nemo_curator.stages.synthetic.nemo_data_designer.data_designer

Module Contents

Classes

Name	Description
`DataDesignerStage`	Data Designer stage.

Data

__all__

API

class nemo_curator.stages.synthetic.nemo_data_designer.data_designer.DataDesignerStage(
    config_builder: data_designer.config.DataDesignerConfigBuilder | None = None,
    data_designer_config_file: str | None = None,
    model_providers: list | None = None,
    verbose: bool = False
)

Dataclass

Bases: ProcessingStage[DocumentBatch, DocumentBatch]

Data Designer stage.

This class provides a Data Designer stage. To request GPUs, use: DataDesignerStage(…).with_(resources=Resources(gpus=X)).

When verbose is False (default), NeMo Data Designer (NDD) log output is suppressed (e.g. “Preview generation in progress”, “Preview complete!”) so the stage is less verbose. Set verbose=True to see full NDD logging.

Optional model_providers: pass a list of :class:data_designer.config.models.ModelProvider to use custom or test endpoints (e.g. a mock LLM server). If None, the default DataDesigner providers are used.

config_builder

DataDesignerConfigBuilder | None = None

data_designer

DataDesigner = field(init=False)

data_designer_config_file

str | None = None

model_providers

list | None = None

verbose

bool = False

nemo_curator.stages.synthetic.nemo_data_designer.data_designer.DataDesignerStage.__post_init__() -> None

nemo_curator.stages.synthetic.nemo_data_designer.data_designer.DataDesignerStage.inputs() -> tuple[list[str], list[str]]

nemo_curator.stages.synthetic.nemo_data_designer.data_designer.DataDesignerStage.outputs() -> tuple[list[str], list[str]]

nemo_curator.stages.synthetic.nemo_data_designer.data_designer.DataDesignerStage.process(
    batch: nemo_curator.tasks.DocumentBatch
) -> nemo_curator.tasks.DocumentBatch

nemo_curator.stages.synthetic.nemo_data_designer.data_designer.__all__ = ['DataDesignerStage']