core.trainer_builder

core.trainer_builder

Builder for the training args and trainer

Classes

Name	Description
HFCausalTrainerBuilder	Build the HuggingFace training args/trainer for causal models and reward modeling
HFPPOTrainerBuilder	HF Factory class for PPO Trainer
HFRLTrainerBuilder	Trainer factory class for TRL-based RLHF trainers (e.g. DPO)
TrainerBuilderBase	Base class for trainer builder.

core.trainer_builder.HFCausalTrainerBuilder(
    self,
    cfg,
    model,
    tokenizer,
    processor=None,
)

Build the HuggingFace training args/trainer for causal models and reward modeling using TRL.

core.trainer_builder.HFPPOTrainerBuilder(
    self,
    cfg,
    model,
    tokenizer,
    processor=None,
)

HF Factory class for PPO Trainer

core.trainer_builder.HFRLTrainerBuilder(
    self,
    cfg,
    model,
    tokenizer,
    processor=None,
)

Trainer factory class for TRL-based RLHF trainers (e.g. DPO)

core.trainer_builder.TrainerBuilderBase(
    self,
    cfg,
    model,
    tokenizer,
    processor=None,
)

Base class for trainer builder.

Name	Description
get_post_trainer_create_callbacks	Callbacks added after the trainer is created, usually b/c these need access to the trainer

core.trainer_builder.TrainerBuilderBase.get_post_trainer_create_callbacks(
    trainer,
)

Callbacks added after the trainer is created, usually b/c these need access to the trainer