ray.rllib.algorithms.algorithm_config.AlgorithmConfig#

class ray.rllib.algorithms.algorithm_config.AlgorithmConfig(algo_class: type | None = None)[source]#

Bases: _Config

A RLlib AlgorithmConfig builds an RLlib Algorithm from a given configuration.

from ray.rllib.algorithms.ppo import PPOConfig
from ray.rllib.algorithms.callbacks import MemoryTrackingCallbacks
# Construct a generic config object, specifying values within different
# sub-categories, e.g. "training".
config = (
    PPOConfig()
    .training(gamma=0.9, lr=0.01)
    .environment(env="CartPole-v1")
    .env_runners(num_env_runners=0)
    .callbacks(MemoryTrackingCallbacks)
)
# A config object can be used to construct the respective Algorithm.
rllib_algo = config.build()

from ray.rllib.algorithms.ppo import PPOConfig
from ray import tune
# In combination with a tune.grid_search:
config = PPOConfig()
config.training(lr=tune.grid_search([0.01, 0.001]))
# Use `to_dict()` method to get the legacy plain python config dict
# for usage with `tune.Tuner().fit()`.
tune.Tuner("PPO", param_space=config.to_dict())

Methods

`__init__`	Initializes an AlgorithmConfig instance.
`api_stack`	Sets the config's API stack settings.
`build`	Builds an Algorithm from this AlgorithmConfig (or a copy thereof).
`build_learner`	Builds and returns a new Learner object based on settings in `self`.
`build_learner_group`	Builds and returns a new LearnerGroup object based on settings in `self`.
`callbacks`	Sets the callbacks configuration.
`checkpointing`	Sets the config's checkpointing settings.
`copy`	Creates a deep copy of this config and (un)freezes if necessary.
`debugging`	Sets the config's debugging settings.
`env_runners`	Sets the rollout worker configuration.
`environment`	Sets the config's RL-environment settings.
`evaluation`	Sets the config's evaluation settings.
`experimental`	Sets the config's experimental settings.
`fault_tolerance`	Sets the config's fault tolerance settings.
`framework`	Sets the config's DL framework settings.
`freeze`	Freezes this config object, such that no attributes can be set anymore.
`from_dict`	Creates an AlgorithmConfig from a legacy python config dict.
`from_state`	Returns an instance constructed from the state.
`get`	Shim method to help pretend we are a dict.
`get_config_for_module`	Returns an AlgorithmConfig object, specific to the given module ID.
`get_default_learner_class`	Returns the Learner class to use for this algorithm.
`get_default_rl_module_spec`	Returns the RLModule spec to use for this algorithm.
`get_evaluation_config_object`	Creates a full AlgorithmConfig object from `self.evaluation_config`.
`get_multi_agent_setup`	Compiles complete multi-agent config (dict) from the information in `self`.
`get_multi_rl_module_spec`	Returns the MultiRLModuleSpec based on the given env/spaces.
`get_rl_module_spec`	Returns the RLModuleSpec based on the given env/spaces.
`get_rollout_fragment_length`	Automatically infers a proper rollout_fragment_length setting if "auto".
`get_state`	Returns a dict state that can be pickled.
`get_torch_compile_worker_config`	Returns the TorchCompileConfig to use on workers.
`is_multi_agent`	Returns whether this config specifies a multi-agent setup.
`items`	Shim method to help pretend we are a dict.
`keys`	Shim method to help pretend we are a dict.
`learners`	Sets LearnerGroup and Learner worker related configurations.
`multi_agent`	Sets the config's multi-agent settings.
`offline_data`	Sets the config's offline data settings.
`overrides`	Generates and validates a set of config key/value pairs (passed via kwargs).
`pop`	Shim method to help pretend we are a dict.
`python_environment`	Sets the config's python environment settings.
`reporting`	Sets the config's reporting settings.
`resources`	Specifies resources allocated for an Algorithm and its ray actors/workers.
`rl_module`	Sets the config's RLModule settings.
`serialize`	Returns a mapping from str to JSON'able values representing this config.
`to_dict`	Converts all settings into a legacy config dict for backward compatibility.
`training`	Sets the training related configuration.
`update_from_dict`	Modifies this AlgorithmConfig via the provided python config dict.
`validate`	Validates all values in this config.
`validate_train_batch_size_vs_rollout_fragment_length`	Detects mismatches for `train_batch_size` vs `rollout_fragment_length`.
`values`	Shim method to help pretend we are a dict.

Attributes

`custom_resources_per_worker`
`delay_between_worker_restarts_s`
`evaluation_num_workers`
`ignore_worker_failures`
`is_atari`	True if if specified env is an Atari env.
`learner_class`	Returns the Learner sub-class to use by this Algorithm.
`max_num_worker_restarts`
`model_config`	Defines the model configuration used.
`num_consecutive_worker_failures_tolerance`
`num_cpus_for_local_worker`
`num_cpus_per_learner_worker`
`num_cpus_per_worker`
`num_envs_per_worker`
`num_gpus_per_learner_worker`
`num_gpus_per_worker`
`num_learner_workers`
`num_rollout_workers`
`recreate_failed_env_runners`
`recreate_failed_workers`
`rl_module_spec`
`total_train_batch_size`
`uses_new_env_runners`
`validate_workers_after_construction`
`worker_health_probe_timeout_s`
`worker_restore_timeout_s`