Added isolet logistic conversion

model_conversion/sklearn-logistic-to-onnx/convert-sklearn-logistic-to-onnx.ipynb

@@ -0,0 +1,232 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# How to Convert sk-learn Logistic Model to ONNX\n",
+    "\n",
+    "The following tutorial is a brief example of how to convert a [scikit-learn](https://scikit-learn.org/stable/) (aka sk-learn) **logistic ML model** to the [ONNX](https://onnx.ai/ ) format for use with Wallaroo.\n",
+    "\n",
+    "This tutorial assumes that you have a Wallaroo instance and are running this Notebook from the Wallaroo Jupyter Hub service."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "This tutorial provides the following:\n",
+    "\n",
+    "* `isolet_logistic_model_numclass.pickle`: a logistic trained sk-learn model.\n",
+    "    This model contains 617 columns.\n",
+    "* `isolet_test_data.tsv`:  A test file that can be used to verify the output of the converted logistic model.\n",
+    "* `try_isolet.ipynb`: This Jupyter Notebook demonstrates the use of the converted sk-learn logistic ML model in ONNX with Wallaroo."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Conversion Process\n",
+    "\n",
+    "### Libraries\n",
+    "\n",
+    "The first step is to import our libraries we will be using."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {
+    "vscode": {
+     "languageId": "plaintext"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "# Used to load the sk-learn model\n",
+    "import pickle\n",
+    "\n",
+    "# Used for the conversion process\n",
+    "import onnx, skl2onnx, onnxmltools\n",
+    "from skl2onnx.common.data_types import FloatTensorType\n",
+    "from skl2onnx.common.data_types import DoubleTensorType\n",
+    "from skl2onnx import convert_sklearn"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Now we can determine the correct ONNX Target Opset for our libraries.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "13"
+      ]
+     },
+     "execution_count": 2,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# figure out the correct opset\n",
+    "\n",
+    "from onnx.defs import onnx_opset_version\n",
+    "from onnxconverter_common.onnx_ex import DEFAULT_OPSET_NUMBER\n",
+    "TARGET_OPSET = min(DEFAULT_OPSET_NUMBER, onnx_opset_version())\n",
+    "TARGET_OPSET"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "With the `TARGET_OPSET` determined, we can convert our sklearn logistic model to onnx.\n",
+    "\n",
+    "* **IMPORTANT NOTE**:  Note that for the conversion process, `zipmap` is **disabled**."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Load the model that we will be converting:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {
+    "vscode": {
+     "languageId": "plaintext"
+    }
+   },
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/opt/conda/lib/python3.8/site-packages/sklearn/base.py:310: UserWarning: Trying to unpickle estimator LogisticRegression from version 0.24.2 when using version 0.24.1. This might lead to breaking code or invalid results. Use at your own risk.\n",
+      "  warnings.warn(\n"
+     ]
+    }
+   ],
+   "source": [
+    "# convert model to ONNX\n",
+    "\n",
+    "# load the model\n",
+    "\n",
+    "# with open(\"isolet_logistic_model.pickle\", \"rb\") as f:\n",
+    "with open(\"./isolet_logistic_model_numclass.pickle\", \"rb\") as f:\n",
+    "    logistic_model = pickle.load(f)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "We already know the number of columns, so we'll set that variable in the next step."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Set the number of columns\n",
+    "\n",
+    "ncols = 617"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Next up is to set the options.  As a reminder **zipmap must be disabled**."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "## Set the options\n",
+    "\n",
+    "initial_type = [('float_input', FloatTensorType([None, ncols]))]\n",
+    "options = {id(logistic_model): {'zipmap': False}} # here we turn off the zipmap"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "With everything ready, we can now convert the sk-learn Logistics model to ONNX, and store it in the variable `onnx_model_converted`."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "## Run the conversion\n",
+    "\n",
+    "onnx_model_converted = convert_sklearn(logistic_model, initial_types=initial_type, options=options,\n",
+    "                       target_opset=TARGET_OPSET)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Now we can save our model to a `onnx` file.  Once complete, we can run it through the `try_isolet.ipynb` notebook to verify it."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {
+    "vscode": {
+     "languageId": "plaintext"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "# Export the model to a file\n",
+    "onnx.save_model(onnx_model_converted, \"isolet_logistic_model_numclass.onnx\")"
+   ]
+  }
+ ],
+ "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": "3.8.6"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}