|
| 1 | +{ |
| 2 | + "cells": [ |
| 3 | + { |
| 4 | + "cell_type": "code", |
| 5 | + "execution_count": 1, |
| 6 | + "metadata": {}, |
| 7 | + "outputs": [ |
| 8 | + { |
| 9 | + "name": "stdout", |
| 10 | + "output_type": "stream", |
| 11 | + "text": [ |
| 12 | + "[2021-10-21 07:08:52] Try to use the default NATS-Bench (topology) path from fast_mode=False and path=/Users/xuanyidong/.torch/NATS-tss-v1_0-3ffb9.pickle.pbz2.\n" |
| 13 | + ] |
| 14 | + } |
| 15 | + ], |
| 16 | + "source": [ |
| 17 | + "from nats_bench import create\n", |
| 18 | + "from nats_bench.api_utils import time_string\n", |
| 19 | + "import numpy as np\n", |
| 20 | + "\n", |
| 21 | + "# Create the API for size search space\n", |
| 22 | + "api_tss = create(None, \"tss\", fast_mode=False, verbose=False)" |
| 23 | + ] |
| 24 | + }, |
| 25 | + { |
| 26 | + "cell_type": "code", |
| 27 | + "execution_count": 2, |
| 28 | + "metadata": {}, |
| 29 | + "outputs": [ |
| 30 | + { |
| 31 | + "name": "stdout", |
| 32 | + "output_type": "stream", |
| 33 | + "text": [ |
| 34 | + "--------------------------------------------------ImageNet16-120--------------------------------------------------\n", |
| 35 | + "Best (10676) architecture on validation: |nor_conv_3x3~0|+|nor_conv_1x1~0|nor_conv_1x1~1|+|skip_connect~0|nor_conv_3x3~1|nor_conv_3x3~2|\n", |
| 36 | + "Best (857) architecture on test: |nor_conv_1x1~0|+|nor_conv_1x1~0|nor_conv_3x3~1|+|skip_connect~0|nor_conv_3x3~1|nor_conv_3x3~2|\n", |
| 37 | + "using validation ::: validation = 46.73, test = 46.20\n", |
| 38 | + "\n", |
| 39 | + "using test ::: validation = 46.38, test = 47.31\n", |
| 40 | + "\n" |
| 41 | + ] |
| 42 | + } |
| 43 | + ], |
| 44 | + "source": [ |
| 45 | + "def get_valid_test_acc(api, arch, dataset):\n", |
| 46 | + " is_size_space = api.search_space_name == \"size\"\n", |
| 47 | + " if dataset == \"cifar10\":\n", |
| 48 | + " xinfo = api.get_more_info(\n", |
| 49 | + " arch, dataset=dataset, hp=90 if is_size_space else 200, is_random=False\n", |
| 50 | + " )\n", |
| 51 | + " test_acc = xinfo[\"test-accuracy\"]\n", |
| 52 | + " xinfo = api.get_more_info(\n", |
| 53 | + " arch,\n", |
| 54 | + " dataset=\"cifar10-valid\",\n", |
| 55 | + " hp=90 if is_size_space else 200,\n", |
| 56 | + " is_random=False,\n", |
| 57 | + " )\n", |
| 58 | + " valid_acc = xinfo[\"valid-accuracy\"]\n", |
| 59 | + " else:\n", |
| 60 | + " xinfo = api.get_more_info(\n", |
| 61 | + " arch, dataset=dataset, hp=90 if is_size_space else 200, is_random=False\n", |
| 62 | + " )\n", |
| 63 | + " valid_acc = xinfo[\"valid-accuracy\"]\n", |
| 64 | + " test_acc = xinfo[\"test-accuracy\"]\n", |
| 65 | + " return (\n", |
| 66 | + " valid_acc,\n", |
| 67 | + " test_acc,\n", |
| 68 | + " \"validation = {:.2f}, test = {:.2f}\\n\".format(valid_acc, test_acc),\n", |
| 69 | + " )\n", |
| 70 | + "\n", |
| 71 | + "def find_best_valid(api, dataset):\n", |
| 72 | + " all_valid_accs, all_test_accs = [], []\n", |
| 73 | + " for index, arch in enumerate(api):\n", |
| 74 | + " valid_acc, test_acc, perf_str = get_valid_test_acc(api, index, dataset)\n", |
| 75 | + " all_valid_accs.append((index, valid_acc))\n", |
| 76 | + " all_test_accs.append((index, test_acc))\n", |
| 77 | + " best_valid_index = sorted(all_valid_accs, key=lambda x: -x[1])[0][0]\n", |
| 78 | + " best_test_index = sorted(all_test_accs, key=lambda x: -x[1])[0][0]\n", |
| 79 | + "\n", |
| 80 | + " print(\"-\" * 50 + \"{:10s}\".format(dataset) + \"-\" * 50)\n", |
| 81 | + " print(\n", |
| 82 | + " \"Best ({:}) architecture on validation: {:}\".format(\n", |
| 83 | + " best_valid_index, api[best_valid_index]\n", |
| 84 | + " )\n", |
| 85 | + " )\n", |
| 86 | + " print(\n", |
| 87 | + " \"Best ({:}) architecture on test: {:}\".format(\n", |
| 88 | + " best_test_index, api[best_test_index]\n", |
| 89 | + " )\n", |
| 90 | + " )\n", |
| 91 | + " _, _, perf_str = get_valid_test_acc(api, best_valid_index, dataset)\n", |
| 92 | + " print(\"using validation ::: {:}\".format(perf_str))\n", |
| 93 | + " _, _, perf_str = get_valid_test_acc(api, best_test_index, dataset)\n", |
| 94 | + " print(\"using test ::: {:}\".format(perf_str))\n", |
| 95 | + "\n", |
| 96 | + "dataset = \"ImageNet16-120\"\n", |
| 97 | + "find_best_valid(api_tss, dataset)" |
| 98 | + ] |
| 99 | + }, |
| 100 | + { |
| 101 | + "cell_type": "code", |
| 102 | + "execution_count": null, |
| 103 | + "metadata": {}, |
| 104 | + "outputs": [], |
| 105 | + "source": [] |
| 106 | + } |
| 107 | + ], |
| 108 | + "metadata": { |
| 109 | + "kernelspec": { |
| 110 | + "display_name": "Python 3", |
| 111 | + "language": "python", |
| 112 | + "name": "python3" |
| 113 | + }, |
| 114 | + "language_info": { |
| 115 | + "codemirror_mode": { |
| 116 | + "name": "ipython", |
| 117 | + "version": 3 |
| 118 | + }, |
| 119 | + "file_extension": ".py", |
| 120 | + "mimetype": "text/x-python", |
| 121 | + "name": "python", |
| 122 | + "nbconvert_exporter": "python", |
| 123 | + "pygments_lexer": "ipython3", |
| 124 | + "version": "3.8.8" |
| 125 | + } |
| 126 | + }, |
| 127 | + "nbformat": 4, |
| 128 | + "nbformat_minor": 4 |
| 129 | +} |
0 commit comments