Skip to main content

​
export_pipeline_to_onnx

def export_pipeline_to_onnx(
    pipeline,
    output_path: Path,
    n_features: int
) -> bool
Exports a scikit-learn pipeline to ONNX format for deployment in various runtime environments.
​
pipeline
sklearn.pipeline.Pipeline
required
Trained scikit-learn pipeline to export. Should contain preprocessing steps and model
​
output_path
Path
required
File path where the ONNX model will be saved. Parent directories will be created if they don’t exist
​
n_features
int
required
Number of input features expected by the pipeline
​
return
bool
  • True if export was successful
  • False if an exception occurred during export

​
Dependencies

This function requires the following packages:
  • skl2onnx: For converting scikit-learn models to ONNX
  • onnx: ONNX runtime dependencies

​
Example

from deployment.onnx_export import export_pipeline_to_onnx
from pathlib import Path
from sklearn.pipeline import Pipeline

# Assume pipeline is already trained
pipeline = Pipeline([...])

success = export_pipeline_to_onnx(
    pipeline=pipeline,
    output_path=Path("models/risk_model.onnx"),
    n_features=15
)

if success:
    print("Model exported successfully")
else:
    print("Export failed")

​
Use Cases

  • Cross-Platform Deployment: Deploy models to various platforms (web, mobile, embedded)
  • Performance Optimization: Use ONNX Runtime for faster inference
  • Framework Interoperability: Run scikit-learn models in non-Python environments
  • Model Serving: Deploy models to production inference servers

Build docs developers (and LLMs) love

Get started for freeTalk to us