About training PopSAN with TD3 algorithm

[Banzang]

Hi, I‘m interested in your work, But I have some confusion when trying td3_cuda_norm.py.
I don't quite understand what parameter “encoder_var” and parameter “mean_range” mean. And why did you choose 0.15 and (-3, 3) as the input values？
Thank you very much！

if name == 'main':
import math
import argparse

parser = argparse.ArgumentParser()
parser.add_argument('--env', type=str, default='Ant-v3')
parser.add_argument('--encoder_pop_dim', type=int, default=10)
parser.add_argument('--decoder_pop_dim', type=int, default=10)
**parser.add_argument('--encoder_var', type=float, default=0.15)**
parser.add_argument('--start_model_idx', type=int, default=0)
parser.add_argument('--num_model', type=int, default=10)
parser.add_argument('--epochs', type=int, default=100)
args = parser.parse_args()

START_MODEL = args.start_model_idx
NUM_MODEL = args.num_model
USE_POISSON = False
if args.env == 'Hopper-v3' or args.env == 'Walker2d-v3':
    USE_POISSON = True
AC_KWARGS = dict(hidden_sizes=[256, 256],
                 encoder_pop_dim=args.encoder_pop_dim,
                 decoder_pop_dim=args.decoder_pop_dim,
                 **mean_range=(-3, 3),**
                 std=math.sqrt(args.encoder_var),
                 spike_ts=5,
                 device=torch.device('cuda'),
                 use_poisson=USE_POISSON)
COMMENT = "td3-popsan-" + args.env + "-encoder-dim-" + str(AC_KWARGS['encoder_pop_dim']) + \
          "-decoder-dim-" + str(AC_KWARGS['decoder_pop_dim'])