evorl.algorithms.contrib.pop_td3

Module Contents

Classes

PopTD3Workflow	Indepentent TD3 agent with shared replay buffer.
WorkflowMetric

Functions

flatten_pop_rollout_trajectory

Flatten the trajectory from [#pop, T, B, …] to [#popTB, …].

API

class evorl.algorithms.contrib.pop_td3.PopTD3Workflow(env: evorl.envs.Env, agent: evorl.agent.Agent, optimizer: optax.GradientTransformation, evaluator: evorl.evaluators.Evaluator, replay_buffer: evorl.replay_buffers.AbstractReplayBuffer, config: omegaconf.DictConfig)

Bases: evorl.algorithms.td3.TD3Workflow

Indepentent TD3 agent with shared replay buffer.

classmethod build_from_config(config: omegaconf.DictConfig, enable_multi_devices: bool = False, enable_jit: bool = True) → typing_extensions.Self

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

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

classmethod name()

setup(key: chex.PRNGKey) → evorl.types.State

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

class evorl.algorithms.contrib.pop_td3.WorkflowMetric

Bases: evorl.metrics.MetricBase

iterations: chex.Array

‘zeros(…)’

sampled_episodes: chex.Array

‘zeros(…)’

sampled_timesteps: chex.Array

‘zeros(…)’

sampled_timesteps_per_agent: chex.Array

‘zeros(…)’

evorl.algorithms.contrib.pop_td3.flatten_pop_rollout_trajectory(trajectory: evorl.sample_batch.SampleBatch) → evorl.sample_batch.SampleBatch

Flatten the trajectory from [#pop, T, B, …] to [#popTB, …].