evorl.algorithms.erl.cemrl_td3.cemrl

Module Contents

Classes

CEMRLWorkflow	1 critic + n actors + 1 replay buffer.
EvaluateMetric

Functions

get_std_statistics

API

class evorl.algorithms.erl.cemrl_td3.cemrl.CEMRLWorkflow(**kwargs)

Bases: evorl.algorithms.erl.cemrl_td3.cemrl_td3_workflow.CEMRLTD3WorkflowTemplate

1 critic + n actors + 1 replay buffer.

We use shard_map to split and parallel the population.

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.cemrl_td3.cemrl.EvaluateMetric

Bases: evorl.metrics.MetricBase

pop_center_episode_lengths: chex.Array

None

pop_center_episode_returns: chex.Array

None

evorl.algorithms.erl.cemrl_td3.cemrl.get_std_statistics(variance)