to_dict

AlgorithmConfig.to_dict() → dict

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

Note: On the new API stack (enable_rl_module_and_learner=True), the effective batch size is derived from train_batch_size_per_learner and num_learners; read the total_train_batch_size property for it. The legacy train_batch_size key in the returned dict is not authoritative on the new stack.

Returns:

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