")
printiv(environment.states)
printiv(environment.actions)
printiv(network_spec)
agent = PPOAgent(
import numpy as np
from tensorforce.agents import PPOAgent
from tensorforce.execution import Runner
"""
import sys
import os
cwd = os.getcwd()
sys.path.append(cwd + "/../Simulation/")
from Env2DCylinder import Env2DCylinder
"""
printi("resume env")
environment = resume_env(plot=200, step=100, dump=100)
printi("define network specs")
network_spec = [
dict(type='dense', size=512),
dict(type='dense', size=512),
]
printi("define agent")
printiv(environment.states)
printiv(environment.actions)
printiv(network_spec)
agent = PPOAgent(
from tensorforce.execution import Runner
import env
import os
import csv
"""
import sys
import os
cwd = os.getcwd()
sys.path.append(cwd + "/../Simulation/")
from Env2DCylinder import Env2DCylinder
"""
printi("resume env")
environment = resume_env(dump=10, single_run=True)
deterministic = True
printi("define network specs")
network_spec = [
dict(type='dense', size=512),
dict(type='dense', size=512),
]
printi("define agent")
printiv(environment.states)
printiv(environment.actions)
printiv(network_spec)
#
# The default running time steps are used for Re=200, and it may run a
# little long time, you can change the value of max_nbr_actuations
#
#----------------------------------------------------------------
import os
import socket
import numpy as np
import csv
from tensorforce.agents import Agent
from tensorforce.execution import ParallelRunner
from env import resume_env, nb_actuations
example_environment = resume_env(plot=False, dump=100, single_run=True)
deterministic = True
network = [dict(type='dense', size=512), dict(type='dense', size=512)]
agent = Agent.create(
# Agent + Environment
agent='ppo', environment=example_environment, max_episode_timesteps=nb_actuations,
# TODO: nb_actuations could be specified by Environment.max_episode_timesteps() if it makes sense...
# Network
network=network,
# Optimization
batch_size=20, learning_rate=1e-3, subsampling_fraction=0.2, optimization_steps=25,
# Reward estimation
likelihood_ratio_clipping=0.2, estimate_terminal=True, # ???