def play():
_, get_action = load_policy(SAVE_PATH)
env = SawyerGraspEnv(n_substeps=5)
n_episode = 10
ep_len, ep_ret, i = 0, 0, 0
max_ep_len = 1000
obs = env.reset()
while i < n_episode:
action = get_action(obs)
print('Action: {}'.format(action))
obs, r, d, reward_info = env.step(action)
# print('Control: {}'.format(env.sim.data.ctrl))
ep_len += 1
ep_ret += r
print('dist_reward: {} grasp_reward: {} terminal_reward: {}'.format(reward_info['dist_reward'],
reward_info['grasp_reward'],
reward_info['terminal_reward']))
env.render()
if d or (ep_len == max_ep_len):
print('DONE: Episode Length: {} Episode Reward {}'.format(ep_len, ep_ret))
obs = env.reset()
ep_len, ep_ret, r = 0, 0, 0
d = False
i += 1
def main():
parser = argparse.ArgumentParser()
parser.add_argument(
"--file",
type=str,
default="",
)
args = parser.parse_args()
_, get_action = load_policy(args.file)
env = gym.make('HumanoidRL-v0')
run_policy(env, get_action)
def play():
_, get_action = load_policy(save_path)
n_episode = 10
ep_len, ep_ret, i = 0, 0, 0
max_ep_len = 500
obs = env.reset()
while i < n_episode:
action = get_action(obs)
# print('Action: {}'.format(action))
obs, r, d, reward_info = env.step(action)
# print('Control: {}'.format(env.sim.data.ctrl))
ep_len += 1
ep_ret += r
env.render()
if d or (ep_len == max_ep_len):
print('DONE: Episode Length: {} Episode Reward {}'.format(
ep_len, ep_ret))
obs = env.reset()
ep_len, ep_ret, r = 0, 0, 0
d = False
i += 1
from spinup.utils.test_policy import load_policy, run_policy, \
load_policy_and_env
# Register custom envs
import gym_match_input_continuous
import deepdrive_zero
import gym
TEST_STATIC_OBSTACLE = False
# TODO: Move configs from env to python that train and test both use
if TEST_STATIC_OBSTACLE:
_, get_action = load_policy(
'/home/c2/src/spinningup/data/dd2d-ppo-intersection/dd2d-ppo-intersection_s0',
use_model_only=False)
env = gym.make('deepdrive-2d-static-obstacle-no-g-pen-v0')
else:
p = '/home/c2/src/tmp/spinningup/data/deepdrive-2d-intersection-no-constrained-controls-example/deepdrive-2d-intersection-no-constrained-controls-example_s0_2020_03-10_13-14.50/best_HorizonReturn/2020_03-11_11-36.27'
if 'no-end-g' in p or 'no-contraint-g' in p or 'no-g' in p or 'no-constrain' in p:
os.environ['END_ON_HARMFUL_GS'] = '0'
os.environ['GFORCE_PENALTY_COEFF'] = '0'
os.environ['JERK_PENALTY_COEFF'] = '0'
if 'no-constrain' in p:
os.environ['CONSTRAIN_CONTROLS'] = '0'
if 'delta-controls' in p or 'deepdrive-2d-intersection-no-g-or-jerk2' in p:
os.environ['EXPECT_NORMALIZED_ACTION_DELTAS'] = '1'
else:
os.environ['EXPECT_NORMALIZED_ACTION_DELTAS'] = '0'
from spinup.utils.test_policy import load_policy, run_policy modelpath = '/home/watate/[email protected]/Python/spinningup/data/ex2_ddpg_10ep_mountaincarcontinuous-v0/ex2_ddpg_10ep_mountaincarcontinuous-v0_s0' len = 0 episodes = 100 norender = False itr = -1 #Only for soft-actor critic deterministic = False # This part is unecessary because load_policy already restores tf_graph #model = spinup.utils.logx.restore_tf_graph(sess, modelpath) env, get_action = load_policy(modelpath, itr if itr >= 0 else 'last', deterministic) #run_policy(env, get_action, len, episodes, not(norender)) ''' for i_episode in range(20): observation = env.reset() while(True): #for t in range(100): env.render() action = env.action_space.sample() observation, reward, done, info = env.step(action) env.close() ''' for i_episode in range(episodes): observation = env.reset()
from spinup.utils.test_policy import load_policy, run_policy
import gym
import gym_foo
_, get_action = load_policy('./output')
env = gym.make('pro-v0')
run_policy(env, get_action)
from spinup.utils.test_policy import load_policy, run_policy
import envs
import gym
policy_name = 'ppo-100e-012-imgclass_as_input'
_, get_action = load_policy('./data/{}/{}_s0'.format(policy_name, policy_name))
env = gym.make('MNISTClassEnv-v0', desired_outputs=[0, 1, 2])
run_policy(env, get_action)
def __init__(self):
_, self.get_action = load_policy('log2/')
from spinup.utils.test_policy import load_policy, run_policy
import gym
import gym_geofriend2
_, get_action = load_policy('./spinupPpo')
env = gym.make("geofriend2-v0") #Pyramid(),
run_policy(env, get_action, max_ep_len=200)
from spinup.utils.test_policy import load_policy, run_policy
import rocket_lander_gym
import gym
_, get_action = load_policy('oupput')
env = gym.make('RocketLander-v0')
run_policy(env, get_action)
from spinup.utils.test_policy import load_policy, run_policy import mycartCont import sys dirname = sys.argv[1:] _, get_action = load_policy(dirname[0]) env = mycartCont.MyCartContEnv() run_policy(env, get_action)
from spinup.utils.test_policy import load_policy, run_policy
_, get_action = load_policy("output_dir2")
import numpy as np
import train
double = train.Doubling()
o = np.array([
0, 2, 0, 0, 0, 3, 0, 3, 0, 0, 1, 0, 3, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0,
0, 0, 0, 0, 0, 5, 0, 3, 0, 0, 0, 1, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0
])
a = get_action(o)
print(a)
#run_policy(double, get_action)
from spinup.utils.test_policy import load_policy, run_policy
import gym
import argparse
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('--env', type=str, required=True)
parser.add_argument('--save_dir', type=str, required=True)
parser.add_argument("-d", '--deterministic', action="store_true")
parser.add_argument("-e", "--episodes", type=int, default=100)
parser.add_argument("-dr", "--dont_render", action="store_false")
args = parser.parse_args()
if args.dont_render == False:
render_it = False
else:
render_it = True
print(args.save_dir)
_, get_action = load_policy(args.save_dir,
deterministic=args.deterministic)
env = gym.make(args.env)
run_policy(env, get_action, num_episodes=args.episodes, render=render_it)
