-
Notifications
You must be signed in to change notification settings - Fork 19
/
train_ppo.py
43 lines (35 loc) · 1.16 KB
/
train_ppo.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
import argparse
import os
from datetime import datetime
from rljax.algorithm import PPO
from rljax.env import make_continuous_env
from rljax.trainer import Trainer
def run(args):
env = make_continuous_env(args.env_id)
env_test = make_continuous_env(args.env_id)
algo = PPO(
num_agent_steps=args.num_agent_steps,
state_space=env.observation_space,
action_space=env.action_space,
seed=args.seed,
)
time = datetime.now().strftime("%Y%m%d-%H%M")
log_dir = os.path.join("logs", args.env_id, f"{str(algo)}-seed{args.seed}-{time}")
trainer = Trainer(
env=env,
env_test=env_test,
algo=algo,
log_dir=log_dir,
num_agent_steps=args.num_agent_steps,
eval_interval=args.eval_interval,
seed=args.seed,
)
trainer.train()
if __name__ == "__main__":
p = argparse.ArgumentParser()
p.add_argument("--env_id", type=str, default="HalfCheetah-v3")
p.add_argument("--num_agent_steps", type=int, default=3 * 10 ** 6)
p.add_argument("--eval_interval", type=int, default=10000)
p.add_argument("--seed", type=int, default=0)
args = p.parse_args()
run(args)