in agents/bc_mle.py [0:0]
def __init__(self,
state_dim,
action_dim,
max_action,
device,
hidden_sizes=[256, 256],
layer_norm=False):
super(GaussianPolicy, self).__init__()
self.layer_norm = layer_norm
self.base_fc = []
last_size = state_dim
for next_size in hidden_sizes:
self.base_fc += [
nn.Linear(last_size, next_size),
nn.LayerNorm(next_size) if layer_norm else nn.Identity(),
nn.ReLU(inplace=True),
]
last_size = next_size
self.base_fc = nn.Sequential(*self.base_fc)
last_hidden_size = hidden_sizes[-1]
self.last_fc_mean = nn.Linear(last_hidden_size, action_dim)
self.last_fc_log_std = nn.Linear(last_hidden_size, action_dim)
self.device = device