`evorl.algorithms.erl.erl_td3.erl_td3_workflow`¶

Module Contents¶

Classes¶

ERLTD3WorkflowTemplate

A template for ERL workflow on TD3 Agent.

Functions¶

`build_erl_rl_update_fn`	K (actor, critic) pairs.
`create_dummy_td3_trainmetric`
`erl_replace_td3_actor_params`
`rollout_episode`

Data¶

DUMMY_TD3_TRAINMETRIC

API¶

evorl.algorithms.erl.erl_td3.erl_td3_workflow.DUMMY_TD3_TRAINMETRIC¶: ‘TD3TrainMetric(…)’

class evorl.algorithms.erl.erl_td3.erl_td3_workflow.ERLTD3WorkflowTemplate(**kwargs)[source]¶

Bases: evorl.algorithms.erl.erl_workflow.ERLWorkflowBase

A template for ERL workflow on TD3 Agent.

LOGGING_WARMUP_FLAG¶: True

setup(key: chex.PRNGKey) → evorl.types.State[source]¶

warmup_step(state: evorl.types.State) → tuple[evorl.algorithms.erl.erl_workflow.ERLTrainMetric, evorl.types.State][source]¶

evorl.algorithms.erl.erl_td3.erl_td3_workflow.build_erl_rl_update_fn(agent: evorl.agent.Agent, optimizer: optax.GradientTransformation, config: omegaconf.DictConfig, agent_state_vmap_axes: evorl.agent.AgentState)[source]¶: K (actor, critic) pairs.

evorl.algorithms.erl.erl_td3.erl_td3_workflow.create_dummy_td3_trainmetric(num: int) → evorl.algorithms.td3.TD3TrainMetric[source]¶

evorl.algorithms.erl.erl_td3.erl_td3_workflow.erl_replace_td3_actor_params(agent_state: evorl.agent.AgentState, pop_actor_params: evorl.algorithms.td3.TD3NetworkParams) → evorl.agent.AgentState[source]¶

evorl.algorithms.erl.erl_td3.erl_td3_workflow.rollout_episode(agent_state: evorl.agent.AgentState, replay_buffer_state, key, *, collector, replay_buffer, agent_state_vmap_axes, num_episodes, num_agents)[source]¶