lmflow.pipeline.rm_inferencer

Attributes

logger

Classes

RewardModelInferencer

Initializes the Inferencer class with given arguments.

Module Contents

lmflow.pipeline.rm_inferencer.logger

class lmflow.pipeline.rm_inferencer.RewardModelInferencer(model_args: lmflow.args.ModelArguments, data_args: lmflow.args.DatasetArguments, inferencer_args: lmflow.args.InferencerArguments, **kwargs)

Bases: lmflow.pipeline.base_pipeline.BasePipeline

Initializes the Inferencer class with given arguments.

Parameters:

model_argsModelArguments object.

Contains the arguments required to load the model.

data_argsDatasetArguments object.

Contains the arguments required to load the dataset.

inferencer_argsInferencerArguments object.

Contains the arguments required to perform inference.

data_args

inferencer_args

model_args

local_rank

world_size

inference(model: lmflow.models.hf_text_regression_model.HFTextRegressionModel, dataset: lmflow.datasets.dataset.Dataset, transform_dataset_in_place: bool = True, use_vllm: bool = False, enable_distributed_inference: bool = False, **kwargs) → lmflow.datasets.dataset.Dataset

_inference(model: lmflow.models.hf_text_regression_model.HFTextRegressionModel, model_input: lmflow.datasets.dataset.Dataset | ray.data.Dataset, enable_distributed_inference: bool = False, **kwargs)

__inference(model: lmflow.models.hf_text_regression_model.HFTextRegressionModel, model_input: lmflow.datasets.dataset.Dataset) → list[float] | list[list[float]]

__distributed_inference(model: lmflow.models.hf_text_regression_model.HFTextRegressionModel, model_input: ray.data.Dataset, num_instances: int, batch_size: int) → list[lmflow.utils.data_utils.RewardModelInferenceResultWithInput]

abstractmethod __vllm_inference(model: lmflow.models.hf_text_regression_model.HFTextRegressionModel, model_input: list[str], enable_distributed_inference: bool = False) → list[float]

__post_process_model_output(model_output: transformers.modeling_outputs.SequenceClassifierOutputWithPast) → list[float]

flatten_list(list_of_list: list[list]) → tuple[list, list[int]]

compress_list(list_to_compress: list, sublist_lengths: list[int]) → list[list]