-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathenjoy.py
133 lines (113 loc) · 4.04 KB
/
enjoy.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
import os
import time
from utils.yaml import read_yaml
import gym
import numpy as np
from stable_baselines3 import PPO, A2C, DQN, HER, SAC, TD3, DDPG
from stable_baselines3.common.noise import NormalActionNoise
from model.episodic_td3 import EpisodicTD3
from utils.env import env_maker, env_save, env_continue_load
import torch as th
from utils.model import policy_kwargs_building
def make_env(env_name, path, rank, seed=0):
def _init():
env = gym.make(env_name)
return env
return _init
algo = "episodic_td3"
env = "FetchReacher-v0"
env = "ALRReacherBalanceIP-v3"
#env = "dmcCheetahDense-v0"
env = 'Meta-dense-soccer-v2'
#env = 'Meta-door-open-v2'
#env = 'HopperXYJump-v0'
#env = 'HopperXYJumpStep-v0'
#env = "MetaBottonPress-v0"
env = "Meta-dense-window-open-v2"
env = "Meta-dense-soccer-v2"
env = "HopperXYJumpStep-v0"
env = "Meta-dense-window-open-v2"
#env = "Meta-dense-reach-v2"
env = "Meta-soccer-v2"
env = "Meta-dense-coffee-push-v2"
env = "Meta-button-press-v2"
env = "Meta-dense-coffee-push-v2"
env = "Meta-dense-soccer-v2"
env = "Meta-context-dense-faucet-open-v2"
env = "HopperXYJumpStep-v0"
env = "BeerPong-v0"
env = "DeepMindBallInCupDense-v0"
env_id = env
path = "logs/episodic_td3/" + env + "_2"
con = 0
if int(con) == 1:
file_name = "context.yml"
else:
file_name = "non_context.yml"
if 'Meta' in env_id:
data = read_yaml(file_name)['Meta-v2']
data['env_params']['env_name'] = data['env_params']['env_name'] + ":" + env
else:
data = read_yaml(file_name)[env_id]
# create log folder
data['path'] = path
# make the environment
env = gym.make(data["env_params"]['env_name'])
# make the model and save the model
ALGO = EpisodicTD3
critic = data['algo_params']['policy']
promp_policy_kwargs = data['promp_params']
critic_kwargs = policy_kwargs_building(data)
model = EpisodicTD3(critic, env,
initial_promp_params=data["algo_params"]['initial_promp_params'],
verbose=1,
critic_network_kwargs=critic_kwargs,
noise_sigma=data["algo_params"]['action_noise_sigma'],
promp_policy_kwargs=promp_policy_kwargs,
critic_learning_rate=data["algo_params"]['critic_learning_rate'],
actor_learning_rate=data["algo_params"]['actor_learning_rate'],
basis_num=data['promp_params']['num_basis'],
data_path=data["path"],
contextual=bool(con),
weight_noise_judge=data["algo_params"]["weight_noise_judge"],
weight_noise=data["algo_params"]["weight_noise"],
)
n_actions = env.action_space.shape[-1]
noise = NormalActionNoise(mean=np.zeros(n_actions), sigma=0 * np.ones(n_actions))
if model.contextual:
model.policy.actor.load_state_dict(th.load(path + '/best_model.pt'))
model.policy.actor.load_state_dict(th.load(path + '/algo_mean.pt'))
algorithm = np.ones((model.basis_num, model.dof))
pos_feature = None
vel_feature = None
else:
algo_path = path + "/best_model.npz"
algo_path = path + "/algo_mean.npz"
algorithm = np.load(algo_path, encoding='bytes', allow_pickle=True)
for i in algorithm:
algorithm = np.array(algorithm[i])
pos = path + "/pos_features.npz"
vel = path + "/vel_features.npz"
pos_feature = np.load(pos, encoding='bytes', allow_pickle=True)
for i in pos_feature:
pos_feature = np.array(pos_feature[i])
vel_feature = np.load(vel, encoding='bytes', allow_pickle=True)
for i in vel_feature:
vel_feature = np.array(vel_feature[i])
#print("algo", algorithm)
basis_num = 10
#algorithm = -1 * np.ones((basis_num, env.action_space.shape[0]))
#algorithm[:, 0] = 10 * np.ones(algorithm[:, 1].shape)
#algorithm[:, 2] = -0.01 #* np.ones(algorithm[:, 2].shape)
#algorithm[10:, 2] = 1 #* np.ones(algorithm[:, 2].shape)
#algorithm[30:, 2] = -1
#algorithm[50:, 2] = 1
#algorithm[70:, 2] = -1
#algorithm[70:, 2] = -1
#algorithm[110:, 2] = -1
#algorithm[90:, 2] = 1
#algorithm[150:, 2] = -1
#algorithm[170:, 2] = -1
#algorithm[190:, 2] = -1
basis_num = data['promp_params']['num_basis']
model.load(algorithm, env, pos=pos_feature, vel=vel_feature)