-
Notifications
You must be signed in to change notification settings - Fork 1
/
example.py
58 lines (48 loc) · 2.39 KB
/
example.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
#!/usr/bin/env python3
import argparse
from functools import partial
import torch
from ppo_pytorch.ppo import PPO, create_atari_kwargs, create_fc_kwargs
from ppo_pytorch.common import GymWrapper, AtariVecEnv, SimpleVecEnv
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='PPO runner')
parser.add_argument('--tensorboard-path', type=str, metavar='DIR', required=True,
help='tensorboard root output folder')
parser.add_argument('--env-name', type=str, metavar='ENV', required=True,
help='gym env name')
parser.add_argument('--steps', type=int, metavar='N', required=True,
help='number of executed environment steps across all actors; '
'one step is four frames for atari, one frame otherwise')
parser.add_argument('--atari', action='store_true', default=False,
help='enable for atari envs')
parser.add_argument('--no-cuda', action='store_true', default=False,
help='disable CUDA training for atari envs')
parser.add_argument('--force-cuda', action='store_true', default=False,
help='enable CUDA training for non-atari envs')
args = parser.parse_args()
assert not args.atari or args.env_name.find('NoFrameskip') != -1, \
'only NoFrameskip atari envs are supported, since library uses custom frameskip implementation.'
assert not args.no_cuda or not args.force_cuda
if args.force_cuda:
args.cuda = torch.cuda.is_available()
elif args.no_cuda:
args.cuda = False
else:
# auto selection
args.cuda = None
# parameters for `PPO` class
alg_params = create_atari_kwargs(args.steps) if args.atari else create_fc_kwargs(args.steps)
if args.cuda is not None:
alg_params.update(dict(cuda_eval=args.cuda, cuda_train=args.cuda))
rl_alg_factory = partial(PPO, **alg_params)
env_factory = partial(AtariVecEnv, args.env_name) if args.atari else partial(SimpleVecEnv, args.env_name)
gym_wrap = GymWrapper(
rl_alg_factory,
env_factory,
log_time_interval=30 if args.atari else 5,
log_path=args.tensorboard_path,
)
print('Training on {} for {} steps, CUDA {}'.format(
args.env_name, int(args.steps),
'enabled' if alg_params['cuda_train'] else 'disabled'))
gym_wrap.train(args.steps)