|
| 1 | +{ |
| 2 | + "cells": [ |
| 3 | + { |
| 4 | + "cell_type": "markdown", |
| 5 | + "id": "e44e86b0", |
| 6 | + "metadata": {}, |
| 7 | + "source": [ |
| 8 | + "# Configurations in Hyrax\n", |
| 9 | + "\n", |
| 10 | + "Hyrax uses configuration files to make it easier to track all of the parameters used for a given process.\n", |
| 11 | + "Hyrax attempts to provide a set of reasonable configuration values for all of the operations a user will perform, with exceptions.\n", |
| 12 | + "\n", |
| 13 | + "When you create a Hyrax instance, the built in defaults will be used automatically." |
| 14 | + ] |
| 15 | + }, |
| 16 | + { |
| 17 | + "cell_type": "code", |
| 18 | + "execution_count": null, |
| 19 | + "id": "42f8f15d", |
| 20 | + "metadata": {}, |
| 21 | + "outputs": [], |
| 22 | + "source": [ |
| 23 | + "from hyrax import Hyrax\n", |
| 24 | + "\n", |
| 25 | + "h = Hyrax()" |
| 26 | + ] |
| 27 | + }, |
| 28 | + { |
| 29 | + "cell_type": "markdown", |
| 30 | + "id": "c161d3b0", |
| 31 | + "metadata": {}, |
| 32 | + "source": [ |
| 33 | + "Once a Hyrax instance is created, the configuration is available as the ``h.config`` dictionary.\n", |
| 34 | + "We can view any particular value by using the keys of the dictionary.\n", |
| 35 | + "For instance, if we wanted to know what the default filename will be for the trained model weights file, we can use the following:" |
| 36 | + ] |
| 37 | + }, |
| 38 | + { |
| 39 | + "cell_type": "code", |
| 40 | + "execution_count": null, |
| 41 | + "id": "2a58fdc0", |
| 42 | + "metadata": {}, |
| 43 | + "outputs": [], |
| 44 | + "source": [ |
| 45 | + "h.config[\"train\"][\"weights_filename\"]" |
| 46 | + ] |
| 47 | + }, |
| 48 | + { |
| 49 | + "cell_type": "markdown", |
| 50 | + "id": "349ccd65", |
| 51 | + "metadata": {}, |
| 52 | + "source": [ |
| 53 | + "If you're interested, the complete Hyrax default configuration file can be seen here:\n", |
| 54 | + "https://hyrax.readthedocs.io/en/latest/hyrax_default_config.html\n", |
| 55 | + "\n", |
| 56 | + "A small portion of the default config can be seen below." |
| 57 | + ] |
| 58 | + }, |
| 59 | + { |
| 60 | + "cell_type": "code", |
| 61 | + "execution_count": null, |
| 62 | + "id": "0ed312fc", |
| 63 | + "metadata": {}, |
| 64 | + "outputs": [], |
| 65 | + "source": [ |
| 66 | + "import json\n", |
| 67 | + "\n", |
| 68 | + "print(json.dumps(h.config[\"general\"], indent=2))" |
| 69 | + ] |
| 70 | + }, |
| 71 | + { |
| 72 | + "cell_type": "markdown", |
| 73 | + "id": "0212fbb8", |
| 74 | + "metadata": {}, |
| 75 | + "source": [ |
| 76 | + "## Updating config values\n", |
| 77 | + "\n", |
| 78 | + "While it is possible to set the values in the dictionaries directly, using `h.set_config()`\n", |
| 79 | + "will rerun the Hyrax configuration resolution algorithm to ensure that any\n", |
| 80 | + "secondary configuration updates are incorporated." |
| 81 | + ] |
| 82 | + }, |
| 83 | + { |
| 84 | + "cell_type": "code", |
| 85 | + "execution_count": null, |
| 86 | + "id": "a2d20e7d", |
| 87 | + "metadata": {}, |
| 88 | + "outputs": [], |
| 89 | + "source": [ |
| 90 | + "print(f\"Default number of epochs: {h.config['train']['epochs']}\")\n", |
| 91 | + "h.set_config(\"train.epochs\", 5)\n", |
| 92 | + "print(f\"Updated number of epochs: {h.config['train']['epochs']}\")" |
| 93 | + ] |
| 94 | + }, |
| 95 | + { |
| 96 | + "cell_type": "markdown", |
| 97 | + "id": "d4fa0d6d", |
| 98 | + "metadata": {}, |
| 99 | + "source": [ |
| 100 | + "> ## Warning\n", |
| 101 | + "> **Use caution with ``None``**\n", |
| 102 | + "> When setting config values, it is generally a good idea to avoid using ``None``.\n", |
| 103 | + "> It is preferable to use ``False`` or an empty string, ``\"\"`` when using .toml." |
| 104 | + ] |
| 105 | + }, |
| 106 | + { |
| 107 | + "cell_type": "markdown", |
| 108 | + "id": "c382d1f7", |
| 109 | + "metadata": {}, |
| 110 | + "source": [ |
| 111 | + "## User configuration files\n", |
| 112 | + "\n", |
| 113 | + "It is not always convenient to manually modify configurations each time\n", |
| 114 | + "Hyrax runs.\n", |
| 115 | + "Configurations can be overridden with a custom .toml file provided to Hyrax during\n", |
| 116 | + "instantiation.\n", |
| 117 | + "\n", |
| 118 | + "In this example, we'll name our file ``user_config.toml`` (any file name is fine), and set the log level and weights filename like so:\n", |
| 119 | + "\n", |
| 120 | + "``` toml\n", |
| 121 | + "[general]\n", |
| 122 | + "log_level = \"error\"\n", |
| 123 | + "\n", |
| 124 | + "[train]\n", |
| 125 | + "weights_filename = \"model_weights.pth\"\n", |
| 126 | + "```\n", |
| 127 | + "\n", |
| 128 | + "> ## Note\n", |
| 129 | + "> User configuration files don't need to be a complete copy of the default\n", |
| 130 | + "> configuration file.\n", |
| 131 | + "> Hyrax will start with the defaults and override the specific keys with values\n", |
| 132 | + "> from the user's configuration file." |
| 133 | + ] |
| 134 | + }, |
| 135 | + { |
| 136 | + "cell_type": "code", |
| 137 | + "execution_count": null, |
| 138 | + "id": "bc1da3ba", |
| 139 | + "metadata": {}, |
| 140 | + "outputs": [], |
| 141 | + "source": [ |
| 142 | + "import json\n", |
| 143 | + "\n", |
| 144 | + "h = Hyrax(config_file=\"./user_config.toml\")\n", |
| 145 | + "\n", |
| 146 | + "print(\"Only the log level has been changed in the 'general' section:\")\n", |
| 147 | + "print(json.dumps(h.config[\"general\"], indent=2))\n", |
| 148 | + "\n", |
| 149 | + "print(\"Additionally, the trained weights file name is now:\")\n", |
| 150 | + "print(h.config[\"train\"][\"weights_filename\"])" |
| 151 | + ] |
| 152 | + }, |
| 153 | + { |
| 154 | + "cell_type": "markdown", |
| 155 | + "id": "e192d03d", |
| 156 | + "metadata": {}, |
| 157 | + "source": [ |
| 158 | + "## Config immutability\n", |
| 159 | + "\n", |
| 160 | + "To support reproducibility, once a Hyrax action starts, the configuration is locked.\n", |
| 161 | + "There are _very few_ exceptions to this rule, and Hyrax clearly logs when the\n", |
| 162 | + "config does change.\n", |
| 163 | + "This ensures that when the config state is saved, it accurately represents the settings\n", |
| 164 | + "used by Hyrax.\n", |
| 165 | + "\n", |
| 166 | + "The state of the configuration will always be saved alongside the output of a Hyrax\n", |
| 167 | + "action in a time stamped directory like ``YYYYmmdd-HHMMSS-train-RAND``.\n", |
| 168 | + "\n", |
| 169 | + "This is another way that Hyrax supports low-code experimentation with machine learning." |
| 170 | + ] |
| 171 | + } |
| 172 | + ], |
| 173 | + "metadata": { |
| 174 | + "kernelspec": { |
| 175 | + "display_name": "hyrax", |
| 176 | + "language": "python", |
| 177 | + "name": "python3" |
| 178 | + }, |
| 179 | + "language_info": { |
| 180 | + "codemirror_mode": { |
| 181 | + "name": "ipython", |
| 182 | + "version": 3 |
| 183 | + }, |
| 184 | + "file_extension": ".py", |
| 185 | + "mimetype": "text/x-python", |
| 186 | + "name": "python", |
| 187 | + "nbconvert_exporter": "python", |
| 188 | + "pygments_lexer": "ipython3", |
| 189 | + "version": "3.12.9" |
| 190 | + } |
| 191 | + }, |
| 192 | + "nbformat": 4, |
| 193 | + "nbformat_minor": 5 |
| 194 | +} |
0 commit comments