-
Notifications
You must be signed in to change notification settings - Fork 0
/
argument.py
25 lines (19 loc) · 1.7 KB
/
argument.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
def add_arguments(parser):
parser.add_argument('--epsilon_start', type=float, default=1, help='Initial Epsilon Value')
parser.add_argument('--epsilon_end', type=float, default=0.1, help='Final Epsilon Value')
parser.add_argument('--decay_start', type=int, default=0, help='When to start decaying epsilon')
parser.add_argument('--decay_end', type=int, default=100000, help='When to end decaying epsilon')
parser.add_argument('--learning_rate', type=float, default=0.0001, help='Learning rate for optimizer')
parser.add_argument('--buffer_size', type=int, default=300000, help='Buffer Size')
parser.add_argument('--batch_size', type=int, default=32, help='Batch size for training')
parser.add_argument('--n_episodes', type=int, default=500000, help='Episodes')
parser.add_argument('--gamma', type=float, default=0.99, help='Gamma')
parser.add_argument('--optimize_interval', type=int, default=4, help='Optimization Interval')
parser.add_argument('--target_update_interval', type=int, default=5000, help='How often to update targe network')
parser.add_argument('--evaluate_interval', type=int, default=10000, help='How often to evaluate')
parser.add_argument('--initialize_weights', type=bool, default=False, help='Initialize manually or let pytorch do it')
parser.add_argument('--clip_grad', type=bool, default=True, help='To clip gradients or not')
parser.add_argument('--load_checkpoint', type=int, default=False, help='The model to load')
parser.add_argument('--path_to_trained_model', type=str, default='trained_models/', help='Path to trained model')
parser.add_argument('--test_n', type=int, default=False, help='The experiment n to save')
return parser