def __init__(self, state_dim, action_dim, hdim):
super(SimpleQFn, self).__init__()
self.Q_head = MLP(dims=[state_dim + action_dim, *hdim, 1])
def __init__(self, state_dim, hdim):
super(SimpleValueFn, self).__init__()
self.value_head = MLP(dims=[state_dim, *hdim, 1])
def __init__(self, state_dim, hdim, action_dim):
super(SimpleBetaReluPolicy, self).__init__()
self.encoder = MLP(dims=[state_dim, *hdim])
self.decoder = BetaReluParams(hdim[-1], action_dim)
self.discrete = False
def __init__(self, state_dim, hdim, action_dim):
super(SimpleBetaSoftPlusPolicy, self).__init__()
self.encoder = MLP(dims=[state_dim, *hdim])
self.decoder = BetaSoftPlusParams(hdim[-1], action_dim)
def __init__(self, state_dim, action_dim):
super(DiscreteCNNPolicy, self).__init__()
self.encoder = CNN(*state_dim)
self.decoder = MLP(
dims=[self.encoder.image_embedding_size, action_dim])
def __init__(self, state_dim, hdim, action_dim):
super(BidPolicyLN, self).__init__()
self.bid_mu = MLP(dims=[state_dim, *hdim, action_dim], zero_init=True)
self.bid_logstd = MLP(dims=[state_dim, *hdim, action_dim],
zero_init=True)
self.discrete = False
def __init__(self, state_dim, hdim, action_dim):
super(DiscretePolicy, self).__init__()
self.network = MLP(dims=[state_dim, *hdim, action_dim])