cognac package
Subpackages
Submodules
cognac.centralized_dql module
- class cognac.centralized_dql.CentralizedDQNAgent(n_agents, state_size, action_size, lr=0.001, gamma=0.95, epsilon=1.0, epsilon_decay=0.995, epsilon_min=0.01)
Bases:
object
- select_action(state)
- store_experience(state, actions, reward, next_state, done)
- train()
- class cognac.centralized_dql.DeepQNetwork(state_size, action_size, hidden_size=16)
Bases:
Module
- forward(x)
Define the computation performed at every call.
Should be overridden by all subclasses.
Note
Although the recipe for forward pass needs to be defined within this function, one should call the
Module
instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.
cognac.centralized_qlearning module
- class cognac.centralized_qlearning.CentralizedQLearningAgent(n_agents, state_size, action_size, alpha=0.001, gamma=0.95, epsilon=1.0, epsilon_decay=0.9995, epsilon_min=0.01)
Bases:
object
- get_action_index(actions)
Converts binary action array to an index.
- get_state_index(state)
Converts binary state array to an index.
- select_action(state)
- update_q_table(state, actions, reward, next_state)
cognac.decentralized_reinforce module
- class cognac.decentralized_reinforce.PolicyNetwork(input_dim, output_dim)
Bases:
Module
- forward(x)
Define the computation performed at every call.
Should be overridden by all subclasses.
Note
Although the recipe for forward pass needs to be defined within this function, one should call the
Module
instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.
- class cognac.decentralized_reinforce.REINFORCEAgent(input_dim, lr=0.01)
Bases:
object
- select_action(state)
- store_transition(log_probs, reward)
- update_policy()
- cognac.decentralized_reinforce.decentralized_reinforce(env: ParallelEnv, episodes=1000)
cognac.idql module
- class cognac.idql.IndependentQLearningAgents(env, alpha=0.1, gamma=0.9, epsilon=0.1)
Bases:
object
- select_action(agent, state)
- test(episodes=10)
- test_all(init_vect_list)
- train(episodes=1000)
- update_q_table(agent, state, action, reward, next_state)
- cognac.idql.encode_state(state)
Encodes a multi-discrete state (NumPy array) as a tuple for dictionary indexing.
cognac.reinforce module
- class cognac.reinforce.PolicyNetwork(input_dim, hidden_dim=32)
Bases:
Module
- forward(state)
Define the computation performed at every call.
Should be overridden by all subclasses.
Note
Although the recipe for forward pass needs to be defined within this function, one should call the
Module
instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.