Implementation:Googleapis Python genai TuningDataset Setup

Knowledge Sources	googleapis/python-genai Google Gen AI Python SDK
Domains	Fine_Tuning, Data_Preparation
Last Updated	2026-02-15 00:00 GMT

Overview

Concrete tool for specifying training datasets for model fine-tuning provided by the google-genai types module.

Description

TuningDataset is a Pydantic model that specifies the training data source for a fine-tuning job. It supports three mutually exclusive sources: gcs_uri (JSONL file in Google Cloud Storage), vertex_dataset_resource (Vertex AI Multimodal Dataset), or examples (inline list of TuningExample objects). Only one source should be set per TuningDataset instance.

Usage

Create a TuningDataset with the appropriate data source. For production workflows, use gcs_uri pointing to a JSONL file. For quick experiments, use inline examples. Pass the TuningDataset to tunings.tune as the training_dataset parameter.

Code Reference

Source Location

• Repository: googleapis/python-genai
• File: google/genai/types.py
• Lines: L12114

Signature

class TuningDataset(_common.BaseModel):
    """Supervised fine-tuning training dataset."""
    gcs_uri: Optional[str] = None
    vertex_dataset_resource: Optional[str] = None
    examples: Optional[list[TuningExample]] = None

Import

from google.genai import types

I/O Contract

Inputs

Name	Type	Required	Description
gcs_uri	Optional[str]	No*	GCS URI of JSONL training data (*one source required)
vertex_dataset_resource	Optional[str]	No*	Vertex AI Dataset resource name
examples	Optional[list[TuningExample]]	No*	Inline training examples

Outputs

Name	Type	Description
TuningDataset	TuningDataset	Dataset specification to pass to tunings.tune

Usage Examples

GCS URI Dataset

from google.genai import types

dataset = types.TuningDataset(
    gcs_uri="gs://my-bucket/training_data.jsonl"
)

Inline Examples

from google.genai import types

dataset = types.TuningDataset(
    examples=[
        types.TuningExample(
            text_input="Classify: I love this product",
            output="positive"
        ),
        types.TuningExample(
            text_input="Classify: Terrible experience",
            output="negative"
        ),
        types.TuningExample(
            text_input="Classify: It was okay",
            output="neutral"
        ),
    ]
)

Related Pages

Implements Principle

• Principle:Googleapis_Python_genai_Training_Dataset_Preparation