evorl.algorithms.erl.erl_td3.erl_eda

Module Contents

Classes

ERLEDAWorkflow	ERL w/ EDA.
EvaluateMetric

Functions

build_rl_update_fn

API

class evorl.algorithms.erl.erl_td3.erl_eda.ERLEDAWorkflow(**kwargs)

Bases: evorl.algorithms.erl.erl_td3.erl_td3_workflow.ERLTD3WorkflowTemplate

ERL w/ EDA.

Configs:

• EC: n actors

• RL: 1 (actor,critic)

• Shared replay buffer

RL will be injected into the pop mean. Support all EDA based ES algorithms.

evaluate(state: evorl.types.State) → tuple[evorl.metrics.MetricBase, evorl.types.State]

learn(state: evorl.types.State) → evorl.types.State

classmethod name()

step(state: evorl.types.State) → tuple[evorl.metrics.MetricBase, evorl.types.State]

class evorl.algorithms.erl.erl_td3.erl_eda.EvaluateMetric

Bases: evorl.metrics.MetricBase

pop_center_episode_lengths: chex.Array

None

pop_center_episode_returns: chex.Array

None

rl_episode_lengths: chex.Array

None

rl_episode_returns: chex.Array

None

evorl.algorithms.erl.erl_td3.erl_eda.build_rl_update_fn(agent: evorl.agent.Agent, optimizer: optax.GradientTransformation, config: omegaconf.DictConfig)