MLTOOLS-4119: add tutorial about using ONNX models in CatBoost

ref:7991db7cc612c7762c46148c0922960aa47cd474

Author: ivankozlov98

convert_onnx_model/tutorial_convert_onnx_models.ipynb

@@ -0,0 +1,151 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "# Using ONNX models in CatBoost\n",
+ "\n",
+ "It is easy to apply ONNX models using CatBoost.\n",
+ "+ Save your model in the ONNX format\n",
+ "+ Load the ONNX model into CatBoost using the load_model() method\n",
+ "+ Apply your model in CatBoost using the predict() method\n",
+ "\n",
+ "Let us follow this scenario step-by-step for a LightGBM model."
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "\n",
+ "Download the MSRank dataset and import the necessary packages:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "from catboost import datasets, CatBoostRegressor\n",
+ "\n",
+ "from lightgbm import LGBMRegressor\n",
+ "\n",
+ "import onnxmltools\n",
+ "from onnxconverter_common import *\n",
+ "\n",
+ "\n",
+ "train_df, _ = datasets.msrank()\n",
+ "X, Y = train_df[train_df.columns[1:]], train_df[train_df.columns[0]]"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "\n",
+ "Build a model:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "[1.30604501 1.60390655 0.35207384 ... 1.18672199 0.55631924 0.54655847]\n"
+ ]
+ }
+ ],
+ "source": [
+ "model = LGBMRegressor()\n",
+ "model.fit(X, Y)\n",
+ "predict = model.predict(X)\n",
+ "print(predict)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "\n",
+ "Save the model in the ONNX format:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 3,
+ "metadata": {
+ "scrolled": false
+ },
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "The maximum opset needed by this model is only 1.\n",
+ "The maximum opset needed by this model is only 1.\n"
+ ]
+ }
+ ],
+ "source": [
+ "features_count = len(X.columns)\n",
+ "onnx_model = onnxmltools.convert_lightgbm(model, name='LightGBM', initial_types=[['input', FloatTensorType([0, features_count])]])\n",
+ "onnxmltools.utils.save_model(onnx_model, 'model.onnx')"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "\n",
+ "Load the ONNX model into CatBoost and compare the CatBoost and LightGBM predictions:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 4,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "[1.30604502 1.60390654 0.35207381 ... 1.18672202 0.55631925 0.54655849]\n"
+ ]
+ }
+ ],
+ "source": [
+ "catboost_model = CatBoostRegressor()\n",
+ "catboost_model.load_model('model.onnx', format='onnx')\n",
+ "catboost_predict = catboost_model.predict(X)\n",
+ "print(catboost_predict)"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.7.3"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}