ray.rllib.algorithms.algorithm_config.AlgorithmConfig.to_dict#

AlgorithmConfig.to_dict() dict[source]#

Converts all settings into a legacy config dict for backward compatibility.

Note: If using the new API stack (enable_rl_module_and_learner=True), the returned dictionary will dynamically overwrite the legacy train_batch_size key with the calculated total_train_batch_size.

Returns:

A complete AlgorithmConfigDict, usable in backward-compatible Tune/RLlib use cases.