def __init__(self, args, input_dim, num_actions):
nn.Module.__init__(self)
self.conv1 = nn.Conv2d(input_dim, 32, 8, stride=4, padding=1)
self.conv2 = nn.Conv2d(32, 64, 4, stride=2)
self.conv3 = nn.Conv2d(64, 64, 3)
self.fc1 = BayesBackpropLinear(3136, 512)
self.fc2 = BayesBackpropLinear(512, num_actions)
initialize_weights(self)
def __init__(self, args, action_space):
nn.Module.__init__(self)
self.features = nn.Sequential(
nn.Linear(args.input_dim, 16),
nn.ReLU(inplace=True),
nn.Linear(16, 16),
nn.ReLU(inplace=True)
)
self.last_layer = nn.Linear(16, action_space)
initialize_weights(self)
def __init__(self, args, input_dim, num_actions):
nn.Module.__init__(self)
self.conv1 = nn.Conv2d(input_dim, 32, 8, stride=4, padding=1)
self.conv2 = nn.Conv2d(32, 64, 4, stride=2)
self.conv3 = nn.Conv2d(64, 64, 3)
self.fc1 = MNFLinear(3136, 512, args.hidden_dim, args.n_hidden,
args.n_flows_q, args.n_flows_r)
self.fc2 = MNFLinear(512, num_actions, args.hidden_dim, args.n_hidden,
args.n_flows_q, args.n_flows_r)
initialize_weights(self)
def __init__(self, args, action_space):
nn.Module.__init__(self)
# self.features = nn.Sequential(
# nn.Linear(args.input_dim, 16),
# nn.ReLU(inplace=True)
# )
self.nheads = args.nheads
self.heads = nn.ModuleList([nn.Sequential(nn.Linear(args.input_dim, 16),
nn.ReLU(inplace=True),
nn.Linear(16, 16),
nn.ReLU(inplace=True),
nn.Linear(16, action_space)) for _ in range(args.nheads)])
initialize_weights(self)
def __init__(self, args, input_dim, num_actions):
nn.Module.__init__(self)
self.features = nn.Sequential(
nn.Conv2d(input_dim, 32, 8, stride=4, padding=1),
nn.ReLU(inplace=True),
nn.Conv2d(32, 64, 4, stride=2),
nn.ReLU(inplace=True),
nn.Conv2d(64, 64, 3),
nn.ReLU(inplace=True)
)
self.nheads = args.nheads
self.heads = nn.ModuleList([nn.Sequential(nn.Linear(3136, 512),
nn.ReLU(inplace=True),
nn.Linear(512, num_actions)) for _ in range(args.nheads)])
initialize_weights(self)
def __init__(self, input_dim, hidden_dim, n_hidden):
nn.Module.__init__(self)
self.input_dim = input_dim
self.hidden_dim = hidden_dim
self.n_hidden = n_hidden
self.first_layer = nn.Sequential(nn.Linear(input_dim, hidden_dim),
nn.Tanh())
hidden_modules = []
for _ in range(n_hidden):
hidden_modules.append(nn.Linear(hidden_dim, hidden_dim))
hidden_modules.append(nn.Tanh())
self.hidden_layer = nn.Sequential(*hidden_modules)
self.mu_layer = nn.Linear(hidden_dim, input_dim)
self.sigma_layer = nn.Sequential(nn.Linear(hidden_dim, input_dim),
nn.Sigmoid())
self.register_buffer('mask', torch.Tensor(input_dim))
initialize_weights(self)