garage.torch.policies package

PyTorch Policies.

class DeterministicMLPPolicy(env_spec, **kwargs)[source]

Bases: garage.torch.modules.mlp_module.MLPModule, garage.torch.policies.base.Policy

Implements a deterministic policy network.

The policy network selects action based on the state of the environment. It uses a PyTorch neural network module to fit the function of pi(s).

forward(input_val)[source]

Forward method.

get_action(observation)[source]

Return a single action.

get_actions(observations)[source]

Return multiple actions.

reset(dones=None)[source]

Reset policy.

class GaussianMLPPolicy(env_spec, **kwargs)[source]

Bases: garage.torch.policies.base.Policy, garage.torch.modules.gaussian_mlp_module.GaussianMLPModule

GaussianMLPPolicy.

A policy that contains a MLP to make prediction based on a gaussian distribution.

Parameters:
  • env_spec (garage.envs.env_spec.EnvSpec) – Environment specification.
  • module – GaussianMLPModule to make prediction based on a gaussian
  • distribution. –
Returns:
forward(inputs)[source]

Forward method.

get_action(observation)[source]

Get a single action given an observation.

get_actions(observations)[source]

Get actions given observations.

get_entropy(observation)[source]

Get entropy given observations.

log_likelihood(observation, action)[source]

Get log likelihood given observations and action.

reset(dones=None)[source]

Reset the environment.

vectorized

Vectorized or not.

class Policy(env_spec)[source]

Bases: abc.ABC

Policy base class without Parameterzied.

Parameters:env_spec (garage.envs.env_spec.EnvSpec) – Environment specification.
action_space

Policy action space.

get_action(observation)[source]

Get action given observation.

get_actions(observations)[source]

Get actions given observations.

observation_space

Observation space.

vectorized

Vectorized or not.