-
Notifications
You must be signed in to change notification settings - Fork 4
/
save_weight.py
37 lines (31 loc) · 1.1 KB
/
save_weight.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
import os
import pickle
import subprocess
import click
import tensorflow as tf
from baselines.her.util import save_weight
@click.command()
@click.option('--policy_file', type=str, default=None)
@click.option('--run_group', type=str, default=None)
@click.option('--epoch', type=int, default=None)
def main(policy_file, run_group, epoch):
import glob
tf.compat.v1.disable_eager_execution()
if policy_file is not None:
policy_file = glob.glob(policy_file)[0]
base = os.path.splitext(policy_file)[0]
with open(policy_file, 'rb') as f:
pretrain = pickle.load(f)
pretrain_weights = save_weight(pretrain.sess)
output_file = open(base + '_weight.pkl', 'wb')
pickle.dump(pretrain_weights, output_file)
output_file.close()
else:
runs = glob.glob(f'logs/{run_group}*/*')
print(runs)
for run in sorted(runs):
policy_file = f'{run}/policy_{epoch}.pkl'
print(policy_file)
subprocess.Popen(['python', 'save_weight.py', f'--policy_file={policy_file}'])
if __name__ == '__main__':
main()