def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.s_dim = s_dim
self.a_dim = a_dim
self.pi1 = nn.Linear(s_dim, 200)
self.pi2 = nn.Linear(200, a_dim)
self.v1 = nn.Linear(s_dim, 100)
self.v2 = nn.Linear(100, 1)
set_init([self.pi1, self.pi2, self.v1, self.v2])
self.distribution = torch.distributions.Categorical
def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.s_dim = s_dim
self.a_dim = a_dim
self.a1 = nn.Linear(s_dim, 100)
self.mu = nn.Linear(100, a_dim)
self.sigma = nn.Linear(100, a_dim)
self.c1 = nn.Linear(s_dim, 100)
self.v = nn.Linear(100, 1)
set_init([self.a1, self.mu, self.sigma, self.c1, self.v])
self.distribution = torch.distributions.Normal
def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.s_dim = s_dim
self.conv1 = nn.Conv2d(3, s_dim, (1, 1), padding=0)
self.conv2 = nn.Conv2d(3, s_dim, (1, 1), padding=0)
self.pi1 = nn.Linear(s_dim * 6 * 6, 100)
self.pi2 = nn.Linear(100, a_dim)
self.v1 = nn.Linear(s_dim * 6 * 6, 100)
self.v2 = nn.Linear(100, 1)
set_init([self.pi1, self.pi2, self.v1, self.v2])
self.distribution = torch.distributions.Categorical
def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.s_dim = s_dim
self.a_dim = a_dim
# actor-net
self.pi1 = nn.Linear(s_dim, 128)
self.pi2 = nn.Linear(128, a_dim)
# critic-net
self.v1 = nn.Linear(s_dim, 128)
self.v2 = nn.Linear(128, 1)
set_init([self.pi1, self.pi2, self.v1, self.v2])
self.distribution = torch.distributions.Categorical
def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.s_dim = s_dim
self.a_dim = a_dim
self.pi1 = nn.Linear(s_dim, 100) # 4x100
self.pi2 = nn.Linear(100, a_dim) # 100x2
self.v1 = nn.Linear(s_dim, 100) # 4x100
self.v2 = nn.Linear(100, 1) # 100x1
set_init([
self.pi1, self.pi2, self.v1, self.v2
]) ## utils.py ## layer.wegiht(mean=0,std=0.1) ## layer.bias = 0.1
self.distribution = torch.distributions.Categorical
def __init__(self, action_d, observation_d):
super(Net, self).__init__()
self.action_d = action_d
self.observation_d = observation_d
self.policy_layer_1 = nn.Linear(self.observation_d, 256)
self.policy_layer_2 = nn.Linear(256, self.action_d)
self.value_layer_1 = nn.Linear(self.observation_d, 256)
self.value_layer_2 = nn.Linear(256, 1)
set_init([
self.policy_layer_1, self.policy_layer_2, self.value_layer_1,
self.value_layer_2
])
self.distribution = torch.distributions.Categorical
def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.s_dim = s_dim
self.a_dim = a_dim
# actor-net, stochastic policy
self.a1 = nn.Linear(s_dim, 200)
self.mu = nn.Linear(200, a_dim)
self.sigma = nn.Linear(200, a_dim)
# critic-net
self.c1 = nn.Linear(s_dim, 100)
self.v = nn.Linear(100, 1)
set_init([self.a1, self.mu, self.sigma, self.c1, self.v])
self.distribution = torch.distributions.Normal
def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.s_dim = s_dim
self.a_dim = a_dim
self.b1 = nn.Linear(s_dim, 32)
# self.bn1 = nn.BatchNorm1d(32, momentum=0.5)
self.b2 = nn.Linear(32, 24)
# self.bn2 = nn.BatchNorm1d(24, momentum=0.5)
# self.b3 = nn.Linear(24,16)
# self.bn3 = nn.BatchNorm1d(16, momentum=0.5)
self.pi = nn.Linear(24, a_dim)
self.v = nn.Linear(24, 1)
set_init([self.b1, self.b2, self.pi, self.v])
self.distribution = torch.distributions.Categorical
def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.a = nn.Linear(s_dim, 512)
self.mu = nn.Linear(512, a_dim)
self.sigma = nn.Linear(512, a_dim)
self.c = nn.Linear(s_dim, 256)
self.v = nn.Linear(256, 1)
# xvaier_init(self.a.weight)
# xvaier_init(self.mu.weight)
# xvaier_init(self.sigma.weight)
# xvaier_init(self.c.weight)
# xvaier_init(self.v.weight)
set_init([self.a, self.mu, self.sigma, self.c, self.v])
self.distribution = torch.distributions.Normal
def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.s_dim = s_dim
self.a_dim = a_dim
self.pi1 = nn.Linear(s_dim, 128)
self.pi12 = nn.Linear(128, 32)
self.pi13 = nn.Linear(32, 16)
self.pi2 = nn.Linear(16, a_dim)
self.v1 = nn.Linear(s_dim, 64)
self.v12 = nn.Linear(64, 16)
self.v13 = nn.Linear(16, 16)
self.v2 = nn.Linear(16, 1)
self.drop = nn.Dropout(0.5)
set_init([self.pi1, self.pi2, self.v1, self.v2])
self.distribution = torch.distributions.Categorical
def __init__(self):
super(Net, self).__init__()
#actor
self.linear_1_a = nn.Linear(S_LEN, 200)
self.linear_2_a = nn.Linear(200, 100)
self.output_a = nn.Linear(100, A_DIM)
#critic
self.linear_1_c = nn.Linear(S_LEN, 200)
self.linear_2_c = nn.Linear(200, 100)
self.output_c = nn.Linear(100, 1)
set_init([self.linear_1_a, self.linear_2_a, self.output_a,
self.linear_1_c, self.linear_2_c, self.output_c])
self.distribution = torch.distributions.Categorical
def __init__(self, s_dim, action_mappings, action_line_mappings):
super(Net, self).__init__()
self.s_dim = s_dim
self.action_mappings = action_mappings
self.action_line_mappings = action_line_mappings
self.pi1 = nn.Linear(s_dim, 896)
self.pi2 = nn.Linear(896, 896)
self.pi3 = nn.Linear(896, action_mappings.shape[0])
self.v1 = nn.Linear(s_dim, 896)
self.v2 = nn.Linear(896, 896)
self.v3 = nn.Linear(896, 1)
set_init([self.pi1, self.pi2, self.pi3, self.v1, self.v2, self.v3])
self.distribution = torch.distributions.Categorical
def __init__(self, s_dim, a_dim):
super(Net, self).__init__()
self.s_dim = s_dim
self.a_dim = a_dim
self.al1 = nn.Linear(s_dim, 200)
self.al2 = nn.Linear(200, 100)
self.al3 = nn.Linear(100, 50)
self.al4 = nn.Linear(50, a_dim)
a_ls = [self.al1, self.al2, self.al3, self.al4]
self.cl1 = nn.Linear(s_dim, 200)
self.cl2 = nn.Linear(200, 100)
self.cl3 = nn.Linear(100, 50)
self.cl4 = nn.Linear(50, 1)
cr_ls = [self.cl1, self.cl2, self.cl3, self.cl4]
set_init(a_ls + cr_ls)
self.distribution = torch.distributions.Categorical
def __init__(self, s_dim, a_dim, CNN=False):
super(A3CNet, self).__init__()
self.s_dim = s_dim
self.a_dim = a_dim
if CNN:
self.conv1 = nn.Conv2d(in_channels=3,
out_channels=6,
kernel_size=3,
stride=1,
padding=(1, 1))
self.pi1 = nn.Linear(1, 32)
self.pi2 = nn.Linear(32, 32)
self.LSTM = nn.LSTM(
input_size=32,
hidden_size=32,
num_layers=1,
)
else:
self.pi1 = nn.Linear(s_dim, 128)
self.pi2 = nn.Linear(128, a_dim)
self.v1 = nn.Linear(s_dim, 128)
self.v2 = nn.Linear(128, 1)
set_init([self.pi1, self.pi2, self.v1, self.v2])
self.distribution = torch.distributions.Categorical