afnio.trainer

class afnio.trainer.Trainer(*, max_epochs=None, enable_checkpointing=True, enable_progress_bar=True, enable_agent_summary=True, default_root_dir=None)

Bases: object

fit(agent, train_dataloader=None, val_dataloader=None, ckpt_path=None, llm_clients=[])

Runs the full optimization routine.

Parameters:

• agent (Module) – AI agent (or flow) to fit.

• train_dataloader (Union[Iterable[Any], DataLoader, None]) – An iterable or DataLoader specifying training samples.

• val_dataloader (Union[Iterable[Any], DataLoader, None]) – An iterable or or DataLoader specifying validation samples.

• ckpt_path (Union[str, Path, None]) – Path of the checkpoint from which training is resumed. Otherwise, if there is no checkpoint file at the path, an exception is raised.

• llm_clients (Optional[List[BaseModel]]) – Optional list of LLM clients used during training. If provided this list is used to calculate the total cost of training (in USD).

Raises:

TypeError – If agent is not Module.

predict()

test(agent, test_dataloader=None, llm_clients=[])

validate(agent, val_dataloader=None, llm_clients=[])

Modules

trainer