Implementation:Sdv dev SDV PARSynthesizer Sample

Knowledge Sources	SDV SDV Documentation
Domains	Time_Series, Data_Generation
Last Updated	2026-02-14 00:00 GMT

Overview

Concrete tool for generating synthetic sequential data from a fitted PARSynthesizer, provided by the SDV library.

Description

The PARSynthesizer.sample method generates synthetic sequences. It samples context rows from the internal GaussianCopulaSynthesizer, assigns sequence keys, then generates temporal data using the PARModel.

Usage

Call this method after fitting a PARSynthesizer. Specify the number of sequences to generate.

Code Reference

Source Location

• Repository: SDV
• File: sdv/sequential/par.py
• Lines: L624-650

Signature

def sample(self, num_sequences, sequence_length=None):
    """Sample new sequences.

    Args:
        num_sequences (int): Number of sequences to sample.
        sequence_length (int or None): Fixed length per sequence.
            None = variable length. Defaults to None.

    Returns:
        pandas.DataFrame: Synthetic sequential data.
    """

Import

from sdv.sequential import PARSynthesizer
# sample is called as: synthesizer.sample(num_sequences=10)

I/O Contract

Inputs

Name	Type	Required	Description
num_sequences	int	Yes	Number of sequences to generate
sequence_length	int or None	No	Fixed length per sequence (None = variable)

Outputs

Name	Type	Description
return value	pd.DataFrame	Synthetic sequential data with sequence keys

Usage Examples

# Generate 10 sequences with variable length
synthetic_data = synthesizer.sample(num_sequences=10)
print(f"Generated {synthetic_data[metadata.get_column_names(sdtype='id')[0]].nunique()} sequences")

# Generate with fixed length
fixed_data = synthesizer.sample(num_sequences=5, sequence_length=50)

Related Pages

Implements Principle

• Principle:Sdv_dev_SDV_Sequential_Data_Sampling

Requires Environment

• Environment:Sdv_dev_SDV_Python_Runtime