data-designer-config

Configuration layer for NeMo Data Designer synthetic data generation framework.

This package provides the configuration API for defining synthetic data generation pipelines. It's a lightweight dependency that can be used standalone for configuration management.

Installation

pip install data-designer-config

Usage

import data_designer.config as dd

# Initialize config builder with model config(s)
config_builder = dd.DataDesignerConfigBuilder(
    model_configs=[
        dd.ModelConfig(
            alias="my-model",
            model="nvidia/nemotron-3.5-lightning-30b-a3b",
            provider="nvidia",
            inference_parameters=dd.ChatCompletionInferenceParams(temperature=0.7),
        ),
    ]
)

# Add columns
config_builder.add_column(
    dd.SamplerColumnConfig(
        name="user_id",
        sampler_type=dd.SamplerType.UUID,
        params=dd.UUIDSamplerParams(prefix="user-"),
    )
)
config_builder.add_column(
    dd.LLMTextColumnConfig(
        name="description",
        prompt="Write a product description",
        model_alias="my-model",
    )
)

# Build configuration
config = config_builder.build()

See main README.md for more information.