from baselines.run import main
args = [
'--num_timesteps',
'50e6',
'--env',
'BreakoutNoFrameskip-v4',
'--alg',
'deepq',
]
main(args)
'10',
'--num_env',
'16',
'--v_net',
'atari',
'--collect_until',
'1e6',
'--vae_buffer_size',
'5e4',
'--vae_batch_size',
'1024',
'--vae_batches_per_epoch',
'20',
'--seed',
str(0),
'--k',
str(k),
'--rl_coef',
str(rlc),
'--tensorboard',
'True',
'--log_interval',
'1',
]
main(args, build_fn=build_pend_env, vae_params=vae_params)
s = get_session()
s.close()
tf.reset_default_graph()
print('done')
'--seed',
str(0),
'--k',
str(k),
'--load_path',
policy_path,
'--play',
'True',
]
v = VAE(load_from=vae_name,
network='atari',
with_opt=False,
session=tf.Session())
model, env = main(args,
build_fn=build_pend_env,
vae_params=vae_params,
just_return=True)
viewer = rendering.SimpleImageViewer()
obs = env.reset()
d = False
max_t = 25e3
t = 0
last_t = 0
max_t_ep = 2000
num_ep = 0
# buffers to be saved
org_buf = []
args.num_env or 1,
seed,
reward_scale=args.reward_scale,
flatten_dict_observations=flatten_dict_observations)
return VecVAEStack(env, k=k, load_from=vae_name.replace('/', ':'))
args = [
'--env',
'PendulumVisual-v0',
'--num_timesteps',
'8e6',
'--alg',
'trpo_mpi',
'--network',
'mlp',
'--num_env',
'16',
'--seed',
str(seed),
'--k',
str(k),
'--tensorboard',
'True',
]
main(args, build_fn=build_pend_env)
s = get_session()
s.close()
tf.reset_default_graph()
pass
from baselines.run import main
args = [
'--env',
'BreakoutNoFrameskip-v4',
'--alg',
'ppo2',
'--num_timesteps',
'10e7',
'--num_env',
'1',
'--log_interval',
'1',
'--seed',
'0',
'--load_path',
'/Users/llach/breakout-nenv16-rlc1.0-seed0-modeld-2019-04-17T20:39/best/',
'--play',
'True',
]
model, env = main(args)
import sys
import baselines.run as run
import highway_env
DEFAULT_ARGUMENTS = [
"--env=highway-parking-v0", "--alg=her", "--num_timesteps=1e4",
"--network=default", "--num_env=0", "--save_path=~/models/latest",
"--load_path=~/models/latest", "--save_video_interval=0", "--play"
]
if __name__ == "__main__":
args = sys.argv
if len(args) <= 1:
args = DEFAULT_ARGUMENTS
run.main(args)
import sys, os
import baselines.run as run
os.environ['OPENAI_LOG_FORMAT'] = 'stdout,log,tensorboard'
os.environ['OPENAI_LOGDIR'] = '../_logdar/'
if __name__ == '__main__':
run.main([
'main.py', '--alg=ppo_moba', '--env=gym_moba:moba-multiplayer-v0',
'--network=multi_unit_mlp', '--num_timesteps=1e7', '--scene_id=13',
'--is_train'
])
'''
run.main(['main.py', '--alg=ppo2', '--env=Breakout-v0',
'--network=cnn', '--num_timesteps=2e7', '--scene_id=10', '--is_train'])
'''
# f = open("../models/test.out", 'w')
# sys.stdout = f
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' # for remove TF warning
DEFAULT_ARGUMENTS = [
"--env=highway-continuous-v0",
"--alg=trpo_mpi",
"--num_timesteps=1e6", # episode * steps = num_timesteps = 1e6
# policy net parameter
"--network=mlp",
"--num_layers=3",
"--num_hidden=124",
"--activation=tf.tanh",
"--num_env=0", # >1 for mpi, disabled for online learning
"--env_json=C:/Users/szhan117/Documents/git_repo/highway-env/scripts/config/IDM.json",
# last save name must be 'latest'
"--save_path=C:/Users/szhan117/Documents/git_repo/highway-env/models/latest",
# "--load_path=C:/Users/szhan117/Documents/git_repo/highway-env/models/latest",
"--save_video_interval=0",
"--play"
]
if __name__ == "__main__":
args = sys.argv
if len(args) <= 1:
args = DEFAULT_ARGUMENTS
run.main(args) # for training
# run.animation(args) # for animation
# f.close()
import sys
import r2048_rl.envs
import gym
from baselines import run
if __name__ == '__main__':
run.main(sys.argv)
import gym_balls # pylint: Add to register gym-balls environments
from baselines.run import main
if __name__ == '__main__':
main()
# logpath = '--log_path=./models/block/harsh_65/IR/fpp_demo25bad{}dim_{}_log'.format(dim,seed)
# perturb = '--perturb=delay'
# algdim = '--algdim={}'.format(dim)
# # if seed >= 100 and seed < 1000: seed = 10
# # elif seed >= 1000 and seed < 10000: seed = 100
# # elif seed >= 10000 and seed < 100000: seed = 1000
# finalargs = defaultargs + [savepath, demofile, logpath, perturb, algdim, '--seed={}'.format(seed)]
# run.main(finalargs)
defaultargs = ['--alg=her', '--env=NuFingers', '--num_timesteps=1e6']
for dim in [4]:
for seed in [10, 100, 1000]:
savepath = '--save_path=./models/NuFingers/IR/fpp_demo25bad{}dim_{}'.format(
dim, seed)
loadpath = '--load_path=./models/NuFingers/Sim_NuFingers_bad{}dim_1000'.format(
dim)
demofile = '--demo_file=./NuFingersObjectDemo_bad{}dim.npz'.format(dim)
logpath = '--log_path=./models/NuFingers/IR/fpp_demo25bad{}dim_{}_log'.format(
dim, seed)
perturb = '--perturb=none'
algdim = '--algdim={}'.format(dim)
finalargs = defaultargs + [
savepath, loadpath, demofile, logpath, perturb, algdim,
'--seed={}'.format(seed)
]
run.main(finalargs)
def run_process(args, output):
res = main(args)
output.put(res)
args = [
'--env', 'BreakoutNoFrameskip-v4',
'--alg', 'ppo2',
'--num_timesteps', '10e7',
'--num_env', '1',
'--log_interval', '1',
'--seed', '0',
'--load_path', f'{home}/breakout-ppo/',
'--play', 'True',
]
frames = [0, 588]
t = 0
model, env = main(args, just_return=True)
obs = env.reset()
d = False
log.info('generating frames')
while not d:
fr = env.render(mode='rgb_array')
if t in frames:
scipy.misc.imsave(f'{figure_path}/frame{t}.png', fr)
print(t)
t += 1
actions, _, _, _ = model.step(obs)
import sys
from baselines.run import main
import rl_od
if __name__ == '__main__':
main(sys.argv)
import gym_turtlebot3
import rospy
import baselines.run as run
import os
env_name = 'TurtleBot3_Circuit_Simple-v0'
alg = 'deepq'
num_timesteps = '1e4'
name_ref = env_name + '_' + num_timesteps + '_' + alg
my_args = [
'--alg=' + alg,
'--env=' + env_name,
'--save_path=./models/' + name_ref + '.pkl',
'--num_timesteps=' + num_timesteps]
os.environ["OPENAI_LOG_FORMAT"] = "csv"
os.environ["OPENAI_LOGDIR"] = './logs/' + name_ref
rospy.init_node(env_name.replace('-', '_'))
run.main(my_args)