{ "cells": [ { "cell_type": "code", "execution_count": null, "metadata": { "collapsed": false }, "outputs": [], "source": [ "%matplotlib inline" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "\n# Experiment Tracking: Coloring a Single Run\n\nThis example demostrates how to use opt-sugar in combination with mlflow\nfor single objective optimization experiment tracking\n\n\n" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "collapsed": false }, "outputs": [], "source": [ "# sphinx_gallery_thumbnail_path = '_static/coloring.png'\nimport datetime\nfrom urllib.parse import urlparse\nfrom itertools import product\nfrom collections import defaultdict\nimport logging\n\nimport gurobipy as gp\nimport mlflow\nfrom mlflow.exceptions import MlflowException\n\n# import sys; sys.path.append('/Users/Juan.ChaconLeon/opt/opt-sugar/src') # when running locally\nfrom opt_sugar import opt_flow\nfrom opt_sugar import low_sugar\n\n# The following function helper is very handy to visualize our colored graphs.\nfrom utils.coloring import get_graph_to_show\n\n# The generate_graph_data generates random graphs given a graph size and an edge probability.\nfrom utils.coloring import generate_graph_data\n\n# TODO: reformat this example similar to\n# https://github.com/scikit-learn/scikit-learn/blob/main/examples/calibration/plot_calibration_multiclass.py" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## The Optimizations Model Builder\n\nThe following class is the builder for the coloring problem.\n\n" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "collapsed": false }, "outputs": [], "source": [ "class ColoringModelBuilder:\n\n def __init__(self, data):\n self.data = data\n self.degree = None\n self.variables = None\n\n def build_variables(self, base_model):\n degrees = defaultdict(int)\n for v1, v2 in self.data[\"edges\"]:\n degrees[v1] += 1\n degrees[v2] += 1\n self.degree = max(degrees.items(), key=lambda x: x[1])[1]\n color_keys = list(product(self.data[\"nodes\"], range(self.degree)))\n color = base_model.addVars(color_keys, vtype=\"B\", name=\"color\")\n max_color = base_model.addVar(lb=0, ub=self.degree, vtype=\"C\", name=\"max_color\")\n self.variables = {\"color\": color, \"max_color\": max_color}\n\n def build_constraints(self, base_model):\n color = self.variables[\"color\"]\n for v1, c in color:\n # if color[v1, c] == 1 -> color[v2, c] == 0 for all v2 such that (v1, v2) or\n # belongs to E\n for v2 in self.data[\"nodes\"]:\n if (v2, v1) in self.data[\"edges\"] or (v1, v2) in self.data[\"edges\"]:\n base_model.addConstr(\n color[v2, c] <= 1 - color[v1, c], name=f\"color_{c}_{v1}_{v2}\"\n )\n\n for v in self.data[\"nodes\"]:\n base_model.addConstr(\n gp.quicksum(color[v, c] for c in range(self.degree)) == 1,\n name=f\"every_node_has_color_{v}\",\n )\n\n max_color = self.variables[\"max_color\"]\n for v, c in color:\n base_model.addConstr(\n c * color[v, c] <= max_color, name=f\"max_color_{v}_{c}\"\n )\n\n def build_objective(self, base_model):\n max_color = self.variables[\"max_color\"]\n base_model.setObjective(max_color, gp.GRB.MINIMIZE)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Tracking an Optimization Experiment\n\nAdd description here.\n\n" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "collapsed": false }, "outputs": [], "source": [ "logging.getLogger(\"mlflow\").setLevel(logging.CRITICAL) # Can be set DEBUG\n\ntry:\n experiment_name = f\"opt_exp_{datetime.datetime.now().strftime('%Y_%m_%d')}\"\n experiment_id = mlflow.create_experiment(name=experiment_name)\nexcept MlflowException:\n experiment_id = mlflow.get_experiment_by_name(name=experiment_name).experiment_id\n\nwith mlflow.start_run(experiment_id=experiment_id):\n def build(data):\n # Create a new model\n m = gp.Model(\"coloring\")\n model_builder = ColoringModelBuilder(data)\n model_builder.build_variables(m)\n model_builder.build_constraints(m)\n model_builder.build_objective(m)\n # setting parameters\n m.setParam('MIPFocus', 2)\n return m\n\n def callback(m):\n objective = m.getObjective()\n color_count = objective.getValue()\n callback_result = {\"color_count\": color_count, \"MIPFocus\": m.getParamInfo('MIPFocus')[2], \"RunTime\": m.RunTime}\n return callback_result\n\n # Generating a graph instance\n data = generate_graph_data(node_count=16, edge_probability=0.5)\n\n # Building\n opt_model = low_sugar.Model(build)\n result = opt_model.optimize(data=data, callback=callback)\n var_values = result[\"vars\"]\n g = get_graph_to_show(data, var_values)\n # g.show(name=\"vis.html\")\n\n # Note: Above is replacement for opt_model.fit(data, fit_callback) and opt_model.predict(data)\n mlflow.log_param(\"MIPFocus\", result[\"callback_result\"][\"MIPFocus\"])\n mlflow.log_param(\"RunTime\", result[\"callback_result\"][\"RunTime\"])\n mlflow.log_metric(\"color_count\", result[\"callback_result\"][\"color_count\"])\n\n tracking_url_type_store = urlparse(mlflow.get_tracking_uri()).scheme\n print(f\"tracking_url_type_store: {tracking_url_type_store}\")\n\n # Register the model\n if tracking_url_type_store != \"file\":\n # There are other ways to use the Model Registry, which depends on the use case,\n # please refer to the doc for more information:\n # https://mlflow.org/docs/latest/model-registry.html#api-workflow\n model_info = mlflow.sklearn.log_model(\n opt_model, \"opt_model\", registered_model_name=\"OptModel\"\n )\n else:\n model_info = mlflow.sklearn.log_model(opt_model, \"opt_model\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Load the Registered Model and Optimize with new Data\n\nAdd description here.\n\n" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "collapsed": false }, "outputs": [], "source": [ "logged_model_uri = model_info.model_uri\nprint(f\"logged_model_uri: {logged_model_uri}\")\n\n# Load model as a PyFuncModel.\nloaded_model = opt_flow.pyfunc.load_model(logged_model_uri)\n\n# Data generation\ndata = generate_graph_data(node_count=6, edge_probability=0.5)\nsolution = loaded_model.optimize(data)\nprint(f\"Optimized Coloring: {solution}\")\n\nvar_values = result[\"vars\"]\ng = get_graph_to_show(data, var_values)\n# g.show(name=\"vis.html\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "\n\n" ] } ], "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": "3.9.15" } }, "nbformat": 4, "nbformat_minor": 0 }