-
Notifications
You must be signed in to change notification settings - Fork 1
/
main_ebm.py
186 lines (160 loc) · 6.72 KB
/
main_ebm.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
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
import numpy as np
import torch
import gym
import argparse
import os
import d4rl
import utils
import TD3
import wandb
from tqdm import tqdm
from time import time
envs_gym = {
"halfcheetah-random-v2": {'alpha': [1.0]},
"hopper-random-v2": {'alpha': [1.0]},
"walker2d-random-v2": {'alpha': [1.0]},
"halfcheetah-medium-v2": {'alpha': [1.0]},
"hopper-medium-v2": {'alpha': [1.0]},
"walker2d-medium-v2": {'alpha': [1.0]},
"halfcheetah-medium-replay-v2": {'alpha': [1.0]},
"hopper-medium-replay-v2": {'alpha': [1.0]},
"walker2d-medium-replay-v2": {'alpha': [1.0]},
"halfcheetah-medium-expert-v2": {'alpha': [1.0]},
"hopper-medium-expert-v2": {'alpha': [1.0]},
"walker2d-medium-expert-v2": {'alpha': [1.0]},
}
# Runs policy for X episodes and returns D4RL score
# A fixed seed is used for the eval environment
def eval_policy(policy, env_name, seed, mean, std, seed_offset=100, eval_episodes=10):
eval_env = gym.make(env_name)
eval_env.seed(seed + seed_offset)
avg_reward = 0.
for _ in range(eval_episodes):
state, done = eval_env.reset(), False
while not done:
state = (np.array(state).reshape(1,-1) - mean)/std
action = policy.select_action(state)
state, reward, done, _ = eval_env.step(action)
avg_reward += reward
avg_reward /= eval_episodes
d4rl_score = eval_env.get_normalized_score(avg_reward) * 100
print("---------------------------------------")
print(f"Evaluation over {eval_episodes} episodes: {avg_reward:.3f}, D4RL score: {d4rl_score:.3f}")
print("---------------------------------------")
return d4rl_score
def main(args):
file_name = f"{args.policy}_{args.env}_{args.alpha}_{args.seed}"
print("---------------------------------------")
print(f"Policy: {args.policy}, Env: {args.env}, Seed: {args.seed}")
print("---------------------------------------")
if not os.path.exists("./results"):
os.makedirs("./results")
exist_files = os.listdir('./results/')
if file_name + '.npy' in exist_files:
print(f'{file_name} exit, skipping')
# return
if not os.path.exists("./infos"):
os.makedirs("./infos")
if not os.path.exists("./ebm_models"):
os.makedirs("./ebm_models")
if args.save_model and not os.path.exists("./models"):
os.makedirs("./models")
env = gym.make(args.env)
# Set seeds
env.seed(args.seed)
env.action_space.seed(args.seed)
torch.manual_seed(args.seed)
np.random.seed(args.seed)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
max_action = float(env.action_space.high[0])
kwargs = {
"state_dim": state_dim,
"action_dim": action_dim,
"max_action": max_action,
"discount": args.discount,
"tau": args.tau,
"policy_noise": args.policy_noise * max_action,
"noise_clip": args.noise_clip * max_action,
"policy_freq": args.policy_freq,
"alpha": args.alpha,
"critic_activation": args.critic_activation,
"layer_norm": args.layer_norm,
}
# Initialize policy
policy = TD3.TD3_BC(**kwargs)
if args.load_model != "":
policy_file = file_name if args.load_model == "default" else args.load_model
policy.load(f"./models/{policy_file}")
replay_buffer = utils.ReplayBuffer(state_dim, action_dim)
replay_buffer.convert_D4RL(d4rl.qlearning_dataset(env))
if args.normalize:
mean, std = replay_buffer.normalize_states()
else:
mean, std = 0, 1
if not os.path.exists(f"./ebm_models/ebm_{args.env}.pt"):
for t in tqdm(range(int(args.ebm_steps)), desc='learning ebm'):
ebm_loss = policy.learn_ebm(replay_buffer, batch_size=args.batch_size)
if (t + 1) % 5000 == 0:
e_pos, e_neg = policy.eval_ebm(replay_buffer, batch_size=args.batch_size)
print(f'ebm_loss: {ebm_loss}, energy_pos: {e_pos}, energy_neg: {e_neg}')
if args.save_ebm:
torch.save(policy.ebm, f'./ebm_models/ebm_{args.env}.pt')
else:
policy.ebm = torch.load(f'./ebm_models/ebm_{args.env}.pt')
evaluations = []
info_log = []
t0 = time()
for t in (range(int(args.max_timesteps))):
info = policy.train(replay_buffer, args.batch_size)
# Evaluate episode
if (t + 1) % args.eval_freq == 0:
info_log.append(info)
eopch_time = (time() - t0) * 1000 / args.eval_freq
print(f"Time steps: {t + 1}, epoch_time: {eopch_time:.2f}", info)
score = eval_policy(policy, args.env, args.seed, mean, std)
evaluations.append(score)
wandb.log({'score': score})
np.save(f"./results/{file_name}", evaluations)
np.save(f"./infos/{file_name}", info_log)
if args.save_model: policy.save(f"./models/{file_name}")
t0 = time()
if __name__ == "__main__":
parser = argparse.ArgumentParser()
# Experiment
parser.add_argument("--policy", default="TD3_BC") # Policy name
parser.add_argument("--env", default="antmaze-large-play-v0") # OpenAI gym environment name
parser.add_argument("--seed", default=42, type=int) # Sets Gym, PyTorch and Numpy seeds
parser.add_argument("--eval_freq", default=5e3, type=int) # How often (time steps) we evaluate
parser.add_argument("--max_timesteps", default=2e6, type=int) # Max time steps to run environment
parser.add_argument("--save_model", action="store_true") # Save model and optimizer parameters
parser.add_argument("--load_model", default="") # Model load file name, "" doesn't load, "default" uses file_name
# TD3
parser.add_argument("--expl_noise", default=0.1) # Std of Gaussian exploration noise
parser.add_argument("--batch_size", default=256, type=int) # Batch size for both actor and critic
parser.add_argument("--discount", default=0.99) # Discount factor
parser.add_argument("--tau", default=0.005) # Target network update rate
parser.add_argument("--policy_noise", default=0.2) # Noise added to target policy during critic update
parser.add_argument("--noise_clip", default=0.5) # Range to clip target policy noise
parser.add_argument("--policy_freq", default=5, type=int) # Frequency of delayed policy updates
parser.add_argument("--normalize", default=True)
parser.add_argument("--layer_norm", default=False)
# EBPC
parser.add_argument("--alpha", default=1.0, type=float)
parser.add_argument("--ebm_steps", default=5e5, type=int) # Max time steps to run environment
parser.add_argument("--save_ebm", default=True, type=bool) # Max time steps to run environment
args = parser.parse_args()
for seed in [0]:
args.seed = seed
for env, env_config in envs_gym.items():
args.env = env
for alpha in env_config['alpha']:
args.alpha = alpha
for critic_activation in ['tanh']:
args.critic_activation = critic_activation
wandb.init(project='EBPC', reinit=True,
settings=wandb.Settings(code_dir='.', log_internal='./null', _disable_stats=True),
group='td3+ebm_tanh', mode='offline', save_code=True)
wandb.config.update(args)
print(args)
main(args)