nip.parameters.trainers

nip.parameters.trainers#

Parameters for the various ML trainers.

Classes

CommonPpoParameters([loss_type, ...])

Common parameters for PPO trainers.

PureTextEiParameters([...])

Additional parameters for the Expert Iteration (EI) trainer.

PureTextMaltParameters([...])

Additional parameters for Multi-Agent LLM Training (MALT) [MSD+24].

ReinforceParameters([use_advantage_and_critic])

Additional parameters for the REINFORCE trainer.

RlTrainerParameters(frames_per_batch, ...)

Additional parameters common to all RL trainers.

SoloAgentParameters([num_epochs, ...])

Additional parameters for running agents in isolation.

SpgParameters([variant, ...])

Additional parameters for SPG [FCR20] and its variants.

TextRlParameters(...)

Additional parameters for the text-based RL trainers.

VanillaPpoParameters()

Additional parameters for the vanilla PPO trainer.