def __init__(self,
seed=0,
noise_start=1.0,
gamma=0.99,
t_stop_noise=30000,
action_size=2,
n_agents=2,
buffer_size=100000,
batch_size=256,
update_every=2,
noise_decay=1.0):
"""
Params
======
action_size (int): dimension of each action
seed (int): Random seed
n_agents (int): number of distinct agents
buffer_size (int): replay buffer size
batch_size (int): minibatch size
gamma (float): discount factor
noise_start (float): initial noise weighting factor
noise_decay (float): noise decay rate
update_every (int): how often to update the network
t_stop_noise (int): max number of timesteps with noise applied in training
"""
self.seed = seed
self.buffer_size = buffer_size
self.batch_size = batch_size
self.update_every = update_every
self.gamma = gamma
self.n_agents = n_agents
self.noise_weight = noise_start
self.noise_decay = noise_decay
self.t_step = 0
self.noise_on = True
self.t_stop_noise = t_stop_noise
# create two agents, each with their own actor and critic
models = [
model.Actor_Critic_Models(n_agents=n_agents)
for _ in range(n_agents)
]
self.agents = [DDPG(i, models[i], self.seed) for i in range(n_agents)]
# create shared replay buffer
self.memory = ReplayBuffer(action_size, self.buffer_size,
self.batch_size, self.seed)
def __init__(self, action_size=2, n_agents=2, seed=0):
"""
Params
======
action_size (int): dimension of each action
seed (int): Random seed
n_agents (int): number of agents
"""
self.n_agents = n_agents
self.t_step = 0
self.noise_on = True
# create two agents, each with their own actor and critic
models = [
model.Actor_Critic_Models(n_agents=n_agents)
for _ in range(n_agents)
]
self.agents = [DDPG(i, models[i]) for i in range(n_agents)]
# create shared replay buffer
self.memory = ReplayBuffer(action_size, BUFFER_SIZE, BATCH_SIZE, seed)
def __init__(self,
action_size=2,
seed=0,
n_agents=2,
buffer_size=10000,
batch_size=256,
gamma=0.99,
update_every=2):
"""
Params
======
action_size (int): dimension of each action
seed (int): Random seed
n_agents (int): number of distinct agents
buffer_size (int): replay buffer size
batch_size (int): minibatch size
gamma (float): discount factor
update_every (int): how often to update the network
"""
self.buffer_size = buffer_size
self.batch_size = batch_size
self.update_every = update_every
self.gamma = gamma
self.n_agents = n_agents
self.t_step = 0
self.noise_on = True
# create two agents, each with their own actor and critic
models = [
model.Actor_Critic_Models(n_agents=n_agents)
for _ in range(n_agents)
]
self.agents = [DDPG(i, models[i]) for i in range(n_agents)]
# create shared replay buffer
self.memory = ReplayBuffer(action_size, self.buffer_size,
self.batch_size, seed)