-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_hparams.py
51 lines (42 loc) · 1.25 KB
/
run_hparams.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
from itertools import product
from rltoolkit import A2C, EvalsWrapper
import multiprocessing as mp
ALGO = A2C
EVALS = 1
ITERATIONS = 200
RETURN_DONE = 190
GAMMA = [0.99]
A_LR = [1e-3, 3e-3, 1e-2, 3e-2]
C_LR = [1e-3, 3e-3, 1e-2, 3e-2]
BATCH_SIZE = [50, 200, 500]
TENSORBOARD_DIR = "tensorboard"
LOG_DIR = "basic_logs"
N_CORES = 4
def run_combination(*args, **kwargs):
evals = EvalsWrapper(*args, **kwargs)
evals.perform_evaluations()
evals.update_tensorboard()
def apply_kwargs(fn, kwargs):
return fn(**kwargs)
if __name__ == "__main__":
combinations = product(A_LR, C_LR, BATCH_SIZE, GAMMA)
kwargs_list = []
for a_lr, c_lr, batch_size, gamma in combinations:
kwargs = {
"Algo": ALGO,
"evals": EVALS,
"iterations": ITERATIONS,
"gamma": gamma,
"actor_lr": a_lr,
"critic_lr": c_lr,
"batch_size": batch_size,
"tensorboard_dir": TENSORBOARD_DIR,
"return_done": RETURN_DONE,
"log_dir": LOG_DIR,
"verbose": 0,
"render": False,
"test_episodes": 5,
}
kwargs_list.append((run_combination, kwargs))
with mp.Pool(N_CORES) as p:
p.starmap(apply_kwargs, kwargs_list)