xgb2code

xgb2code - generate code for an XGB model

Description

This program takes an XGB model (in JSON format) and generates code for it.

Generating code for a model avoids having to call out to a different language (e.g., C) as well as avoids the need for using the XGB libraries at runtime.

Model Support

The following XGBoost objectives are supported:

• Binary classification: binary:logistic and binary:logitraw.
• Regression: reg:logistic, reg:squarederror, reg:linear, reg:absoluteerror, reg:pseudohubererror, and reg:quantileerror.

Supported Languages

Currently xgb2code supports generating Go code.

Usage

$ ./xgb2code -h
Usage of ./xgb2code:
  -function-name string
        The function name to use. Must be a valid Go function name.
  -go-package-name string
        The package name to use when generating Go code. Must be a valid Go package name.
  -input-json string
        Path to the model as JSON
  -language string
        Language to generate code for. Currently 'go' is supported. (default "go")
  -output-file string
        The file to write to

Example Usage

$ ./xgb2code -function-name predict \
             -go-package-name main \
             -input-json testdata/small-model/model.json \
             -language go \
             -output-file predict.go

produces a file predict.go where the primary model prediction function has the signature:

func predict(data []*float32, predMargin bool) float32 {

When predMargin is true, the function returns the raw margin (the summed tree outputs plus the base_score intercept). Otherwise, the sigmoid is applied for the logistic objectives (binary:logistic and reg:logistic); for all other objectives the margin is the final prediction, so predMargin has no effect.

Library Usage

The code generation functionality is also available as a library via the gen package:

package main

import (
	"log"

	"github.com/maxmind/xgb2code/gen"
)

func main() {
	err := gen.GenerateFile(
		"testdata/small-model/model.json", // input model JSON
		"main",                            // Go package name
		"predict",                         // function name
		"predict.go",                      // output file
	)
	if err != nil {
		log.Fatal(err)
	}
}

Installation

Release binaries and packages have been made available for several popular platforms. Simply download the binary for your platform and run it.

Bug Reports

Please report bugs by filing an issue with our GitHub issue tracker at https://github.com/maxmind/xgb2code/issues.

Copyright and License

This software is Copyright (c) 2022 - 2026 by MaxMind, Inc.

This is free software, licensed under the Apache License, Version 2.0 or the MIT License, at your option.