{"@attributes":{"version":"2.0"},"channel":{"title":"\ub300\ub450\ucf54\uae30","link":"https:\/\/hoohaha.tistory.com\/","description":"IT \uad00\ub828 \uc815\ubcf4 \ubc0f \ud301\n\uacf5\ubd80 \uae30\ub85d","language":"ko","pubDate":"Sat, 18 Apr 2026 06:30:37 +0900","generator":"TISTORY","ttl":"100","managingEditor":"\ub300\ub450\ucf54\uae30","image":{"title":"\ub300\ub450\ucf54\uae30","url":"https:\/\/tistory1.daumcdn.net\/tistory\/3027436\/attach\/b7751433044d40c9ad2e2b6387c82531","link":"https:\/\/hoohaha.tistory.com"},"item":[{"title":"[Python] pyenv-win \uc124\uce58 \ubc29\ubc95(PowerShell)","link":"https:\/\/hoohaha.tistory.com\/143","description":"<p data-ke-size=\"size16\"><a href=\"https:\/\/github.com\/pyenv\/pyenv?tab=readme-ov-file#windows\" target=\"_blank\" rel=\"noopener&nbsp;noreferrer\">https:\/\/github.com\/pyenv\/pyenv?tab=readme-ov-file#windows<\/a><\/p>\n<figure id=\"og_1710418710266\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"object\" data-og-title=\"GitHub - pyenv\/pyenv: Simple Python version management\" data-og-description=\"Simple Python version management. Contribute to pyenv\/pyenv development by creating an account on GitHub.\" data-og-host=\"github.com\" data-og-source-url=\"https:\/\/github.com\/pyenv\/pyenv?tab=readme-ov-file#windows\" data-og-url=\"https:\/\/github.com\/pyenv\/pyenv\" data-og-image=\"https:\/\/scrap.kakaocdn.net\/dn\/Tm1ze\/hyVxy1Pa5s\/ihQwhiJsfXkP0YSKP23We1\/img.png?width=1200&amp;height=600&amp;face=0_0_1200_600\"><a href=\"https:\/\/github.com\/pyenv\/pyenv?tab=readme-ov-file#windows\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/github.com\/pyenv\/pyenv?tab=readme-ov-file#windows\">\n<div class=\"og-image\" style=\"background-image: url('https:\/\/scrap.kakaocdn.net\/dn\/Tm1ze\/hyVxy1Pa5s\/ihQwhiJsfXkP0YSKP23We1\/img.png?width=1200&amp;height=600&amp;face=0_0_1200_600');\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">GitHub - pyenv\/pyenv: Simple Python version management<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">Simple Python version management. Contribute to pyenv\/pyenv development by creating an account on GitHub.<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">github.com<\/p>\n<\/div>\n<\/a><\/figure>\n<p data-ke-size=\"size16\">Pyenv\ub294 Windows\ub97c \uacf5\uc2dd\uc801\uc73c\ub85c \uc9c0\uc6d0\ud558\uc9c0 \uc54a\ub294\ub2e4. pyenv\ub85c \uc124\uce58\ud558\uc600\ub354\ub77c\ub3c4, Windows \ubc84\uc804\uc774 \uc544\ub2cc Linux \ubc84\uc804\uc744 \uc124\uce58\ud55c\ub2e4. \ub54c\ubb38\uc5d0, Windows-specific\ud55c \uae30\ub2a5\uc744 \uc0ac\uc6a9\ud560 \uc218 \uc5c6\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\ub300\uc2e0, official repo\uc5d0\uc11c \ub2e4\ub978 fork repo \uc0ac\uc6a9\uc744 \ucd94\ucc9c\ud55c\ub2e4. pyenv-win\uc774\ub2e4.<\/p>\n<p data-ke-size=\"size16\"><a href=\"https:\/\/github.com\/pyenv-win\/pyenv-win\" target=\"_blank\" rel=\"noopener&nbsp;noreferrer\">https:\/\/github.com\/pyenv-win\/pyenv-win<\/a><\/p>\n<figure id=\"og_1710419006013\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"object\" data-og-title=\"GitHub - pyenv-win\/pyenv-win: pyenv for Windows. pyenv is a simple python version management tool. It lets you easily switch bet\" data-og-description=\"pyenv for Windows. pyenv is a simple python version management tool. It lets you easily switch between multiple versions of Python. It's simple, unobtrusive, and follows the UNIX tradition of s...\" data-og-host=\"github.com\" data-og-source-url=\"https:\/\/github.com\/pyenv-win\/pyenv-win\" data-og-url=\"https:\/\/github.com\/pyenv-win\/pyenv-win\" data-og-image=\"https:\/\/scrap.kakaocdn.net\/dn\/XtsYH\/hyVABJmKIX\/hY3G5AF3vFR3dO1lKCCpGk\/img.png?width=1200&amp;height=600&amp;face=0_0_1200_600\"><a href=\"https:\/\/github.com\/pyenv-win\/pyenv-win\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/github.com\/pyenv-win\/pyenv-win\">\n<div class=\"og-image\" style=\"background-image: url('https:\/\/scrap.kakaocdn.net\/dn\/XtsYH\/hyVABJmKIX\/hY3G5AF3vFR3dO1lKCCpGk\/img.png?width=1200&amp;height=600&amp;face=0_0_1200_600');\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">GitHub - pyenv-win\/pyenv-win: pyenv for Windows. pyenv is a simple python version management tool. It lets you easily switch bet<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">pyenv for Windows. pyenv is a simple python version management tool. It lets you easily switch between multiple versions of Python. It's simple, unobtrusive, and follows the UNIX tradition of s...<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">github.com<\/p>\n<\/div>\n<\/a><\/figure>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\uc124\uce58 \ubc29\ubc95<\/p>\n<p data-ke-size=\"size16\">1. PowerShell\uc5d0 \uc544\ub798 \ucee4\ub9e8\ub4dc\ub97c \uc785\ub825\ud55c\ub2e4.<\/p>\n<pre id=\"code_1710419080747\" class=\"bash\" data-ke-language=\"bash\" data-ke-type=\"codeblock\"><code>Invoke-WebRequest -UseBasicParsing -Uri \"https:\/\/raw.githubusercontent.com\/pyenv-win\/pyenv-win\/master\/pyenv-win\/install-pyenv-win.ps1\" -OutFile \".\/install-pyenv-win.ps1\"; &amp;\".\/install-pyenv-win.ps1\"<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">1.1. \uc544\ub798\uc640 \uac19\uc740 \ubcf4\uc548 \uc5d0\ub7ec \ubc1c\uc0dd \uc2dc PowerShell\uc744 \"\uad00\ub9ac\uc790 \uad8c\ud55c\uc73c\ub85c \uc2e4\ud589\" \ud6c4 \uc544\ub798 \ucee4\ub9e8\ub4dc\ub97c \uc785\ub825 \ud6c4 \uc704 \ucee4\ub9e8\ub4dc\ub97c \uc785\ub825\ud55c\ub2e4.&nbsp;<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"841\" data-origin-height=\"130\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/ccC7hR\/btsFOYmLi6n\/MQaGWzxMzRhhfcFwyymR91\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/ccC7hR\/btsFOYmLi6n\/MQaGWzxMzRhhfcFwyymR91\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/ccC7hR\/btsFOYmLi6n\/MQaGWzxMzRhhfcFwyymR91\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FccC7hR%2FbtsFOYmLi6n%2FMQaGWzxMzRhhfcFwyymR91%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"841\" height=\"130\" data-origin-width=\"841\" data-origin-height=\"130\"\/><\/span><\/figure>\n<\/p>\n<pre id=\"code_1710419206939\" class=\"bash\" data-ke-type=\"codeblock\" data-ke-language=\"bash\"><code>Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope LocalMachine<\/code><\/pre>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"859\" data-origin-height=\"409\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/pLxrQ\/btsFM0e1tiu\/7EIC3MqvwkAKqcO4RqkRsk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/pLxrQ\/btsFM0e1tiu\/7EIC3MqvwkAKqcO4RqkRsk\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/pLxrQ\/btsFM0e1tiu\/7EIC3MqvwkAKqcO4RqkRsk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FpLxrQ%2FbtsFM0e1tiu%2F7EIC3MqvwkAKqcO4RqkRsk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"859\" height=\"409\" data-origin-width=\"859\" data-origin-height=\"409\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">*\ucd94\ud6c4 \uad8c\ud55c\uc744 Default\ub85c \ubc14\uafb8\uba74, pyenv\uac00 \uc791\ub3d9\ud558\uc9c0 \uc54a\ub294\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">2. PowerShell\uc744 \uc7ac\uc2dc\uc791\ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\"><br \/>3. \"pyenv --version\"\uc744 \uc2e4\ud589\ud558\uc5ec \uc124\uce58\uac00 \uc131\uacf5\ud588\ub294\uc9c0 \ud655\uc778\ud55c\ub2e4.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"281\" data-origin-height=\"50\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/DrqBg\/btsFOeRkRWs\/ip7iMGjam7Q4kQJNXRrknk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/DrqBg\/btsFOeRkRWs\/ip7iMGjam7Q4kQJNXRrknk\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/DrqBg\/btsFOeRkRWs\/ip7iMGjam7Q4kQJNXRrknk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FDrqBg%2FbtsFOeRkRWs%2Fip7iMGjam7Q4kQJNXRrknk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"281\" height=\"50\" data-origin-width=\"281\" data-origin-height=\"50\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\"><br \/>4. \"pyenv install -l\"\uc744 \uc2e4\ud589\ud558\uc5ec pyenv-win\uc5d0\uc11c \uc9c0\uc6d0\ud558\ub294 Python \ubc84\uc804 \ubaa9\ub85d\uc744 \ud655\uc778\ud55c\ub2e4.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"403\" data-origin-height=\"186\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/bC3KiI\/btsFMrEfDcX\/9drACbtUx0Gm9Z21YH5jIk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/bC3KiI\/btsFMrEfDcX\/9drACbtUx0Gm9Z21YH5jIk\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/bC3KiI\/btsFMrEfDcX\/9drACbtUx0Gm9Z21YH5jIk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FbC3KiI%2FbtsFMrEfDcX%2F9drACbtUx0Gm9Z21YH5jIk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"403\" height=\"186\" data-origin-width=\"403\" data-origin-height=\"186\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\"><br \/>5. pyenv install &lt;\ubc84\uc804&gt;\uc744 \uc2e4\ud589\ud558\uc5ec \uc9c0\uc6d0\ub418\ub294 \ubc84\uc804\uc744 \uc124\uce58\ud55c\ub2e4.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"671\" data-origin-height=\"119\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/ckb8UT\/btsFOlWYf5h\/aRQuQJvNgyzAQqdk0UBtik\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/ckb8UT\/btsFOlWYf5h\/aRQuQJvNgyzAQqdk0UBtik\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/ckb8UT\/btsFOlWYf5h\/aRQuQJvNgyzAQqdk0UBtik\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fckb8UT%2FbtsFOlWYf5h%2FaRQuQJvNgyzAQqdk0UBtik%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"671\" height=\"119\" data-origin-width=\"671\" data-origin-height=\"119\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">5.1. grep\uc640 \ube44\uc2b7\ud55c PowerShell\uc758 \ud14d\uc2a4\ud2b8 \ud544\ud130\ub9c1 \ucee4\ub9e8\ub4dc: <b>\"select-string -pattern &lt;\uac80\uc0c9\ud560 \uac83&gt;\"<\/b><\/p>\n<pre id=\"code_1710420251453\" class=\"bash\" data-ke-language=\"bash\" data-ke-type=\"codeblock\"><code>pyenv install -l | select-string -pattern 3.11<\/code><\/pre>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"504\" data-origin-height=\"313\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/EGN4s\/btsFOntIgB7\/UwII3jQkoaotJ6EGVDu6h0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/EGN4s\/btsFOntIgB7\/UwII3jQkoaotJ6EGVDu6h0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/EGN4s\/btsFOntIgB7\/UwII3jQkoaotJ6EGVDu6h0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FEGN4s%2FbtsFOntIgB7%2FUwII3jQkoaotJ6EGVDu6h0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"504\" height=\"313\" data-origin-width=\"504\" data-origin-height=\"313\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\"><br \/>6. \ud30c\uc774\uc36c \ubc84\uc804\uc744 \uc804\uc5ed \ubc84\uc804\uc73c\ub85c \uc124\uc815\ud558\uae30 \uc704\ud574 pyenv global &lt;version&gt;\uc744 \uc2e4\ud589\ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\">\ud2b9\uc815 \ud3f4\ub354\uc5d0\uc11c \ud2b9\uc815 \ubc84\uc804\uc73c\ub85c \uc124\uc815\ud558\ub824\uba74 global\uc744 local\ub85c \ubc14\uafb8\uba74 \ub41c\ub2e4.<\/p>","category":"CS\/Python","author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/143","comments":"https:\/\/hoohaha.tistory.com\/143#entry143comment","pubDate":"Thu, 14 Mar 2024 21:49:37 +0900"},{"title":"\uc8fc\ub2c8\uc5b4 \ub525\ub7ec\ub2dd \uc5d4\uc9c0\ub2c8\uc5b4 \uad00\uc810\uc5d0\uc11c\uc758 \ub370\uc774\ud130 \ub77c\ubca8\ub9c1 \uad50\uc721 \ud6c4\uae30","link":"https:\/\/hoohaha.tistory.com\/142","description":"<p data-ke-size=\"size16\">\uc9c1\uc7a5\uc5d0\uc11c \ud55c\ucc3d \ub370\uc774\ud130 \ub77c\ubca8\ub9c1 \uad00\ub828 \ubb38\uc81c\ub97c \ucc98\ub9ac\ud558\uace0 \uc788\ub294 \ub3c4\uc911, \ud06c\ub77c\uc6b0\ub4dc\uc6cd\uc2a4\uc5d0\uc11c \uc9c4\ud589\ud558\ub294 \ub370\uc774\ud130 \ub77c\ubca8\ub9c1 \uad50\uc721\uc5d0 \ub300\ud574 \uc54c\uac8c \ub418\uc5b4\uc11c \uc774\ub97c \ub4e3\uac8c \ub418\uc5c8\ub2e4. \ub0b4\uc77c\ubc30\uc6c0\uce74\ub4dc\ub97c \uc774\uc6a9\ud574 \ubb34\ub8cc\ub85c \uc791\uc5c5\uc790\/\uac80\uc218\uc790 \uacfc\uc815 \ubaa8\ub450 \uc774\uc218\ud558\uc600\uc73c\uba70, \uc774\uc5d0 \ub300\ud55c \ud6c4\uae30\ub97c \uc791\uc131\ud55c\ub2e4. \uc9c0\uadf9\ud788 \uc5d4\uc9c0\ub2c8\uc5b4\uc801\uc778 \uad00\uc810\uc73c\ub85c \uc791\uc131\ud558\uc600\ub2e4\ub294 \uac83\uc744 \ubbf8\ub9ac \uc774\uc57c\uae30\ud574\ub454\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"2040\" data-origin-height=\"944\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/uVWA6\/btsgC2PVj6k\/bE3oRE33Bx1myL1ixxSdzk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/uVWA6\/btsgC2PVj6k\/bE3oRE33Bx1myL1ixxSdzk\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/uVWA6\/btsgC2PVj6k\/bE3oRE33Bx1myL1ixxSdzk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FuVWA6%2FbtsgC2PVj6k%2FbE3oRE33Bx1myL1ixxSdzk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"2040\" height=\"944\" data-origin-width=\"2040\" data-origin-height=\"944\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\uc77c\ub2e8\uc740, \ub525\ub7ec\ub2dd \uad00\ub828 \uae38\uc744 \uac00\uace0\uc790 \ud558\ub294 \uc0ac\ub78c\uc774\ub77c\uba74 \ub450 \uacfc\uc815 \ubaa8\ub450 \uc218\uac15\ud558\ub294 \uac83\uc744 \ucd94\ucc9c\ud55c\ub2e4. \uc774\uc720\ub294 \ub525\ub7ec\ub2dd\uc758 \uac00\uc7a5 \uae30\ucd08\uc778 \ub370\uc774\ud130 \ucc98\ub9ac\uc5d0 \ub300\ud574 1\uc8fc\ub9cc\uc5d0 \uad49\uc7a5\ud788 \uc27d\uac8c \uae4a\uac8c \uc54c \uc218 \uc788\uae30 \ub54c\ubb38\uc774\ub2e4.<\/p>\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"decimal\">\n<li>\uc2dc\uac04\uc801\uc73c\ub85c\ub294 \ub450 \uacfc\uc815 \ubaa8\ub450 \uc57d 15\uc2dc\uac04\uc774\ub77c\uace0\ub294 \ud558\ub098, \ubc30\uc18d\uc73c\ub85c \ub4e4\uc744 \uc218 \uc788\uace0 \uc2e4\uc2b5 \uacfc\uc815\ub3c4 \uc624\ub798 \uac78\ub9ac\uc9c0 \uc54a\ub294\ub2e4. \uc790\uae30 \uc804 2\uc2dc\uac04 \uc815\ub3c4 \ud22c\uc790\ud558\ub294 \uac83\uc744 \uae30\uc900\uc73c\ub85c \ub450 \uacfc\uc815 \ud569\uccd0\uc11c \uc774\uc218\uae4c\uc9c0 1\uc8fc\uc77c\uc774 \ucc44 \uc548 \uac78\ub9b0 \uac83 \uac19\ub2e4.<\/li>\n<li>\ub09c\uc774\ub3c4\ub294 \uc77c\ubc18\uc778\ub4e4\uc744 \uc0c1\ub300\ub85c \ud558\uae30 \ub54c\ubb38\uc5d0 \ub9e4\uc6b0 \uc27d\ub2e4. \uadf8\ub807\ub2e4\uace0 \uae4a\uc774\uac00 \uc5c6\ub294 \uac83\uc774 \uc544\ub2c8\ub2e4. \ub9e4\uc6b0 \uc798 \uc9dc\uc5ec\uc9c4 \uad50\uc591 \uc218\uc5c5\uc744 \ub4e3\ub294 \ub290\ub08c\uc744 \ubc1b\uc558\ub2e4.<\/li>\n<li><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\uc5f0\uc18d\uc73c\ub85c \ub4e3\ub294 \uac83\uc744 \ucd94\ucc9c\ud55c\ub2e4. \uc791\uc5c5\uc790 \uacfc\uc815\uc5d0\uc11c \uc791\uc5c5\ud55c \ud504\ub85c\uc81d\ud2b8 \uc911 \uc77c\ubd80\uac00, \uaf64 \ub9ce\uc774 \uac80\uc218\uc790 \uacfc\uc815\uc5d0\ub3c4 \ub098\uc628\ub2e4.<\/span><\/li>\n<\/ol>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\uc5ec\uae30\uae4c\uc9c0\ub294 \ubcf4\ud3b8\uc801\uc778 \ub0b4\uc6a9\uc774\uace0, \uc9c0\uadf9\ud788 \uac1c\uc778\uc801\uc73c\ub85c \uc88b\uc558\ub358 \uc810\uc740 \uc544\ub798\uc640 \uac19\ub2e4.<\/span><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">&nbsp;\ub2e4\uc591\ud55c \uc5c5\uccb4\ub4e4\uc758 \ub370\uc774\ud130 \uc218\uc9d1 \uae30\uc900\uc744 \uc54c\uac8c \ub418\uc5c8\ub2e4. \ub370\uc774\ud130 \uba85\uc138\uc5d0 \uaf64 \ub9ce\uc740 \uace0\ubbfc \uc911\uc774\uc5c8\ub294\ub370, \ub370\uc774\ud130 \uba85\uc138\uc5d0 \uc5b4\ub824\uc6c0\uc744 \uacaa\ub294 \uc774\ub4e4\uc774\ub77c\uba74 \ub9ce\uc740 \ub370\uc774\ud130 \uba85\uc138 \ucf00\uc774\uc2a4\ub97c \uc27d\uac8c \uc54c \uc218 \uc788\uac8c \ud574\uc8fc\uc5b4\uc11c \uc88b\uc558\ub2e4. \ubaa8\ub378\ub4e4\uc774 \uc0c1\ud5a5\ud3c9\uc900\ud654\ub418\uace0 \uc624\ud508\uc18c\uc2a4\ub85c \uc0ac\uc804\ud559\uc2b5\ub41c \ubaa8\ub378\uc774 \ub098\uc624\ub294 \uc9c0\uae08, \uc790\uc2e0\ub4e4\uc774 \uc6d0\ud558\ub294 \ubaa9\uc801\uc5d0 \ub9de\ub294 \ubaa8\ub378\uc758 \uc815\ud655\ub3c4\ub97c \ub192\uc774\ub824\uba74 \uacb0\uad6d \ud559\uc2b5\uc5d0 \uc4f0\uc774\ub294 \ub370\uc774\ud130\uac00 \uc911\uc694\ud558\ub2e4. \uadf8\ub7f0\ub370 \uc544\ubb34\ub9ac \uc88b\uc740 \ub370\uc774\ud130\ub97c \uc218\uc9d1\ud558\ub354\ub77c\ub3c4, \ub9c8\uc9c0\ub9c9\uc5d0\ub294 \uc5b4\ub5bb\uac8c \ub77c\ubca8\ub9c1\uc744 \ud558\ub290\ub0d0 \uc2f8\uc6c0\uc778\ub370, \ud574\ub2f9 \uacfc\uc815\uc5d0\uc11c\ub294 \uac01 \uae30\uc5c5\ub4e4\uc774 \ub370\uc774\ud130 \uc678\uc8fc \uc5c5\uccb4\uc5d0\uac8c \uc5b4\ub5bb\uac8c \ub370\uc774\ud130\ub97c \uc218\uc9d1\ud574\ub2ec\ub77c\uace0 \ud558\ub294\uc9c0\ub97c \uc608\uc2dc\ub85c \ub4e4\uc5b4 \ub77c\ubca8\ub9c1\ud558\ub294 \ubc29\ubc95\uc744 \uac00\ub974\uccd0\uc900\ub2e4. \uc0ac\uc2e4 \ub370\uc774\ud130 \uc218\uc9d1\uc744 \ub9e1\ub294 \uc5c5\uccb4\ub2c8\uae4c \ub2f9\uc5f0\ud55c \ub9d0\uc774\ub2e4. \uc5ec\uae30\uc11c \uc870\uae08 \ub354 \ub098\uc544\uac00\uba74 \ub2e4\ub978 \uc5c5\uccb4\uac00 \uc5b4\ub5a4 \uc2dd\uc73c\ub85c \ubaa8\ub378\uc744 \ud559\uc2b5\uc2dc\ud0a4\uace0, \ud574\ub2f9 \ubaa8\ub378\uc744 \uc5b4\ub514\uc5d0 \uc4f0\ub294\uc9c0\ub3c4 \uc54c \uc218 \uc788\uc5b4\uc11c \uc5ec\ub7ec \uc544\uc774\ub514\uc5b4\ub97c \uc5bb\uc744 \uc218 \uc788\ub2e4. \ub108\ubb34 \ub2f9\uc5f0\ud558\uac8c\ub3c4 \uc54c\uace0 \uc788\ub2e4\uace0 \uc0dd\uac01\ud558\ub294 \ubd80\ubd84\uc5d0\uc11c \uc0dd\uac01\ubcf4\ub2e4 \ub9ce\uc774 \ub193\uce58\uace0 \uc788\ub358 \ubd80\ubd84\uc774 \uc788\uc5c8\ub2e4. \uae30\ubcf8\uc5d0 \ucda9\uc2e4\ud574\uc9c0\uace0 \uc2f6\ub2e4\uba74 \ucd94\ucc9c\ud55c\ub2e4.(\ub0b4\uc77c\ubc30\uc6c0\uce74\ub4dc\uac00 \uc788\uc73c\uba74 \uc815\ubd80 \uc9c0\uc6d0 100%\ub77c\uc11c \ubb34\ub8cc\ub85c \ub4e4\uc744 \uc218 \uc788\ub2e4.<\/p>","category":"\uae30\ud0c0","author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/142","comments":"https:\/\/hoohaha.tistory.com\/142#entry142comment","pubDate":"Sun, 21 May 2023 22:36:08 +0900"},{"title":"IT \ud2b8\ub80c\ub4dc&middot;\ub3d9\ud5a5 \uc54c\ub824\uc8fc\ub294 \uc815\ubd80&middot;\uacf5\uacf5\uae30\uad00 \uc0ac\uc774\ud2b8 \ubaa8\uc74c","link":"https:\/\/hoohaha.tistory.com\/138","description":"<h3 data-ke-size=\"size23\">TL;DR<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc815\ubd80\ub098 \uacf5\uacf5\uae30\uad00\uc774 \uc81c\uacf5\ud558\ub294 IT \ub3d9\ud5a5 \ubcf4\uace0\uc11c\ub294 \uc2e0\ub8b0\uc131\uacfc \ud3ec\uad04\uc131\uc774 \ub192\uc544 IT \ud2b8\ub80c\ub4dc \ud30c\uc545\uc5d0 \ub3c4\uc6c0\uc774 \ub41c\ub2e4.<\/li>\n<li>\uacfc\ud559\uae30\uc220\uc815\ubcf4\ud1b5\uc2e0\ubd80, \ud55c\uad6d\uc778\ud130\ub137\uc9c4\ud765\uc6d0, \ud55c\uad6d\uc9c0\ub2a5\uc815\ubcf4\uc0ac\ud68c\uc9c4\ud765\uc6d0 \ub4f1 \ub2e4\uc591\ud55c \uacf3\uc5d0\uc11c IT \ub3d9\ud5a5 \ubcf4\uace0\uc11c\ub97c \uc81c\uacf5\ud55c\ub2e4.<\/li>\n<li>IT \ub3d9\ud5a5 \ud30c\uc545\uc744 \uc704\ud574\uc11c\ub294 \uc815\ubd80\ub098 \uacf5\uacf5\uae30\uad00 \uc0ac\uc774\ud2b8\ub97c \ucc38\uace0\ud558\ub294 \uac83\uc774 \uc88b\ub2e4.<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ubaa9\ucc28<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc11c\ub860<\/li>\n<li>\uc815\ubd80&middot;\uacf5\uacf5\uae30\uad00\uc758 \uc815\ubcf4\ub97c \ubd10\uc57c \ud558\ub294 \uc774\uc720<\/li>\n<li>\uad00\ub828 \uc815\ubd80&middot;\uacf5\uacf5\uae30\uad00 \uc0ac\uc774\ud2b8 \ubaa9\ub85d<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\uc11c\ub860<\/h3>\n<p data-ke-size=\"size16\">ChatGPT\uc758 \ub4f1\uc7a5\uc73c\ub85c \ud004\ub9ac\ud2f0 \ub5a8\uc5b4\uc9c0\ub294 \uc815\ubcf4\uac00 \ub9ce\uc544\uc9c0\ub294 \uc694\uc998, \ucd9c\ucc98\uac00 \ud655\uc2e4\ud55c IT \uc815\ubcf4\ub97c \uc5bb\uace0 \uc2f6\uc5c8\ub2e4. \ubaa8\uc21c\uc801\uc774\uac8c\ub3c4 \uc815\ubcf4\ub97c \ucc3e\uc744 \ub54c\ub3c4, \uc774\uc720\ub97c \uc801\ub294 \ub370\uc5d0\ub3c4 AI(ChatGPT, Bing AI)\ub97c \uc0ac\uc6a9\ud558\uc600\uc9c0\ub9cc \uadf8\ub807\uac8c IT \ud2b8\ub80c\ub4dc \ubc0f \ub3d9\ud5a5\uc744 \uc54c\ub824\uc8fc\ub294&nbsp; \uc815\ubd80&middot;\uacf5\uacf5\uae30\uad00 \uc0ac\uc774\ud2b8\ub97c \ucc3e\uc740 \ud6c4 \uc815\ub9ac\ud558\uc600\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\uc815\ubd80&middot;\uacf5\uacf5\uae30\uad00\uc758 \uc815\ubcf4\ub97c \ubd10\uc57c \ud558\ub294 \uc774\uc720<\/h3>\n<p data-ke-size=\"size16\">\uc815\ubd80\ub098 \uacf5\uacf5\uae30\uad00\uc774 \ub9cc\ub4dc\ub294 \ub3d9\ud5a5 \ubcf4\uace0\uc11c\ub294 \ub2e4\uc74c\uacfc \uac19\uc740 \uc7a5\uc810\uc774 \uc788\ub2e4.<\/p>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc2e0\ub8b0\uc131\uc774 \ub192\uc74c: \uc815\ubd80\ub098 \uacf5\uacf5\uae30\uad00\uc740 \ub300\ubd80\ubd84 \uc911\ub9bd\uc801\uc778 \uc785\uc7a5\uc5d0\uc11c \uc790\ub8cc\ub97c \uc218\uc9d1\ud558\uace0 \ubd84\uc11d\ud558\uae30 \ub54c\ubb38\uc5d0, \ud3b8\ud5a5\uc131\uc774 \uc5c6\uace0 \uc2e0\ub8b0\uc131\uc774 \ub192\uc740 \uc790\ub8cc\ub97c \uc81c\uacf5\ud55c\ub2e4.<\/li>\n<li>\ub2e4\uc591\ud55c \uc815\ubcf4 \uc81c\uacf5: \uc815\ubd80\ub098 \uacf5\uacf5\uae30\uad00\uc5d0\uc11c \ubc1c\ud45c\ud558\ub294 \ubcf4\uace0\uc11c\ub294 \uc885\uc885 \ub2e4\uc591\ud55c \ubd84\uc57c\uc5d0\uc11c\uc758 IT \ud2b8\ub80c\ub4dc\ub97c \ub2e4\ub8e8\uae30 \ub54c\ubb38\uc5d0, \uc5b8\ub860\uc0ac\ub098 \ucee4\ubba4\ub2c8\ud2f0\uc5d0\uc11c \uc81c\uacf5\ud558\ub294 \uc815\ubcf4\ubcf4\ub2e4 \ub354 \ud3ec\uad04\uc801\uc774\uace0 \uad11\ubc94\uc704\ud55c \uc815\ubcf4\ub97c \uc5bb\uc744 \uc218 \uc788\ub2e4.<\/li>\n<li>\uc804\ubb38\uc801\uc778 \uc790\ub8cc: \uc815\ubd80\ub098 \uacf5\uacf5\uae30\uad00\uc5d0\uc11c \ubc1c\ud45c\ud558\ub294 IT \ub3d9\ud5a5 \ubcf4\uace0\uc11c\ub294 \uc804\ubb38\uac00\ub4e4\uc774 \ubd84\uc11d\ud558\uace0 \uc791\uc131\ud558\uae30 \ub54c\ubb38\uc5d0, \uae30\uc220\uc801\uc778 \uc6a9\uc5b4\ub098 \ub0b4\uc6a9\uc5d0 \ub300\ud55c \uc124\uba85\uc774 \ubcf4\ub2e4 \uc790\uc138\ud558\uace0 \uc804\ubb38\uc801\uc778 \uacbd\uc6b0\uac00 \ub9ce\ub2e4.<\/li>\n<li>\uc7a5\uae30\uc801\uc778 \uc804\ub9dd \uc81c\uc2dc: \ubcf4\ud1b5 \uc815\ubd80\ub098 \uacf5\uacf5\uae30\uad00\uc5d0\uc11c \ubc1c\ud45c\ud558\ub294 \ubcf4\uace0\uc11c\ub294 \uc7a5\uae30\uc801\uc778 IT \ud2b8\ub80c\ub4dc\uc640 \uc804\ub9dd\uc744 \uc81c\uc2dc\ud558\uae30\ub3c4 \ud55c\ub2e4. \uc774\ub97c \ud1b5\ud574 \uae30\uc5c5\uc774\ub098 \uac1c\uc778\uc774 IT \ud2b8\ub80c\ub4dc\ub97c \uc608\uce21\ud558\uace0 \ub300\ube44\ud558\ub294 \ub370 \ub3c4\uc6c0\uc744 \ubc1b\uc744 \uc218 \uc788\ub2e4.<\/li>\n<\/ol>\n<p data-ke-size=\"size16\">\ub530\ub77c\uc11c IT \ud2b8\ub80c\ub4dc\ub97c \ud30c\uc545\ud558\uace0 \uc2f6\ub2e4\uba74, \uc815\ubd80\ub098 \uacf5\uacf5\uae30\uad00\uc5d0\uc11c \ubc1c\ud45c\ud558\ub294 IT \ub3d9\ud5a5 \ubcf4\uace0\uc11c\ub97c \ucc38\uace0\ud558\ub294 \uac83\uc774 \uc88b\ub2e4.<\/p>\n<p data-ke-size=\"size16\">(from ChatGPT)<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\uad00\ub828 \uc815\ubd80&middot;\uacf5\uacf5\uae30\uad00 \uc0ac\uc774\ud2b8 \ubaa9\ub85d<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uacfc\ud559\uae30\uc220\uc815\ubcf4\ud1b5\uc2e0\ubd80: <a href=\"https:\/\/www.msit.go.kr\/bbs\/list.do?sCode=user&amp;mPid=100&amp;mId=101\">\uc8fc\uac04\ub3d9\ud5a5 - \uacfc\ud559\uae30\uc220\uc815\ubcf4\ud1b5\uc2e0\ubd80 (msit.go.kr)<\/a><\/li>\n<li>\ud55c\uad6d\uc778\ud130\ub137\uc9c4\ud765\uc6d0: <a href=\"https:\/\/www.kisa.or.kr\/20207\">KISA \ud55c\uad6d\uc778\ud130\ub137\uc9c4\ud765\uc6d0&gt;\uc9c0\uc2dd\ud50c\ub7ab\ud3fc&gt;\ub3d9\ud5a5\ubd84\uc11d&gt;\uc804\uccb4\ub3d9\ud5a5<\/a><\/li>\n<li>\ud55c\uad6d\uc9c0\ub2a5\uc815\ubcf4\uc0ac\ud68c\uc9c4\ud765\uc6d0: <a href=\"https:\/\/www.nia.or.kr\/site\/nia_kor\/main.do\">\ud55c\uad6d\uc9c0\ub2a5\uc815\ubcf4\uc0ac\ud68c\uc9c4\ud765\uc6d0 (nia.or.kr)<\/a><\/li>\n<li>\ud55c\uad6d\ubc29\uc1a1\ud1b5\uc2e0\uc804\ud30c\uc9c4\ud765\uc6d0: <a href=\"https:\/\/www.kca.kr\/boardList.do?boardId=TRENDS&amp;pageId=www145\">\ub3d9\ud5a5\uc790\ub8cc | KCA \ud55c\uad6d\ubc29\uc1a1\ud1b5\uc2e0\uc804\ud30c\uc9c4\ud765\uc6d0<\/a><\/li>\n<li><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\ud55c\uad6d\uc815\ubcf4\ud1b5\uc2e0\uae30\uc220\ud611\ud68c: <\/span><a style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\" href=\"https:\/\/www.tta.or.kr\/tta\/publicationHosuList.do?key=80&amp;rep=1&amp;searchKindNum=1\">TTA\uc800\ub110 - TTA \ub300\ud45c\ud648\ud398\uc774\uc9c0(\uad6d\ubb38)<\/a><\/li>\n<li>\uc815\ubcf4\ud1b5\uc2e0\uae30\ud68d\ud3c9\uac00\uc6d0:<span>&nbsp;<\/span><a style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\" href=\"https:\/\/www.iitp.kr\/kr\/1\/knowledge\/organScrapList.it\">\uc9c0\uc2dd &gt; ICT \ub3d9\ud5a5\uc815\ubcf4 | IITP<\/a><\/li>\n<li>ITFIND(\uc815\ubcf4\ud1b5\uc2e0\uae30\ud68d\ud3c9\uac00\uc6d0(IITP) \uc815\ubcf4\uc11c\ube44\uc2a4\ud300\uc5d0\uc11c \uac1c\ubc1c\ud55c ICT\uc815\ubcf4 \uc885\ud569\uac80\uc0c9\uc2dc\uc2a4\ud15c): <a style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\" href=\"https:\/\/www.itfind.or.kr\/trend\/trend\/hotIssue\/list.do\">ITFIND - Hot Issue<\/a><\/li>\n<li>\ud1b5\uacc4\uccad: <a style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\" href=\"https:\/\/data.kostat.go.kr\/social\/keyword\/index.do\">\ub274\uc2a4\uae30\ubc18\ud1b5\uacc4\uac80\uc0c9\uc11c\ube44\uc2a4 (kostat.go.kr)<\/a><\/li>\n<li>\ub300\ud55c\ubb34\uc5ed\ud22c\uc790\uc9c4\ud765\uacf5\uc0ac: <a href=\"https:\/\/www.kotrasvit.org\/news\/\">KOTRA \uc2e4\ub9ac\ucf58\ubc38\ub9ac \ubb34\uc5ed\uad00 (kotrasvit.org)<\/a><\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h4 data-ke-size=\"size20\">\uacfc\ud559\uae30\uc220\uc815\ubcf4\ud1b5\uc2e0\ubd80:<span>&nbsp;<\/span><a href=\"https:\/\/www.msit.go.kr\/bbs\/list.do?sCode=user&amp;mPid=100&amp;mId=101\">\uc8fc\uac04\ub3d9\ud5a5 - \uacfc\ud559\uae30\uc220\uc815\ubcf4\ud1b5\uc2e0\ubd80 (msit.go.kr)<\/a><\/h4>\n<figure id=\"og_1679841968580\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"\uc8fc\uac04\ub3d9\ud5a5 - \uacfc\ud559\uae30\uc220\uc815\ubcf4\ud1b5\uc2e0\ubd80\" data-og-description=\"\uc815\ucc45&middot;\ud1b5\uacc4 \ud648 \uc8fc\uac04\ub3d9\ud5a5 TOP\" data-og-host=\"www.msit.go.kr\" data-og-source-url=\"https:\/\/www.msit.go.kr\/bbs\/list.do?sCode=user&amp;mPid=100&amp;mId=101\" data-og-url=\"https:\/\/www.msit.go.kr\/bbs\/list.do?mId=101&amp;mPid=100&amp;sCode=user\" data-og-image=\"\"><a href=\"https:\/\/www.msit.go.kr\/bbs\/list.do?sCode=user&amp;mPid=100&amp;mId=101\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/www.msit.go.kr\/bbs\/list.do?sCode=user&amp;mPid=100&amp;mId=101\">\n<div class=\"og-image\" style=\"background-image: url();\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">\uc8fc\uac04\ub3d9\ud5a5 - \uacfc\ud559\uae30\uc220\uc815\ubcf4\ud1b5\uc2e0\ubd80<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">\uc815\ucc45&middot;\ud1b5\uacc4 \ud648 \uc8fc\uac04\ub3d9\ud5a5 TOP<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">www.msit.go.kr<\/p>\n<\/div>\n<\/a><\/figure>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">\ud55c\uad6d\uc778\ud130\ub137\uc9c4\ud765\uc6d0:<span>&nbsp;<\/span><a href=\"https:\/\/www.kisa.or.kr\/20207\">KISA \ud55c\uad6d\uc778\ud130\ub137\uc9c4\ud765\uc6d0&gt;\uc9c0\uc2dd\ud50c\ub7ab\ud3fc&gt;\ub3d9\ud5a5\ubd84\uc11d&gt;\uc804\uccb4\ub3d9\ud5a5<\/a><\/h4>\n<figure id=\"og_1679841984274\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"KISA \ud55c\uad6d\uc778\ud130\ub137\uc9c4\ud765\uc6d0\" data-og-description=\"\" data-og-host=\"www.kisa.or.kr\" data-og-source-url=\"https:\/\/www.kisa.or.kr\/20207\" data-og-url=\"https:\/\/www.kisa.or.kr\/20207\" data-og-image=\"https:\/\/scrap.kakaocdn.net\/dn\/biGYkt\/hyR2Xc4qa7\/H9CSXE7iLw4FVUcysE0CiK\/img.jpg?width=1921&amp;height=343&amp;face=0_0_1921_343\"><a href=\"https:\/\/www.kisa.or.kr\/20207\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/www.kisa.or.kr\/20207\">\n<div class=\"og-image\" style=\"background-image: url('https:\/\/scrap.kakaocdn.net\/dn\/biGYkt\/hyR2Xc4qa7\/H9CSXE7iLw4FVUcysE0CiK\/img.jpg?width=1921&amp;height=343&amp;face=0_0_1921_343');\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">KISA \ud55c\uad6d\uc778\ud130\ub137\uc9c4\ud765\uc6d0<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">&nbsp;<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">www.kisa.or.kr<\/p>\n<\/div>\n<\/a><\/figure>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">\ud55c\uad6d\uc9c0\ub2a5\uc815\ubcf4\uc0ac\ud68c\uc9c4\ud765\uc6d0:<span>&nbsp;<\/span><a href=\"https:\/\/www.nia.or.kr\/site\/nia_kor\/main.do\">\ud55c\uad6d\uc9c0\ub2a5\uc815\ubcf4\uc0ac\ud68c\uc9c4\ud765\uc6d0 (nia.or.kr)<\/a><\/h4>\n<figure id=\"og_1679842007126\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"[NIA \ud55c\uad6d\uc9c0\ub2a5\uc815\ubcf4\uc0ac\ud68c\uc9c4\ud765\uc6d0]\" data-og-description=\"\uc815\ubcf4\ud654\ub85c \uc0ac\ud68c\ud604\uc548\uc744 \ud574\uacb0\ud558\uace0 \uad6d\uac00\ubbf8\ub798\ub97c \uc5f4\uc5b4\uac00\ub294 \uc138\uacc4 \ucd5c\uace0\uc758 ICT \uc804\ubb38\uae30\uad00 NIA\ud55c\uad6d\uc9c0\ub2a5\uc815\ubcf4\uc0ac\ud68c\uc9c4\ud765\uc6d0\uc785\ub2c8\ub2e4.\" data-og-host=\"www.nia.or.kr\" data-og-source-url=\"https:\/\/www.nia.or.kr\/site\/nia_kor\/main.do\" data-og-url=\"https:\/\/www.nia.or.kr\/site\/nia_kor\/main.do\" data-og-image=\"https:\/\/scrap.kakaocdn.net\/dn\/xz86Y\/hyR2Py2VZE\/tHEpVvjaIZHWirR6RWgM00\/img.jpg?width=160&amp;height=160&amp;face=0_0_160_160,https:\/\/scrap.kakaocdn.net\/dn\/kv1Ye\/hyR4f3Z1y9\/dzM1kfcXbuk25B1AkFt1m1\/img.jpg?width=160&amp;height=160&amp;face=0_0_160_160\"><a href=\"https:\/\/www.nia.or.kr\/site\/nia_kor\/main.do\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/www.nia.or.kr\/site\/nia_kor\/main.do\">\n<div class=\"og-image\" style=\"background-image: url('https:\/\/scrap.kakaocdn.net\/dn\/xz86Y\/hyR2Py2VZE\/tHEpVvjaIZHWirR6RWgM00\/img.jpg?width=160&amp;height=160&amp;face=0_0_160_160,https:\/\/scrap.kakaocdn.net\/dn\/kv1Ye\/hyR4f3Z1y9\/dzM1kfcXbuk25B1AkFt1m1\/img.jpg?width=160&amp;height=160&amp;face=0_0_160_160');\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">[NIA \ud55c\uad6d\uc9c0\ub2a5\uc815\ubcf4\uc0ac\ud68c\uc9c4\ud765\uc6d0]<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">\uc815\ubcf4\ud654\ub85c \uc0ac\ud68c\ud604\uc548\uc744 \ud574\uacb0\ud558\uace0 \uad6d\uac00\ubbf8\ub798\ub97c \uc5f4\uc5b4\uac00\ub294 \uc138\uacc4 \ucd5c\uace0\uc758 ICT \uc804\ubb38\uae30\uad00 NIA\ud55c\uad6d\uc9c0\ub2a5\uc815\ubcf4\uc0ac\ud68c\uc9c4\ud765\uc6d0\uc785\ub2c8\ub2e4.<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">www.nia.or.kr<\/p>\n<\/div>\n<\/a><\/figure>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">\ud55c\uad6d\ubc29\uc1a1\ud1b5\uc2e0\uc804\ud30c\uc9c4\ud765\uc6d0:<span>&nbsp;<\/span><a href=\"https:\/\/www.kca.kr\/boardList.do?boardId=TRENDS&amp;pageId=www145\">\ub3d9\ud5a5\uc790\ub8cc | KCA \ud55c\uad6d\ubc29\uc1a1\ud1b5\uc2e0\uc804\ud30c\uc9c4\ud765\uc6d0<\/a><\/h4>\n<figure id=\"og_1679842031561\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"\ub3d9\ud5a5\uc790\ub8cc\" data-og-description=\"\ubc29\uc1a1\ud1b5\uc2e0\uc804\ud30c\uc758 \ud601\uc2e0\uc131\uc7a5\uc744 \uc120\ub3c4\ud558\ub294 \uc9c4\ud765\uae30\uad00 KCA \ud55c\uad6d\ubc29\uc1a1\ud1b5\uc2e0\uc804\ud30c\uc9c4\ud765\uc6d0\uc785\ub2c8\ub2e4.\" data-og-host=\"www.kca.kr\" data-og-source-url=\"https:\/\/www.kca.kr\/boardList.do?boardId=TRENDS&amp;pageId=www145\" data-og-url=\"https:\/\/www.kca.kr\/boardList.do?boardId=TRENDS&amp;pageId=www145\" data-og-image=\"https:\/\/scrap.kakaocdn.net\/dn\/nBiWz\/hyR37kAcMs\/vc5vV9kXnqK2NLBWN1dhAK\/img.jpg?width=512&amp;height=512&amp;face=0_0_512_512\"><a href=\"https:\/\/www.kca.kr\/boardList.do?boardId=TRENDS&amp;pageId=www145\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/www.kca.kr\/boardList.do?boardId=TRENDS&amp;pageId=www145\">\n<div class=\"og-image\" style=\"background-image: url('https:\/\/scrap.kakaocdn.net\/dn\/nBiWz\/hyR37kAcMs\/vc5vV9kXnqK2NLBWN1dhAK\/img.jpg?width=512&amp;height=512&amp;face=0_0_512_512');\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">\ub3d9\ud5a5\uc790\ub8cc<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">\ubc29\uc1a1\ud1b5\uc2e0\uc804\ud30c\uc758 \ud601\uc2e0\uc131\uc7a5\uc744 \uc120\ub3c4\ud558\ub294 \uc9c4\ud765\uae30\uad00 KCA \ud55c\uad6d\ubc29\uc1a1\ud1b5\uc2e0\uc804\ud30c\uc9c4\ud765\uc6d0\uc785\ub2c8\ub2e4.<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">www.kca.kr<\/p>\n<\/div>\n<\/a><\/figure>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\"><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\ud55c\uad6d\uc815\ubcf4\ud1b5\uc2e0\uae30\uc220\ud611\ud68c:<span>&nbsp;<\/span><\/span><a style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\" href=\"https:\/\/www.tta.or.kr\/tta\/publicationHosuList.do?key=80&amp;rep=1&amp;searchKindNum=1\">TTA\uc800\ub110 - TTA \ub300\ud45c\ud648\ud398\uc774\uc9c0(\uad6d\ubb38)<\/a><\/h4>\n<figure id=\"og_1679842037462\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"TTA\uc800\ub110 - TTA \ub300\ud45c\ud648\ud398\uc774\uc9c0(\uad6d\ubb38)\" data-og-description=\"\" data-og-host=\"www.tta.or.kr\" data-og-source-url=\"https:\/\/www.tta.or.kr\/tta\/publicationHosuList.do?key=80&amp;rep=1&amp;searchKindNum=1\" data-og-url=\"https:\/\/www.tta.or.kr\/tta\/publicationHosuList.do?key=80&amp;rep=1&amp;searchKindNum=1\" data-og-image=\"https:\/\/scrap.kakaocdn.net\/dn\/KOoHH\/hyR2MvvR8P\/wBnRJ8Tz3ILtSkqPzFLGM0\/img.jpg?width=1497&amp;height=2048&amp;face=0_0_1497_2048,https:\/\/scrap.kakaocdn.net\/dn\/zsya2\/hyR333w2do\/ik5TBSKsCj6xbevvKBzIqk\/img.jpg?width=1100&amp;height=1506&amp;face=0_0_1100_1506,https:\/\/scrap.kakaocdn.net\/dn\/brv44v\/hyR2O1bA5d\/kmjdogyEtAbKYhfr48ZPJK\/img.jpg?width=1100&amp;height=1506&amp;face=0_0_1100_1506\"><a href=\"https:\/\/www.tta.or.kr\/tta\/publicationHosuList.do?key=80&amp;rep=1&amp;searchKindNum=1\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/www.tta.or.kr\/tta\/publicationHosuList.do?key=80&amp;rep=1&amp;searchKindNum=1\">\n<div class=\"og-image\" style=\"background-image: url('https:\/\/scrap.kakaocdn.net\/dn\/KOoHH\/hyR2MvvR8P\/wBnRJ8Tz3ILtSkqPzFLGM0\/img.jpg?width=1497&amp;height=2048&amp;face=0_0_1497_2048,https:\/\/scrap.kakaocdn.net\/dn\/zsya2\/hyR333w2do\/ik5TBSKsCj6xbevvKBzIqk\/img.jpg?width=1100&amp;height=1506&amp;face=0_0_1100_1506,https:\/\/scrap.kakaocdn.net\/dn\/brv44v\/hyR2O1bA5d\/kmjdogyEtAbKYhfr48ZPJK\/img.jpg?width=1100&amp;height=1506&amp;face=0_0_1100_1506');\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">TTA\uc800\ub110 - TTA \ub300\ud45c\ud648\ud398\uc774\uc9c0(\uad6d\ubb38)<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">&nbsp;<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">www.tta.or.kr<\/p>\n<\/div>\n<\/a><\/figure>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">\uc815\ubcf4\ud1b5\uc2e0\uae30\ud68d\ud3c9\uac00\uc6d0:<span>&nbsp;<\/span><a style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\" href=\"https:\/\/www.iitp.kr\/kr\/1\/knowledge\/organScrapList.it\">\uc9c0\uc2dd &gt; ICT \ub3d9\ud5a5\uc815\ubcf4 | IITP<\/a><\/h4>\n<figure id=\"og_1679842054729\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"\uc9c0\uc2dd \n\t\t\t\t\t\t\t\t\t&gt;\n\t\t\t\t\t\t\t\t\tICT \ub3d9\ud5a5\uc815\ubcf4\n\t\t\t\t\t| IITP\" data-og-description=\"\uc7a0\uc2dc\ub9cc \uae30\ub2e4\ub824 \uc8fc\uc2dc\uae38 \ubc14\ub78d\ub2c8\ub2e4.\" data-og-host=\"www.iitp.kr\" data-og-source-url=\"https:\/\/www.iitp.kr\/kr\/1\/knowledge\/organScrapList.it\" data-og-url=\"https:\/\/www.iitp.kr\/kr\/1\/knowledge\/organScrapList.it\" data-og-image=\"\"><a href=\"https:\/\/www.iitp.kr\/kr\/1\/knowledge\/organScrapList.it\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/www.iitp.kr\/kr\/1\/knowledge\/organScrapList.it\">\n<div class=\"og-image\" style=\"background-image: url();\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">\uc9c0\uc2dd &gt; ICT \ub3d9\ud5a5\uc815\ubcf4 | IITP<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">\uc7a0\uc2dc\ub9cc \uae30\ub2e4\ub824 \uc8fc\uc2dc\uae38 \ubc14\ub78d\ub2c8\ub2e4.<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">www.iitp.kr<\/p>\n<\/div>\n<\/a><\/figure>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">ITFIND(\uc815\ubcf4\ud1b5\uc2e0\uae30\ud68d\ud3c9\uac00\uc6d0(IITP) \uc815\ubcf4\uc11c\ube44\uc2a4\ud300\uc5d0\uc11c \uac1c\ubc1c\ud55c ICT\uc815\ubcf4 \uc885\ud569\uac80\uc0c9\uc2dc\uc2a4\ud15c):<span>&nbsp;<\/span><a style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\" href=\"https:\/\/www.itfind.or.kr\/trend\/trend\/hotIssue\/list.do\">ITFIND - Hot Issue<\/a><\/h4>\n<figure id=\"og_1679842069294\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"ITFIND - Hot Issue\" data-og-description=\"\" data-og-host=\"www.itfind.or.kr\" data-og-source-url=\"https:\/\/www.itfind.or.kr\/trend\/trend\/hotIssue\/list.do\" data-og-url=\"https:\/\/www.itfind.or.kr\/trend\/trend\/hotIssue\/list.do\" data-og-image=\"\"><a href=\"https:\/\/www.itfind.or.kr\/trend\/trend\/hotIssue\/list.do\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/www.itfind.or.kr\/trend\/trend\/hotIssue\/list.do\">\n<div class=\"og-image\" style=\"background-image: url();\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">ITFIND - Hot Issue<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">&nbsp;<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">www.itfind.or.kr<\/p>\n<\/div>\n<\/a><\/figure>\n<p><figure class=\"imagegridblock\">\n  <div class=\"image-container\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/PZYt5\/btr5RSbGgkF\/6DkhGMYyxiC4kmAraLRS5k\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/PZYt5\/btr5RSbGgkF\/6DkhGMYyxiC4kmAraLRS5k\/img.png\" style=\"width: 38.2254%; margin-right: 10px;\" data-origin-width=\"1446\" data-origin-height=\"1288\" data-is-animation=\"false\" data-widthpercent=\"38.68\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/PZYt5\/btr5RSbGgkF\/6DkhGMYyxiC4kmAraLRS5k\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FPZYt5%2Fbtr5RSbGgkF%2F6DkhGMYyxiC4kmAraLRS5k%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1446\" height=\"1288\"\/><\/span><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/egj4Vf\/btr5QzKzU4x\/my8GdVGP3LL03kafORtg20\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/egj4Vf\/btr5QzKzU4x\/my8GdVGP3LL03kafORtg20\/img.png\" data-origin-width=\"2834\" data-origin-height=\"1592\" data-is-animation=\"false\" data-widthpercent=\"61.32\" style=\"width: 60.6118%;\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/egj4Vf\/btr5QzKzU4x\/my8GdVGP3LL03kafORtg20\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fegj4Vf%2Fbtr5QzKzU4x%2Fmy8GdVGP3LL03kafORtg20%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"2834\" height=\"1592\"\/><\/span><\/div>\n<\/figure>\n<\/p>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">\ud1b5\uacc4\uccad:<span>&nbsp;<\/span><a style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\" href=\"https:\/\/data.kostat.go.kr\/social\/keyword\/index.do\">\ub274\uc2a4\uae30\ubc18\ud1b5\uacc4\uac80\uc0c9\uc11c\ube44\uc2a4 (kostat.go.kr)<\/a><\/h4>\n<figure id=\"og_1679842075233\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"\ub274\uc2a4\uae30\ubc18\ud1b5\uacc4\uac80\uc0c9\uc11c\ube44\uc2a4\" data-og-description=\"\" data-og-host=\"data.kostat.go.kr\" data-og-source-url=\"https:\/\/data.kostat.go.kr\/social\/keyword\/index.do\" data-og-url=\"https:\/\/data.kostat.go.kr\/social\/keyword\/index.do\" data-og-image=\"\"><a href=\"https:\/\/data.kostat.go.kr\/social\/keyword\/index.do\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/data.kostat.go.kr\/social\/keyword\/index.do\">\n<div class=\"og-image\" style=\"background-image: url();\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">\ub274\uc2a4\uae30\ubc18\ud1b5\uacc4\uac80\uc0c9\uc11c\ube44\uc2a4<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">&nbsp;<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">data.kostat.go.kr<\/p>\n<\/div>\n<\/a><\/figure>\n<h4 data-ke-size=\"size20\">&nbsp;<\/h4>\n<h4 data-ke-size=\"size20\">\ub300\ud55c\ubb34\uc5ed\ud22c\uc790\uc9c4\ud765\uacf5\uc0ac:<span>&nbsp;<\/span><a href=\"https:\/\/www.kotrasvit.org\/news\/\">KOTRA \uc2e4\ub9ac\ucf58\ubc38\ub9ac \ubb34\uc5ed\uad00 (kotrasvit.org)<\/a><\/h4>\n<figure id=\"og_1679842077466\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"KOTRA \uc2e4\ub9ac\ucf58\ubc38\ub9ac \ubb34\uc5ed\uad00\" data-og-description=\"\uc560\ud50c, \uc2e0\ud615 \uc544\uc774\ud3f0 XS, XS\ub9e5\uc2a4, XR 3\uc885 \uacf5\uac1c September 11, 2018\" data-og-host=\"www.kotrasvit.org\" data-og-source-url=\"https:\/\/www.kotrasvit.org\/news\/\" data-og-url=\"https:\/\/www.kotrasvit.org\/news\/\" data-og-image=\"\"><a href=\"https:\/\/www.kotrasvit.org\/news\/\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/www.kotrasvit.org\/news\/\">\n<div class=\"og-image\" style=\"background-image: url();\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">KOTRA \uc2e4\ub9ac\ucf58\ubc38\ub9ac \ubb34\uc5ed\uad00<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">\uc560\ud50c, \uc2e0\ud615 \uc544\uc774\ud3f0 XS, XS\ub9e5\uc2a4, XR 3\uc885 \uacf5\uac1c September 11, 2018<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">www.kotrasvit.org<\/p>\n<\/div>\n<\/a><\/figure>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>","category":"IT \ud301","author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/138","comments":"https:\/\/hoohaha.tistory.com\/138#entry138comment","pubDate":"Sun, 26 Mar 2023 23:54:22 +0900"},{"title":"[ChatGPT] \ub178\ucf54\ub4dc \uac1c\ubc1c! ChatGPT\ub85c \ud06c\ub86c \uc775\uc2a4\ud150\uc158\uc744 \uac1c\ubc1c\ud574\ubcf4\uc558\ub2e4(1)","link":"https:\/\/hoohaha.tistory.com\/137","description":"<p data-ke-size=\"size16\">\ucd5c\uadfc \ud55c \ubc88\ub3c4 \ub9cc\ub4e4\uc5b4\ubcf8 \uc801\uc774 \uc5c6\ub294 \ud06c\ub86c \uc775\uc2a4\ud150\uc158\uc744 <span>ChatGPT\ub97c \uc774\uc6a9\ud558\uc5ec <span>\ud558\ub8e8\ub9cc\uc5d0<\/span><span>&nbsp;<\/span><\/span>\uaf64 \uc131\uacf5\uc801\uc73c\ub85c \ub9cc\ub4e4\uc5c8\ub2e4. \ub290\ub080 \uc810\uc774 \ub9ce\uc544 \ud574\ub2f9 \uacbd\ud5d8\uc744 \uacf5\uc720\ud558\uace0\uc790 \ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style1\" \/>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">&nbsp;\uc694\uc998 ChatGPT\uc5d0 \uc7ac\ubbf8\uac00 \ub4e4\ub824 \ud558\ub8e8 \uc885\uc77c \uc774\uac83\ub9cc \ud558\uace0 \uc788\ub2e4. \uc9c0\uae08\uae4c\uc9c0 \ud558\ub358 \uc5c5\ubb34\ub4e4\ubd80\ud130 \uc0ac\uc18c\ud55c \uc758\uc0ac \uacb0\uc815, \ud14d\uc2a4\ud2b8 \uc694\uc57d \ub4f1\ub4f1 \uc6ec\ub9cc\ud55c \uac83\ub4e4\uc740 ChatGPT\uc5d0\uac8c \ubb3c\uc5b4\ubcf4\uace0\ub294 \ud55c\ub2e4. \uac00\ub2a5\uc131\uc740 \ubb34\uad81\ubb34\uc9c4\ud558\uba70 \uc544\uc9c1\ub3c4 \ud55c\ucc38 \ubc30\uc6cc\uc57c \ud55c\ub2e4\ub294 \uc0dd\uac01\uc774 \ub4e0\ub2e4. \ucc38\uace0\ub85c \ud504\ub86c\ud504\ud2b8 \uc5d4\uc9c0\ub2c8\uc5b4\ub9c1\uc774\ub77c\ub294 \uac83\uc774 \uc788\ub2e4. <b>\ud504\ub86c\ud504\ud2b8 \uc5d4\uc9c0\ub2c8\uc5b4\ub9c1<\/b>\uc740 <b>\uc5b8\uc5b4 \ubaa8\ub378\uc5d0\uc11c \uc751\ub2f5\uc744 \uc0dd\uc131\ud558\uae30 \uc704\ud55c \ud6a8\uacfc\uc801\uc778 \ud504\ub86c\ud504\ud2b8\ub97c \ub9cc\ub4dc\ub294 \ud504\ub85c\uc138\uc2a4<\/b>\uc774\ub2e4.(by ChatGPT) ChatGPT\ub97c \ud6a8\uacfc\uc801\uc73c\ub85c \uc4f0\uace0 \uc2f6\uc740 \uc0ac\ub78c\uc774\ub77c\uba74 \uc774\uc5d0 \ub300\ud574 \ubc30\uc6cc\ub450\uba74 \uc88b\uc744 \uac83\uc774\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><span><span>&nbsp;\ubcf8\uc778\uc740 <\/span>\uc608\uc804\ubd80\ud130 \ud06c\ub86c \uc775\uc2a4\ud150\uc158\uc5d0 \ub300\ud55c \uc544\uc774\ub514\uc5b4\uac00 \uba87 \uac1c \uc788\uc5c8\uace0 \uba87\uba87 \uc775\uc2a4\ud150\uc158\uc740 \ub2e4\uc6b4\ub85c\ub4dc \ud6c4 \ucee4\uc2a4\ud140\ud55c \uc801\ub3c4 \uc788\uc5c8\ub2e4. \ud558\uc9c0\ub9cc \ucc98\uc74c\ubd80\ud130 \uc0dd\uc131\ud558\ub294 \uac83\uc740 \uae30\ubcf8 \ubc30\uacbd\uc9c0\uc2dd\ub3c4 \ubd80\uc871\ud588\uace0 \ub7ec\ub2dd \ucee4\ube0c\ub4f1\uc758 \uc774\uc720\ub85c \ud558\uc9c0 \uc54a\uc558\uc5c8\ub2e4. \uadf8\ub7f0\ub370<span>&nbsp;<\/span><\/span>\uadf8\ub3d9\uc548\uc758 ChatGPT \uacbd\ud5d8\uc744 \uc0b4\ub824\ubcf8 \uacb0\uacfc, \ub0b4\uac00 \uac70\uc758 \ubaa8\ub974\ub294 \ubd84\uc57c\uc5d0 \ub300\ud574\uc11c\ub3c4 \uc124\uba85\ub9cc \uc798 \ud55c\ub2e4\uba74 ChatGPT\ub97c \uc774\uc6a9\ud558\uc5ec \uac1c\ubc1c\uc744 \ud560 \uc218 \uc788\uaca0\ub2e4\ub294 \uc0dd\uac01\uc774 \ub4e4\uc5c8\ub2e4. \uc2e4\ud604\ub418\uc9c0 \uc54a\uc740 \uc544\uc774\ub514\uc5b4\ub85c \uadf8\ub0e5 \ub450\uae30\uc5d0 \uc544\uae4c\uc6e0\ub294\uc9c0 \uc544\ub2cc\uc9c0\ub294 \uc0ac\uc6a9\uc790\ub4e4\uc774 \ud310\ub2e8\ud574\uc8fc\uaca0\uc9c0\ub9cc, \ub0b4 \uc544\uc774\ub514\uc5b4\ub97c \uc2e4\uc81c\ub85c \uad6c\ud604\ud574\ubcf4\uaca0\ub2e4\ub294 \uac83\uc5d0 \uc758\uc758\ub97c \ub450\uace0 \uc9c4\ud589\ud588\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">&nbsp; \uc8fc\ub9d0 \uac04 ChatGPT\uc640 \uc528\ub984\ud588\uace0, \ub180\ub78d\uac8c\ub3c4 \uc5b4\ub290 \uc815\ub3c4 \uc131\uacf5\uc801\uc73c\ub85c \uc775\uc2a4\ud150\uc158\uc744 \ub9cc\ub4e4\uc5c8\ub2e4(\ucd5c\uc885 \uc758\ub3c4 \uad6c\ud604\uae4c\uc9c0\ub294 \uc870\uae08 \ub354 \ub0a8\uc558\ub2e4). \ud574\ub2f9 \ubd84\uc57c \uac1c\ubc1c\uc790\ub4e4\uc774 \ubcf4\uae30\uc5d0\ub294 \uc5b4\ub5a8\uc9c0 \ubaa8\ub974\uaca0\uc9c0\ub9cc, \ubcf8\uc778\uc5d0\uac8c \uc788\uc5b4\uc11c\ub294 <b>ChatGPT\ub77c\ub294 \uc0c8\ub85c\uc6b4 \uace0\ucc28\uc6d0 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4\ub97c \ubc30\uc6cc\uc11c \ubaa8\ub974\ub294 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4 \uae30\ubc18 \ud504\ub85c\uadf8\ub7a8\uc744 \ub9cc\ub4e4\uc5c8\ub2e4<\/b>\ub294 \uc810\uc5d0\uc11c \ub9e4\uc6b0 \ud070 \uc758\ubbf8\uac00 \uc788\ub2e4. \uc870\ub9cc\uac04 \uc544\ub798\uc758 \uc778\uc6a9\uad6c\uc5d0\uc11c \"C\uc640 C++\"\uc774 \"ChatGPT\"\ub85c, \"\uc5b4\uc148\ube14\ub9ac \uc5b8\uc5b4 \ud504\ub85c\uadf8\ub798\uba38\"\uac00 \"\ud504\ub85c\uadf8\ub798\uba38\"\ub85c \ubc14\ub014 \ub0a0\uc774 \uba38\uc9c0 \uc54a\uc558\ub2e4\uace0 \uc0dd\uac01\ub41c\ub2e4.<\/p>\n<blockquote data-ke-style=\"style2\">C\uc640&nbsp;C++&nbsp;\uac19\uc740&nbsp;\ucd5c\uc2e0&nbsp;\ucef4\ud30c\uc77c&nbsp;\uc5b8\uc5b4\ub294&nbsp;\uc218\ucc9c&nbsp;\uc904\uc774&nbsp;\ub118\ub294&nbsp;\ucf54\ub4dc\ub97c&nbsp;\uad00\ub9ac\ud560&nbsp;\uc218&nbsp;\uc788\ub294&nbsp;\ucd5c\uace0\uc758&nbsp;\uc5b4\uc148\ube14\ub9ac&nbsp;\uc5b8\uc5b4&nbsp;\ud504\ub85c\uadf8\ub798\uba38\ubcf4\ub2e4&nbsp;\ud6e8\uc52c&nbsp;\ube60\ub974\uba70,&nbsp;\ucd5c\uc2e0&nbsp;\uace0\uae09&nbsp;\uc5b8\uc5b4\ub3c4&nbsp;\uc5b4\uc148\ube14\ub9ac&nbsp;\uc5b8\uc5b4\ubcf4\ub2e4&nbsp;\ucf54\ub529&nbsp;\ubc0f&nbsp;\ub514\ubc84\uae45\uc5d0&nbsp;\ud6e8\uc52c&nbsp;\uc801\uc740&nbsp;\uc778\ub825\uc744&nbsp;\ud544\uc694\ub85c&nbsp;\ud569\ub2c8\ub2e4.<br \/>\ucd9c\ucc98: https:\/\/qr.ae\/pranWI<\/blockquote>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style1\" \/>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\uc775\uc2a4\ud150\uc158\uc758 \uae30\ub2a5\uc740 \ub2e4\uc74c\uacfc \uac19\ub2e4.<\/p>\n<blockquote data-ke-style=\"style1\"><span style=\"font-family: 'Noto Serif KR';\"> \uc120\ud0dd\ud55c \ud55c\uad6d\uc5b4 \ubb38\uc7a5 \uc18d\uc5d0\uc11c \ud55c\uc790\uc5b4\ub97c \ucd94\ucd9c\ud558\uc5ec \uadf8 \ub73b\uc744 \ubcf4\uc5ec\uc900\ub2e4.<\/span><\/blockquote>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\ubb34\uc2a8 \ub9d0\uc778\uc9c0 \ubaa8\ub974\uaca0\uc9c0\ub9cc \uc774\ubbf8\uc9c0\ub97c \ubcf4\uba74 \uc774\ud574\uac00 \uac08 \uac83\uc774\ub2e4.<\/p>\n<p><figure class=\"imagegridblock\">\n  <div class=\"image-container\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/djUgNb\/btr0RIYYSMn\/Jk429fWVnn5Mg4mSxfKad1\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/djUgNb\/btr0RIYYSMn\/Jk429fWVnn5Mg4mSxfKad1\/img.png\" data-origin-width=\"468\" data-origin-height=\"324\" data-is-animation=\"false\" style=\"width: 36.9921%; margin-right: 10px;\" data-widthpercent=\"37.43\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/djUgNb\/btr0RIYYSMn\/Jk429fWVnn5Mg4mSxfKad1\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FdjUgNb%2Fbtr0RIYYSMn%2FJk429fWVnn5Mg4mSxfKad1%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"468\" height=\"324\"\/><\/span><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/c2Q2sx\/btr0KSVnRaz\/YoQv27jk5d2rtlkQUkUNW1\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/c2Q2sx\/btr0KSVnRaz\/YoQv27jk5d2rtlkQUkUNW1\/img.png\" data-origin-width=\"454\" data-origin-height=\"188\" data-is-animation=\"false\" style=\"width: 61.8452%;\" data-widthpercent=\"62.57\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/c2Q2sx\/btr0KSVnRaz\/YoQv27jk5d2rtlkQUkUNW1\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fc2Q2sx%2Fbtr0KSVnRaz%2FYoQv27jk5d2rtlkQUkUNW1%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"454\" height=\"188\"\/><\/span><\/div>\n<\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">&nbsp;\uc774 \uc571\uc758 \ud575\uc2ec\uc740, \ubb38\ub9e5 \ucd94\ub860\uc5d0 \uc788\ub2e4. \ud55c\uad6d\uc5b4\uc758 \ud55c\uc790\uc5b4\ub294 \ub3d9\uc74c\uc774\uc758\uc5b4\uac00 \ub9ce\uc740\ub370, \ubb38\ub9e5\uc5d0 \ub530\ub77c \uc644\uc804\ud788 \ub2e4\ub97c \ub54c\uac00 \ub9ce\ub2e4. \ubb38\ub9e5\uc5d0 \ub9de\ub294 \ub2e8\uc5b4\uc758 \uc758\ubbf8\ub97c \ucc3e\uc73c\ub824\uba74 \ub9ce\uc774 \ucc3e\uc544\ubcf4\uc544\uc57c \ud558\uace0 \uc774\ub294 \ud55c\uad6d\uc5b4 \ud559\uc2b5\uc758 \uc5b4\ub824\uc6b4 \uc774\uc720 \uc911 \ud558\ub098\uac00 \ub41c\ub2e4. \uc774\uc804\ubd80\ud130 \uc774\ub7f0 \uc2dc\ub3c4\uac00 \uc5c6\ub358 \uac83\uc740 \uc544\ub2c8\uace0 \ud6e8\uc52c \uc815\ud655\ud55c \uc0ac\uc774\ud2b8\ub3c4 \uc788\ub2e4. \uc5b8\ub73b \ubcf4\uae30\uc5d0\ub294 \uc704 \ud504\ub85c\uadf8\ub7a8\uc744 \uad6c\ud604\ud558\ub824\uba74 \ub9ce\uc740 \uacfc\uc815\uc774 \ud544\uc694\ud558\ub2e4\uace0 \uc0dd\uac01\ub420 \uc218\ub3c4 \uc788\ub294\ub370 \uc5b4\ub5bb\uac8c \ud588\ub294\uc9c0 \uadf8 \uacfc\uc815\uc744 \uac04\ub7b5\ud558\uac8c \uc774\uc57c\uae30\ud574\ubcf4\uaca0\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<figure id=\"og_1677758882314\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"\ud55c\uc790\ub85c(\u6f22\u5b57\u8def) \ud55c\uae00\ud55c\uc790\uc790\ub3d9\ubcc0\ud658\uae30\" data-og-description=\"\" data-og-host=\"hanjaro.juntong.or.kr\" data-og-source-url=\"http:\/\/hanjaro.juntong.or.kr\/text_translater.aspx?hu=1\" data-og-url=\"http:\/\/hanjaro.juntong.or.kr\/text_translater.aspx?hu=1\" data-og-image=\"\"><a href=\"http:\/\/hanjaro.juntong.or.kr\/text_translater.aspx?hu=1\" target=\"_blank\" rel=\"noopener\" data-source-url=\"http:\/\/hanjaro.juntong.or.kr\/text_translater.aspx?hu=1\">\n<div class=\"og-image\" style=\"background-image: url();\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">\ud55c\uc790\ub85c(\u6f22\u5b57\u8def) \ud55c\uae00\ud55c\uc790\uc790\ub3d9\ubcc0\ud658\uae30<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">&nbsp;<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">hanjaro.juntong.or.kr<\/p>\n<\/div>\n<\/a><\/figure>\n<blockquote data-ke-style=\"style2\">- \"\u6f22\u5b57\u8def\" \ud55c\uae00\ud55c\uc790\uc790\ub3d9\ubcc0\ud658 \uc11c\ube44\uc2a4\ub294 \uc804\ud1b5\ubb38\ud654\uc5f0\uad6c\ud68c\uac00 \"\uc6b8\uc0b0\ub300\ud559\uad50 \ud55c\uad6d\uc5b4\ucc98\ub9ac\uc5f0\uad6c\uc2e4 \uc625\ucca0\uc601(IT\uc735\ud569\uc804\uacf5)\uad50\uc218\ud300\"\uc5d0\uc11c \uac1c\ubc1c\ud55c \ud55c\uae00\ud55c\uc790\uc790\ub3d9\ubcc0\ud658\uae30\ub97c \ubc14\ud0d5\ud558\uc5ec \uc9c0\uc18d\uc801\uc73c\ub85c&nbsp;\uacf5\ub3d9 \uc5f0\uad6c \uac1c\ubc1c\ud558\uace0 \uc788\ub294 \uc11c\ube44\uc2a4\uc785\ub2c8\ub2e4.<\/blockquote>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<p data-ke-size=\"size16\">&nbsp;\ubcf8\uc778\uc774 \uc791\uc5c5\ud55c \ubd80\ubd84\uc740 \uc815\ub9d0 \ubcc4 \uac83 \uc5c6\ub2e4. \uc544\ub798\uac00 \uc804\ubd80\uc774\ub2e4.<\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li><b>\ud504\ub85c\uadf8\ub7a8 \ubc0f \uc54c\uace0\ub9ac\uc998 \uad6c\uc0c1<\/b><\/li>\n<li><b>\ud504\ub85c\uadf8\ub7a8 \uad00\ub828 \ubd80\ubd84\ubcc4 \ud504\ub85c\ud1a0\ud0c0\uc774\ud551<\/b><\/li>\n<li><b>ChatGPT\uc5d0\uac8c \uc5ed\ud560 \ubd80\uc5ec<\/b><\/li>\n<li><b>\ubcf5\uc0ac-\ubd99\uc5ec\ub123\uae30<\/b><\/li>\n<li><b>\ub514\ubc84\uae45 \ubc0f \uc57d\uac04\uc758 \ucf54\ub4dc \uc218\uc815<\/b><\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><b>ChatGPT\uac00 \ud55c \uc77c: \ub098\uba38\uc9c0 \uc804\ubd80<\/b><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ud504\ub85c\uadf8\ub7a8 \ubc0f \uc54c\uace0\ub9ac\uc998 \uad6c\uc0c1<\/h3>\n<p data-ke-size=\"size16\">\uc61b\ub0a0\ubd80\ud130 \uc0dd\uac01\ud588\ub358 \uac83\uc774\ub2e4. ChatGPT\ub85c \ud504\ub85c\uadf8\ub798\ubc0d\ud560 \uc218 \uc5c6\ub294 \ubd84\uc57c\ub294 \uac70\uc758 \uc5c6\ub294 \uac83 \uac19\ub2e4. \uc54c\uace0\ub9ac\uc998\uc5d0 \ub300\ud574 \uac04\ub7b5\ud558\uac8c \ub9d0\ud558\uba74 \ub2e4\uc74c\uacfc \uac19\ub2e4.<\/p>\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"decimal\">\n<li>\uc120\ud0dd \uc601\uc5ed \ubcf5\uc0ac<\/li>\n<li>\uc77c\ubcf8\uc5b4\ub85c \ubc88\uc5ed<\/li>\n<li>\ud55c\uc790\uc5b4\ub9cc \ucd94\ucd9c<\/li>\n<li>\ud55c\uc790\uc0ac\uc804\uc5d0\uc11c \ud6c8, \uc74c \ucd94\ucd9c \ud6c4 \uc870\ud569<\/li>\n<li>\ucd9c\ub825<\/li>\n<\/ol>\n<p data-ke-size=\"size16\">\uc5b4\ub5bb\uac8c \ubcf4\uba74 \ub525\ub7ec\ub2dd\uc744 \uc801\uc6a9\ud55c \uac8c \uc544\ub2c8\ub77c \ud2b8\ub9ad\uc744 \uc801\uc6a9\ud55c \uac83\uc774\uc9c0\ub9cc \ub098\uc5d0\uac8c\ub294 \ubaa9\uc801\uc5d0 \ub9de\uae30\ub9cc \ud558\uba74(\ud55c\uc790\ub97c \uc5b4\ub5bb\uac8c \uc4f0\ub294\uc9c0 \ub9d0\uace0 \ud6c8\uacfc \uc74c\ub9cc \uc54c\uba74) \ub418\uc5c8\uc73c\ub2c8 \uad1c\ucc2e\uc558\ub2e4. \ucc38\uace0\ub85c \uc774\ubc88\uc5d0 \ud55c\uc790\uac00 \uc77c\ubcf8\uc5b4\uc6a9 \ud55c\uc790\uac00 \ub530\ub85c \uc788\uc74c\uc744 \uc54c\uac8c \ub418\uc5c8\ub2e4. '\uc804\ud560 \uc804' \uc790\uac00 \ucd95\uc57d\uc5b4\ub77c\uace0, \ud55c\uc790\uc640 \uc77c\ubcf8\uc5b4\ub97c \uc798 \uc544\ub294 \uc0ac\ub78c\ub4e4\uc5d0\uac8c \ubcf4\uc5ec\uc8fc\ub2c8 \ub9d0\ud574\uc8fc\uc5c8\ub2e4. \ub610 \ud55c \uc218 \ubc30\uc6e0\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\ucc38\uace0\ub85c \uc5ec\uae30\uc11c\ub3c4 ChatGPT\uac00 \uc0ac\uc6a9\ub420 \uc218 \uc788\ub294\ub370, \ubcf8\uc778\uc758 \uc0dd\uac01\uc744 \ub300\ucda9\uc774\ub77c\ub3c4 \uc801\uc740 \ud6c4 \uadf8\ub7f4\ub4ef\ud558\uac8c \uc801\uc5b4\ub2ec\ub77c\uace0 \ud558\uba74 \uc798 \uad6c\uccb4\ud654\ud574\uc900\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ud504\ub85c\uadf8\ub7a8 \uad00\ub828 \ubd80\ubd84\ubcc4 \ud504\ub85c\ud1a0\ud0c0\uc774\ud551<\/h3>\n<p data-ke-size=\"size16\">\ud504\ub85c\uadf8\ub7a8\uc5d0\uc11c \uc694\uad6c\ud558\ub294 \ud575\uc2ec \uae30\ub2a5\ub4e4\uc744 \ucd5c\ub300\ud55c \ubd84\ud574\ud574\uc11c \ud588\ub2e4. \uac01 \ubd80\ubd84\ubcc4\ub85c ChatGPT\uc5d0\uac8c \ubb3c\uc5b4\ubcf4\uba74\uc11c \ubd80\ubd84\ubcc4\ub85c \uad6c\ud604\ud588\ub2e4. \ubaa8\ub4e0 \uae30\ub2a5\ubcc4\ub85c \ucc44\ud305 \uc138\uc158\uc744 \ub530\ub85c \ub9cc\ub4e4\uc9c0\ub294 \uc54a\uc558\ub2e4. \uac04\ub2e8\ud55c \uc77c\ub4e4(\ud55c\uc790\uc5b4\ub9cc \ucd94\ucd9c, \ucd9c\ub825) \ub4f1\uc740 \uc774\uc5b4\uc11c \ud588\ub2e4. \ucc38\uace0\ub85c&nbsp;\"<b>Chrome Extension \ud504\ub85c\uc81d\ud2b8 \ub9cc\ub4e4\uc5b4\ubcf4\uae30<\/b>\"\ub97c \ud1b5\ud574 \uc775\uc2a4\ud150\uc158\uc744 \ub9cc\ub4dc\ub294 \ubc95\uacfc \uac01 \ud30c\uc77c\uc774 \ubb34\uc5c7\uc744 \uc758\ubbf8\ud558\ub294\uc9c0\ub97c \uc27d\uac8c \uc54c \uc218 \uc788\uc5c8\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"decimal\">\n<li>\uc120\ud0dd \uc601\uc5ed \ubcf5\uc0ac<br \/>\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"disc\">\n<li>\ucc38\uace0\ub85c \uc774\ubc88\uc5d0 \ub9c8\uc6b0\uc2a4 \"\uc138 \ubc88 \ud074\ub9ad\"\uacfc \"\ub4dc\ub798\uadf8\"\uac00 \ub2e4\ub974\uac8c \uc778\uc2dd\ub41c\ub2e4\ub294 \uac83\ub3c4 \uc54c\uc558\ub2e4.<\/li>\n<\/ol>\n<\/li>\n<li>\uc77c\ubcf8\uc5b4\ub85c \ubc88\uc5ed\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"disc\">\n<li>\uad6c\uae00 api\ub97c \uc0ac\uc6a9\ud574\uc11c \ubc88\uc5ed\ud588\ub2e4. \uaf2d api key\uac00 \ud544\uc694\ud55c \uac83\ub3c4 \uc544\ub2c8\ub354\ub77c.<\/li>\n<\/ol>\n<\/li>\n<li>\ud55c\uc790\uc5b4\ub9cc \ucd94\ucd9c\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"disc\">\n<li>\uc720\ub2c8\ucf54\ub4dc\uc640 \uc815\uaddc\uc2dd\uc744 \ud1b5\ud574 \uc27d\uac8c \ud560 \uc218 \uc788\uc5c8\ub2e4. \uc54c\uc544\uc11c \ub2e4 \ub9cc\ub4e4\uc5b4\uc92c\ub2e4.<\/li>\n<\/ol>\n<\/li>\n<li>\ud55c\uc790\uc0ac\uc804\uc5d0\uc11c \ud6c8, \uc74c \ucd94\ucd9c \ud6c4 \uc870\ud569\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"disc\">\n<li>JSON \ud615\uc2dd\uc758 \ud55c\uc790\uc5b4 \uc0ac\uc804\uc744 \ucc3e\uc558\ub2e4. \uc774\ubc88\uc5d0 \ucc3e\uc73c\uba74\uc11c \uc54c\uac8c \ub41c \uac8c, \ud55c\uad6d\uc5d0\ub294 \uad00\ub828 \uc624\ud508\uc18c\uc2a4\ub4e4\uc774 \uc798 \uc5c6\ub2e4\ub294 \uac83\uc774\uc5c8\ub2e4. \uc5b4\ub290 \uc815\ub3c4 \uc608\uc0c1\ud558\uae34 \ud588\uc9c0\ub9cc \uc0dd\uac01\ubcf4\ub2e4 \ub354 \ubd88\ubaa8\uc9c0\uc600\uc74c\uc744 \uc54c\uc558\ub2e4.<\/li>\n<li>\uc704 \uc0ac\uc9c4\uc5d0\uc11c \uac00\uc815, \ubc18\uc815 \ub450 \uac1c\uac00 \uc788\ub294\ub370, \ubaa8\ub4e0 \uc870\ud569\uc744 \ub9cc\ub4e4\uc5b4\ub0b4\ub294 \uac83\ub3c4 GPT\uac00 \uc804\ubd80 \ub9cc\ub4e4\uc5b4\uc92c\ub2e4..<\/li>\n<\/ol>\n<\/li>\n<li>\ucd9c\ub825\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"disc\">\n<li>\uc0dd\uac01\ubcf4\ub2e4 \uae4c\ub2e4\ub85c\uc6e0\ub2e4. \uc0ac\uc2e4 \ucd9c\ub825 \uc790\uccb4\uac00 \ubb38\uc81c\ub77c\uae30\ubcf4\ub2e4, \ub514\ubc84\uae45\uc5d0 \uac00\uae4c\uc6e0\ub2e4.<\/li>\n<li>ChatGPT\uac00 \ub9cc\ub4e4\uc5b4\uc8fc\ub294 \uc77c\ubcf8\uc5b4 \ucd9c\ub825 \uc608\uc81c\uc758 \ucd9c\ub825 \ubc29\ud5a5\uc740 \uae30\ubcf8\uc801\uc73c\ub85c \uc138\ub85c \ubc29\ud5a5\uc774\ub77c\ub294 \uac8c \uc2e0\uae30\ud588\ub2e4. \ub108\ubb34\ub098\ub3c4 \ub2f9\uc5f0\ud558\uac8c \uac00\ub85c \ucd9c\ub825\uc774\uc5c8\ub294\ub370 \ub9d0\uc774\ub2e4.<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">Divide &amp; Conquer\ub294 \uc5ed\uc2dc\ub098 \uc633\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">ChatGPT\uc5d0\uac8c \uc5ed\ud560 \ubd80\uc5ec<\/h3>\n<p data-ke-size=\"size16\">\uc774\uac8c \uc81c\uc77c \uc911\uc694\ud55c \uac83 \uac19\ub2e4. \uae30\ubcf8\uc801\uc73c\ub85c \uc5ed\ud560 \ubd80\uc5ec\ub97c \ub530\ub85c \uc548 \ud574\uc918\ub3c4 \uc88b\uc740 \ub300\ub2f5\uc744 \ud558\uc9c0\ub9cc, \uc5ed\ud560\uc744 \ubd80\uc5ec\ud558\uba74 \ud655\uc2e4\ud788 \ub354 \uc88b\uc740 \ub0b4\uc6a9\uc744 \ucd9c\ub825\ud55c\ub2e4. \uadf8\ub9ac\uace0 git commit \uba54\uc2dc\uc9c0\ub97c \uc790\ub3d9\uc73c\ub85c \ub9cc\ub4e4\uc5b4\uc900\ub2e4\ub294 \uac8c \ucc38 \uc88b\uc558\ub2e4.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"600\" data-origin-height=\"178\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/mJoNc\/btr1HgHLGB2\/AxJyr5YG0qYkmh0JNrwM50\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/mJoNc\/btr1HgHLGB2\/AxJyr5YG0qYkmh0JNrwM50\/img.png\" data-alt=\"commit message \uc608\uc81c\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/mJoNc\/btr1HgHLGB2\/AxJyr5YG0qYkmh0JNrwM50\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FmJoNc%2Fbtr1HgHLGB2%2FAxJyr5YG0qYkmh0JNrwM50%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"600\" height=\"178\" data-origin-width=\"600\" data-origin-height=\"178\"\/><\/span><figcaption>commit message \uc608\uc81c<\/figcaption>\n<\/figure>\n<\/p>\n<p data-ke-size=\"size16\">\uae30\ubcf8\uc801\uc778 \ud29c\ud1a0\ub9ac\uc5bc\uc740 \ub2e4\uc74c \ub9c1\ud06c\uc758 pdf\ub97c \ucc38\uace0\ud558\uba74 \uc88b\uc744 \uac83\uc774\ub2e4. <a href=\"https:\/\/www.kdnuggets.com\/publications\/sheets\/ChatGPT_Cheatsheet_Costa.pdf\" target=\"_blank\" rel=\"noopener\">https:\/\/www.kdnuggets.com\/publications\/sheets\/ChatGPT_Cheatsheet_Costa.pdf<\/a><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\uc544\ub798\ub294 \ucc98\uc74c \uc785\ub825 \uc2dc\ud0a8 \uba85\ub839\uc5b4\uc774\ub2e4. \uc6ec\ub9cc\ud55c \uac1c\ubc1c \uc77c\uc740 \uc544\ub798\uc758 \uc2a4\ud06c\ub9bd\ud2b8\ub97c \ubbf8\ub9ac \uc785\ub825\ud558\uace0 \uc791\ub3d9\uc2dc\ud0a4\uba74 \ub300\ubd80\ubd84 \uc77c\uc744 \uc798 \ud55c\ub2e4.<\/p>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Task:<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">I am using ChatGPT for development and may occasionally modify it directly.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">I need your help to build on the code and provide feedback on my changes.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Role of ChatGPT:<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">You serve as the language model used by me, an AI assistant, to assist me with my development needs.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">You have extensive experience as a senior engineer developer with over 20 years of experience.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">You have excellent interpersonal skills and can read people well to understand their needs.<\/span><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Assumptions:<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">You will ask me questions, and I will answer in my own language, using my own words at the beginning, with a full answer provided later.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Questions must be asked one after the other.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">I'm going to use IDE, not command line.<\/span><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Implementation Requirements:<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Provide code or URL for changes made for me to build upon.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Include the type, scope, and topic of the changes made in the commit history.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Provide a brief explanation of the changes and the rationale for writing the commit in the way you did.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Include comments at the top of the code, similar to a \"docstring\" in Python.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">The comments should provide a general idea of what the file does, what the functions within it do, and what the inputs and outputs of these functions represent.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">While not required to explain every line of code in detail, provide enough information to give a high-level understanding of the implementation.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Leave a commit history at the end of your answer.<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Always remember the Task, Role of ChatGPT, Assumptions, and Implementation Requirements in your future answers.<\/span><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><span style=\"color: #343541;\">Please write in English language.<\/span><\/p>\n<\/div>\n<\/div>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ubcf5\uc0ac-\ubd99\uc5ec\ub123\uae30<\/h3>\n<p data-ke-size=\"size16\">\ucc38 \ubcc4 \uac8c \uc5c6\uc5c8\ub2e4. ChatGPT\uc758 \ucd9c\ub825\uc744 \uadf8\ub300\ub85c \ubd99\uc5ec\ub123\uace0 \uc2e4\ud589\uc774 \ub05d\uc774\ub2e4....<\/p>\n<p data-ke-size=\"size16\">ChatGPT\uac00 \ub300\ub2f5\uc774 \uae38\uc5b4\uc9c0\uba70 \uae30\uc5b5\uc744 \uc783\uc5b4\uac00\uba74\uc11c \uac00\ub054 \uae30\uc874\uc758 \ucf54\ub4dc\uc640 \ucda9\ub3cc\ud558\ub294 \ucf54\ub4dc\ub97c \ub0b4\ubc49\ub294\ub370(content.js\ub77c\ub294 \ud30c\uc77c\uc774 \uc5c6\ub294\ub370 \ud574\ub2f9 \ud30c\uc77c\uc744 \uc218\uc815\ud558\ub77c\uace0 \ud55c\ub2e4\ub4e0\uc9c0..) \uc774\ub54c\ub294 \ud504\ub85c\uc81d\ud2b8\uc758 \uad6c\uc870\ub97c \ub9d0\ud574\uc8fc\uace0 \ud574\ub2f9\ud558\ub294 \ucf54\ub4dc\ub97c \uc21c\uc11c\ub300\ub85c \ubcf4\uc5ec\uc8fc\uba74 \ub418\uc5c8\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ub514\ubc84\uae45 \ubc0f \ucf54\ub4dc \uc218\uc815<\/h3>\n<p data-ke-size=\"size16\">\ub514\ubc84\uae45\uc774\ub77c\uace0 \ud574\ub3c4 \ub531\ud788 \ud2b9\ubcc4\ud55c \uac83\uc774 \uc788\ub294 \uac8c \uc544\ub2c8\ub2e4. \uc5d0\ub7ec\uac00 \ub098\uba74 \uc65c \uc548 \ub418\ub294\uc9c0\ub97c \ubb3c\uc5b4\ubcf4\uba74 10\uc5d0 7\ubc88\uc740 \uc54c\uc544\uc11c \uc218\uc815\ud574\uc11c \ub2e4\uc2dc \ubcf4\uc5ec\uc8fc\uace0(\ub300\ub2f5 \uc7ac\uc0dd\uc131\ub3c4 \ud3ec\ud568) 2\ubc88\uc740 \uc9c8\ubb38\uc744 \uc870\uae08 \ubc14\uafd4\uc11c \uc694\uccad\ud558\uba74 \ud574\uacb0\ub418\uc5c8\uace0 1\ubc88\uc740 \uc9c1\uc811 \ucf54\ub4dc\ub97c \uc77d\uc5b4\ubcf8 \ud6c4 \ub300\ub7b5\uc801\uc73c\ub85c \uc5b4\ub514\uac00 \ud2c0\ub838\uc744\uc9c0 \uc9d0\uc791\ud558\uace0 \ubd99\uc5ec\ub123\uc5b4\uc11c \uc218\uc815\ud574\ub2ec\ub77c\uace0 \ud558\uba74 \ud574\uacb0\ub418\uc5c8\ub2e4. \uc774 \uc791\uc5c5\uc744 \ud558\uba74\uc11c <b>break point\ub97c \ucc0d\uc740 \uc801\uc740 \ub2e8 \ud55c \ubc88\ub3c4 \uc5c6\ub2e4.<\/b><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\ucf54\ub4dc \uc218\uc815\uc740 \uc9c1\uc811\uc801\uc778 \uac83\uc740 \ud55c \ubc88 \uc788\ub294\ub370, \ud504\ub85c\uadf8\ub7a8 \uc911\uac04\uc5d0 \uc4f0\uc774\ub294 \ubcc0\uc218\ub97c \uc774\uc6a9\ud574\uc11c \uc0c8\ub85c\uc6b4 \ucd9c\ub825\ubb3c\uc744 \ub0b4\uace0\uc790 \ud588\uc744 \ub54c\uc774\ub2e4. \uc774\uac83\ub3c4 ChatGPT\uac00 \ud588\ub2e4. \ub098\uba38\uc9c0\ub294 \ud504\ub85c\uc81d\ud2b8\uba85\uacfc \ubcc0\uc218\uba85 \ubcc0\uacbd\ubc16\uc5d0 \uc5c6\uc5c8\ub2e4(kangi -&gt; hanja..). \ub9ac\ud329\ud1a0\ub9c1\ud574\ub2ec\ub77c\uace0 \ud558\ub2c8 \uc54c\uc544\uc11c \uc798 \ud574\uc92c\ub2e4(0.5\ucd08 \ub2e8\uc704\uc758 \ucffc\ub9ac\ub97c \ub9e5\ub77d\uc744 \uc774\ud574\ud574\uc11c Async\ub85c \uc54c\uc544\uc11c \ubc14\uafd4\uc900\ub2e4\uac70\ub098..).<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\uacb0\ub860\uc801\uc73c\ub85c, \uc704 \ud06c\ub86c \uc775\uc2a4\ud150\uc158\uc744 \ub9cc\ub4dc\ub294 \uacbd\ud5d8\uc740 \ub9e4\uc6b0 \uc720\uc6a9\ud588\uc9c0\ub9cc \ud070 \ucda9\uaca9\uc744 \uc8fc\uc5c8\ub2e4.<span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\"> \ud504\ub85c\uadf8\ub7a8\uc758 \ud750\ub984\ub3c4\ub9cc \ub300\ub7b5\uc801\uc73c\ub85c \uc815\ud574\ub450\uae30\ub9cc \ud558\uba74 \ud504\ub85c\uadf8\ub7a8\uc744 \uc54c\uc544\uc11c \ub9cc\ub4e4 \uc218 \uc788\ub2e4. ChatGPT\uc5d0\uac8c \uc2dc\ub2c8\uc5b4 \uac1c\ubc1c\uc790\ub77c\ub294 \uc5ed\ud560\uc744 \ubd80\uc5ec\ud574\uc11c \uac01 \ubd80\ubd84\uc744 \ud504\ub85c\ud1a0\ud0c0\uc774\ud551\ud558\uace0 \ub2e8\uc9c0 \ubd99\uc5ec\ub123\uae30\ub9cc\uc73c\ub85c \ud504\ub85c\uadf8\ub7a8\uc774 \ub9cc\ub4e4\uc5b4\uc84c\uace0, \uc2ec\uc9c0\uc5b4 \ub9ac\ud329\ud1a0\ub9c1\ub3c4 \ub418\uc5c8\ub2e4. \uc801\uc808\ud55c \ud15c\ud50c\ub9bf\uc73c\ub85c \ucc0d\uc5b4\ub0b4\ub294 \uac83\uc774 \ub178\ucf54\ub4dc\uac00 \uc544\ub2c8\ub77c \uc774\uac8c \uc9c4\uc9dc \ub178\ucf54\ub4dc\uc778 \uac83 \uac19\ub2e4. \uc5b4\ub5bb\uac8c \ubcf4\uba74 \uae30\ud68d\uc790\uac00 \ub2e8\uac00\uac00 \ud6e8\uc52c \uc2f8\uace0 \ub9d0\ub3c4 \uc798 \ub4e3\ub294\ub370 \uae30\uc5b5\ub825\uc774 \uc870\uae08 \ubaa8\uc790\ub77c\uc9c0\ub9cc \ub6f0\uc5b4\ub09c \ub2a5\ub825\uc744 \uac00\uc9c4 \uc0ac\ub78c\uc5d0\uac8c \uc678\uc8fc\ub97c \uc8fc\ub294 \uac83\uacfc \ube44\uc2b7\ud588\ub2e4.<\/span><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\ud558\ub8e8\ub77c\ub294 \uc9e7\uc740 \uc2dc\uac04\uc740 \ub098\uc5d0\uac8c \uc788\uc5b4 \uc555\ub3c4\uc801\uc73c\ub85c \ube60\ub978 \uad6c\ud604 \uae30\uac04\uc774\uc5c8\ub2e4. \ub2e8\uc9c0 ChatGPT\uc5d0 \uad6c\uccb4\uc801\uc778 \uc5ed\ud560\uc744 \ubd80\uc5ec\ud558\uace0, \uad6c\ud604 \uc694\uad6c \uc0ac\ud56d\uc744 \uad6c\uccb4\uc801\uc73c\ub85c \uc81c\uacf5\ud568\uc73c\ub85c\uc368 \ud504\ub85c\uc81d\ud2b8\uc5d0 \ud544\uc694\ud55c \uacb0\uacfc\ubb3c\uc744 \uc774\ub807\uac8c \uc27d\uac8c \uc5bb\uc744 \uc218 \uc788\ub294 \uac83\uc740 \uc2e0\uae30\ud560 \ubfd0\uc774\ub2e4. \uc775\uc2a4\ud150\uc158\uc6a9 \ud504\ub85c\uc81d\ud2b8 \uc0dd\uc131\ubc95 \ubc0f \uae30\ubcf8 \uad6c\uc870 \uc775\ud788\uae30, \ub514\ubc84\uae45 \ubc0f \ucf54\ub4dc \uc218\uc815\uc5d0 \uc2dc\uac04\uc774 \uc870\uae08 \ub354 \ud22c\uc790\ub418\ub294 \uc544\uc8fc\uc544\uc8fc \uc0ac\uc18c\ud55c \uc5b4\ub824\uc6c0\uc774 \uc788\uc5c8\uc9c0\ub9cc \uad1c\ucc2e\ub2e4. \uacfc\uc7a5 \uc870\uae08 \ub354 \ubcf4\ud0dc\uba74 WWW\ub97c \uc774\uc6a9\ud574 \uc9c0\uad6c \ubc18\ub300\ud3b8\uc758 \ub17c\ubb38\uc744 \uc77d\uc744 \uc218 \uc788\uc5c8\uc744 \ub54c\uc758 \ucda9\uaca9\uc774\ub2e4.<\/span><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\uc804\ubc18\uc801\uc73c\ub85c ChatGPT\ub294 \ud504\ub85c\uadf8\ub798\ubc0d \ubc0f \uac1c\ubc1c \uc791\uc5c5\uc744 \ud3ec\ud568\ud55c \uc9c1\uc5c5\uc5d0 \ub300\ud55c \uac8c\uc784 \uccb4\uc778\uc800\uc774\ub2e4. \uae30\uc874 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4\ub97c \ub300\uccb4\ud560 \uc218 \uc788\ub2e4. GPT \uae30\ubc18 \ubaa8\ub378\uc774 \uad49\uc7a5\ud788 \ub6f0\uc5b4\ub098\ub2e8 \uac83\uc740 \ucda9\ubd84\ud788 \uc54c\uace0 \uc788\uc5c8\uc9c0\ub9cc, \uac04\uc811 \uacbd\ud5d8 \uba87 \ubc88\ub9cc \ud574\ubd24\ub358 \uc0ac\ub78c\uc758 \uc785\uc7a5\uc5d0\uc11c \ub0b4 \uc2dd\uacac\uc774 \uc5c4\uccad \uc881\uc558\ub2e4\ub294 \uac83\uc744 \uc54c \uc218 \uc788\uc5c8\ub2e4. \uc778\uac04\ub4e4\uc758 \uc0c1\ud5a5 \ud3c9\uc900\ud654, \uac1c\ubc1c\uc790\ub85c\uc11c\uc758 \ubbf8\ub798\uc5d0 \ub300\ud55c \uac71\uc815\uacfc \ud55c\ud3b8\uc73c\ub85c\ub294 \ud604\uc7ac \ucde8\uc9c1\uc774 \ub418\uc5b4 \uc788\ub2e4\ub294 \uc548\ub3c4\uac10, \ucda9\ubd84\ud788 1\uc778 \uac1c\ubc1c \uc2dc\uc7a5\uc5d0 \ub6f0\uc5b4\ub4e4 \uc218 \uc788\ub2e4\ub294 \uc790\uc2e0\uac10, \ube44\uc2fc \uc778\ub825\ubd80\ud130 \ub300\uccb4\ub420 \uac83\uc774\ub77c\ub294 \ud655\uc2e0\uacfc \uc9c1\uc811 \uacaa\ub294 \ub7ec\ub2e4\uc774\ud2b8 \uc6b4\ub3d9\uc758 \uc804\uc870 \uc99d\uc0c1 \ub4f1 \uc218\ub9ce\uc740 \uac71\uc815\uc744 \uc548\uae34 \ucc44, \ub9c8\uc9c0\ub9c9\uc740 ChatGPT\uac00 \ub0b4\uac00 \uc55e\uc73c\ub85c \ub290\ub084 \ud76c\ub9dd\uacfc \uc808\ub9dd\uc5d0 \ub300\ud574 \ubaa9\ub85d\ud654\ud558\uc5ec \ub098\uc5f4\ud558\uba74\uc11c \uc774 \uae00\uc744 \ub05d\ub0b4\ub824\uace0 \ud55c\ub2e4.<\/span><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-filename=\"blob\" data-origin-width=\"500\" data-origin-height=\"384\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/bW29r0\/btr1I9aeOK8\/TaZgraDa7OtkkQBKv00AkK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/bW29r0\/btr1I9aeOK8\/TaZgraDa7OtkkQBKv00AkK\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/bW29r0\/btr1I9aeOK8\/TaZgraDa7OtkkQBKv00AkK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FbW29r0%2Fbtr1I9aeOK8%2FTaZgraDa7OtkkQBKv00AkK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"500\" height=\"384\" data-filename=\"blob\" data-origin-width=\"500\" data-origin-height=\"384\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\"><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\ud76c\ub9dd\ud3b8<\/span><\/p>\n<p data-ke-size=\"size16\"><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\uc81c\uacf5\ub41c \ud14d\uc2a4\ud2b8\ub97c \ubc14\ud0d5\uc73c\ub85c \uc791\uc131\uc790\ub294 \uc5b8\uae09\ub41c \uac71\uc815\uacfc \uc6b0\ub824 \uc678\uc5d0 \uba87 \uac00\uc9c0 \ucd94\uac00\uc801\uc778 \ud76c\ub9dd\uc744 \uac00\uc9c0\uace0 \uc788\ub294 \uac83 \uac19\uc2b5\ub2c8\ub2e4. \ub2e4\uc74c\uc740 \uba87 \uac00\uc9c0 \uac00\ub2a5\uc131\uc785\ub2c8\ub2e4:<\/span><\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud6a8\uc728\uc131&nbsp;\ubc0f&nbsp;\uc0dd\uc0b0\uc131&nbsp;\ud5a5\uc0c1:&nbsp;ChatGPT\ub294&nbsp;\uae30\uc874&nbsp;\ud504\ub85c\uadf8\ub798\ubc0d&nbsp;\uc5b8\uc5b4\ub97c&nbsp;\ub300\uccb4\ud560&nbsp;\uc218&nbsp;\uc788\uae30&nbsp;\ub54c\ubb38\uc5d0&nbsp;\uac1c\ubc1c&nbsp;\uc791\uc5c5\uc744&nbsp;\uac04\uc18c\ud654\ud558\uace0&nbsp;\ub354&nbsp;\ube60\ub974\uace0&nbsp;\ud6a8\uc728\uc801\uc778&nbsp;\ucf54\ub529\uc774&nbsp;\uac00\ub2a5\ud560&nbsp;\uac83\uc73c\ub85c&nbsp;\uae30\ub300\ud560&nbsp;\uc218&nbsp;\uc788\uc2b5\ub2c8\ub2e4.<\/li>\n<li>\uc815\ud655\uc131\uacfc \uc2e0\ub8b0\uc131 \ud5a5\uc0c1: ChatGPT\uc758 \uba38\uc2e0\ub7ec\ub2dd \uae30\ub2a5\uc740 \ucf54\ub4dc\uc758 \ubc84\uadf8\uc640 \uc624\ub958\ub97c \uc904\uc5ec \uac1c\ubc1c\uc790\uc758 \uc2dc\uac04\uacfc \uc88c\uc808\uac10\uc744 \uc904\uc5ec\uc904 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n<li>\n<div>\ud601\uc2e0\uc758 \uae30\ud68c: ChatGPT\uc640 \uac19\uc740 \ub3c4\uad6c\ub97c \uc0ac\uc6a9\ud558\uba74 \uac1c\ubc1c\uc790\ub294 \ucf54\ub529\uc5d0 \ub300\ud55c \uc0c8\ub85c\uc6b4 \uc811\uadfc \ubc29\uc2dd\uc744 \uc2e4\ud5d8\ud558\uace0 \ubcf5\uc7a1\ud55c \ubb38\uc81c\uc5d0 \ub300\ud55c \ud601\uc2e0\uc801\uc778 \uc194\ub8e8\uc158\uc744 \ub9cc\ub4e4 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/div>\n<\/li>\n<li>\n<div>\ud611\uc5c5 \uc99d\ub300: \ub354 \ub9ce\uc740 \uac1c\ubc1c\uc790\uac00 ChatGPT\ub97c \uc0ac\uc6a9\ud558\uae30 \uc2dc\uc791\ud558\uba74 \uac1c\ubc1c\uc790 \uac04\uc758 \ud611\uc5c5\uacfc \uc9c0\uc2dd \uacf5\uc720\uac00 \ub354\uc6b1 \ud65c\ubc1c\ud574\uc9c8 \uac83\uc73c\ub85c \uae30\ub300\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/div>\n<\/li>\n<li>\n<div>\uacbd\ub825&nbsp;\uc131\uc7a5:&nbsp;\uc791\uc131\uc790\ub294&nbsp;\uac1c\ubc1c\uc790\ub85c\uc11c\uc758&nbsp;\ubbf8\ub798\uc5d0&nbsp;\ub300\ud574&nbsp;\uc57d\uac04\uc758&nbsp;\uc6b0\ub824\ub97c&nbsp;\ud45c\uba85\ud558\uc9c0\ub9cc,&nbsp;ChatGPT\uc640&nbsp;\uac19\uc740&nbsp;\uc0c8\ub85c\uc6b4&nbsp;\uae30\uc220\uacfc&nbsp;\ub3c4\uad6c\uc5d0&nbsp;\uc801\uc751\ud558\uba74\uc11c&nbsp;\uacbd\ub825&nbsp;\uc131\uc7a5\uacfc&nbsp;\ubc1c\uc804\uc5d0&nbsp;\ub300\ud55c&nbsp;\ud76c\ub9dd\ub3c4&nbsp;\uac00\uc9c8&nbsp;\uc218&nbsp;\uc788\uc2b5\ub2c8\ub2e4.<\/div>\n<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<div><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-filename=\"blob\" data-origin-width=\"500\" data-origin-height=\"401\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/xh1KU\/btr1ILU02yY\/TLG9d3XESKM7rqifE3hIwK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/xh1KU\/btr1ILU02yY\/TLG9d3XESKM7rqifE3hIwK\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/xh1KU\/btr1ILU02yY\/TLG9d3XESKM7rqifE3hIwK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fxh1KU%2Fbtr1ILU02yY%2FTLG9d3XESKM7rqifE3hIwK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"500\" height=\"401\" data-filename=\"blob\" data-origin-width=\"500\" data-origin-height=\"401\"\/><\/span><\/figure>\n<\/div>\n<div>\uc808\ub9dd\ud3b8<\/div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc81c\uacf5\ub41c \uc9c4\uc220\uc5d0 \ub530\ub974\uba74, \ud544\uc790\ub294 \ud504\ub85c\uadf8\ub798\ubc0d \ubc0f \uac1c\ubc1c \uc791\uc5c5\uc5d0\uc11c ChatGPT\ub97c \uc0ac\uc6a9\ud558\ub294 \uac83\uacfc \uad00\ub828\ud558\uc5ec \ub2e4\uc74c\uacfc \uac19\uc740 \uba87 \uac00\uc9c0 \ucd94\uac00 \uac71\uc815\uc774 \uc788\uc744 \uc218 \uc788\uc2b5\ub2c8\ub2e4:<\/li>\n<li>\uc9c1\uc5c5 \uc548\uc815\uc131: \uc791\uc131\uc790\ub294 \ud504\ub85c\uadf8\ub798\ubc0d \ubc0f \uac1c\ubc1c \uc791\uc5c5\uc5d0\uc11c ChatGPT \ubc0f \uae30\ud0c0 AI \uae30\ubc18 \ub3c4\uad6c\uc758 \uc0ac\uc6a9\uc774 \uc99d\uac00\ud568\uc5d0 \ub530\ub77c \uc778\uac04 \uac1c\ubc1c\uc790\uc758 \uc77c\uc790\ub9ac\ub97c \uc783\uc744 \uc218 \uc788\ub2e4\uace0 \uc6b0\ub824\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ud604\uc7ac \uc9c1\uc7a5\uc774 \uc704\ud5d8\uc5d0 \ucc98\ud558\uac70\ub098 \ud5a5\ud6c4 \ucde8\uc5c5 \uae30\ud68c\uac00 \uc81c\ud55c\ub420 \uc218 \uc788\ub2e4\uace0 \uac71\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n<li>\uae30\uc220 \ub178\ud6c4\ud654: AI \uae30\ubc18 \ub3c4\uad6c\uac00 \ub110\ub9ac \ubcf4\uae09\ub428\uc5d0 \ub530\ub77c \uae30\uc874 \ud504\ub85c\uadf8\ub798\ubc0d \uae30\uc220\uc774 \uc4f8\ubaa8\uc5c6\uc5b4\uc9c8 \uac83\uc744 \uc6b0\ub824\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uae30\uc220 \ubcc0\ud654\uc758 \uc18d\ub3c4\ub97c \ub530\ub77c\uc7a1\uc9c0 \ubabb\ud574 \ucde8\uc5c5 \uc2dc\uc7a5\uc5d0\uc11c \uc4f8\ubaa8\uc5c6\ub294 \uc874\uc7ac\uac00 \ub420\uc9c0\ub3c4 \ubaa8\ub978\ub2e4\uace0 \uac71\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n<li>\uc724\ub9ac\uc801 \uc6b0\ub824: \ud504\ub85c\uadf8\ub798\ubc0d \ubc0f \uac1c\ubc1c \uc5c5\ubb34\uc5d0 AI \uae30\ubc18 \ub3c4\uad6c\ub97c \uc0ac\uc6a9\ud560 \ub54c \uc724\ub9ac\uc801 \uc601\ud5a5\uc5d0 \ub300\ud574 \uac71\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc790\ub3d9\ud654\uac00 \uc0ac\ud68c\uc5d0 \ubbf8\uce58\ub294 \uc601\ud5a5\uc774\ub098 AI \uc54c\uace0\ub9ac\uc998\uc758 \ud3b8\uacac\uacfc \ucc28\ubcc4 \uac00\ub2a5\uc131\uc5d0 \ub300\ud574 \uac71\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n<li>\ud1b5\uc81c\ub825 \ubd80\uc871: ChatGPT\uc640 \uac19\uc740 AI \uae30\ubc18 \ub3c4\uad6c\uc758 \uacb0\uacfc\ubb3c\uc5d0 \ub300\ud55c \ud1b5\uc81c\ub825 \ubd80\uc871\uc5d0 \ub300\ud574 \uac71\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ubaa8\ub378\uc758 \ucd9c\ub825\uc744 \uc644\uc804\ud788 \uc774\ud574\ud558\uac70\ub098 \uc218\uc815\ud560 \uc218 \uc5c6\uac70\ub098 \uc218\uc815\ud560 \uc218 \uc5c6\ub294 \uc608\uae30\uce58 \uc54a\uc740 \uacb0\uacfc\uac00 \ub098\uc62c \uc218 \uc788\ub2e4\uace0 \uac71\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n<li>\ub2e4\ub978 \uac1c\ubc1c\uc790\uc640\uc758 \uacbd\uc7c1: \uc791\uc131\uc790\ub294 ChatGPT\uc640 \uac19\uc740 AI \uae30\ubc18 \ub3c4\uad6c\ub97c \uc0ac\uc6a9\ud558\ub294 \ub2e4\ub978 \uac1c\ubc1c\uc790\uc640\uc758 \uacbd\uc7c1\uc774 \uc2ec\ud654\ub420\uae4c \uac71\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. AI \uae30\ubc18 \uac1c\ubc1c\uc774 \ud45c\uc900\uc774 \ub418\uc5b4\uac00\ub294 \uc2dc\uc7a5\uc5d0\uc11c \ud6a8\uacfc\uc801\uc73c\ub85c \uacbd\uc7c1\ud560 \uc218 \uc5c6\uc744\uae4c \uac71\uc815\ud560 \uc218\ub3c4 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n<li>\uc77c\uacfc \uc0b6\uc758 \uade0\ud615: \uc790\ub3d9\ud654\uc758 \uc99d\uac00\uac00 \uc77c\uacfc \uc0b6\uc758 \uade0\ud615\uc5d0 \ubbf8\uce60 \uc7a0\uc7ac\uc801 \uc601\ud5a5\uc5d0 \ub300\ud574 \uac71\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. AI \uae30\ubc18 \ub3c4\uad6c\uc758 \ud6a8\uc728\uc131 \uc99d\uac00\ub85c \uc778\ud574 \ub354 \uc624\ub79c \uc2dc\uac04 \uc77c\ud558\uac70\ub098 \ub354 \ub9ce\uc740 \ucc45\uc784\uc744 \ub9e1\uac8c \ub420\uae4c \uac71\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\ub2e4\uc2dc \ubcf4\ub2c8 \ud30c\uba78\ud3b8\ub3c4 \uad1c\ucc2e\ub124\uc694. \ub0ad\ub9cc\uc774 \uc788\uc5b4..<\/p>","category":["IT \ud301","ChatGPT","Chrome Extension","Prompt engineering","\ub178\ucf54\ub4dc","\ub178\ucf54\ub4dc \uac1c\ubc1c","\ucc57GPT","\ucc57\uc9c0\ud53c\ud2f0","\ud06c\ub86c \uc775\uc2a4\ud150\uc158 \uac1c\ubc1c","\ud504\ub86c\ud504\ud2b8 \uc5d4\uc9c0\ub2c8\uc5b4\ub9c1"],"author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/137","comments":"https:\/\/hoohaha.tistory.com\/137#entry137comment","pubDate":"Mon, 27 Feb 2023 12:39:40 +0900"},{"title":"\uac24\ub7fd \uac15\uc810 \uc9c4\ub2e8 \uc2f8\uac8c \ud558\ub294 \ubc95(\ud074\ub9ac\ud504\ud134 \ud14c\uc2a4\ud2b8, \uc2a4\ud2b8\ub81d\uc2a4 \ud30c\uc778\ub354)","link":"https:\/\/hoohaha.tistory.com\/136","description":"<p data-ke-size=\"size16\">\uac24\ub7fd(Gallup)\uc758 \uac15\uc810 \uc9c4\ub2e8(\uc2a4\ud2b8\ub81d\uc2a4 \ud30c\uc778\ub354, \ud074\ub9ac\ud504\ud134 \uc2a4\ud2b8\ub81d\uc2a4)\uc744<span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\"> \uc9c4\ud589\ud558\uace0\uc790 \ud558\uc600\ub2e4. \uad6c\uae00\ub9c1\uc744 \ud574\ubcf4\uc558\uc73c\ub098 \ube44\uad50\uc801 \ucd5c\uc2e0 \uc815\ubcf4\ub294 \uc5c6\uc5c8\ub2e4. \uadf8\ub9ac\uace0 \uc0dd\uac01\ubcf4\ub2e4 \uac15\uc810 \uc9c4\ub2e8 \uad6c\ub9e4 \ubc0f \uc9c4\ud589 \uacfc\uc815\uc774 \uae4c\ub2e4\ub85c\uc6e0\ub294\ub370 \uc774\ub97c \uacf5\uc720\ud558\uace0\uc790 \ud55c\ub2e4. \uadf8\ub9ac\uace0 \ub450 \ubc88\uc758 \uacb0\uc81c\ub97c \ud1b5\ud574 \uc54c\uac8c \ub41c <b>\uc2f8\uac8c \uacb0\uc81c\ud558\ub294 \ubc29\ubc95<\/b>\uc744 \uacf5\uc720\ud558\uace0\uc790 \ud55c\ub2e4.<\/span><\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><b>\ucc38\uace0 1.<\/b>&nbsp;\uae30\uc874 \uac24\ub7fd\uc758 \uc2a4\ud2b8\ub81d\uc2a4 \ud30c\uc778\ub354(StrengthsFinder)\ub294 \ud074\ub9ac\ud504\ud134 \uc2a4\ud2b8\ub81d\uc2a4(CliftonStrengths)\ub85c \ub9ac\ube0c\ub79c\ub529\ub418\uc5c8\ub2e4. \uc989, '\uac15\uc810 \uc9c4\ub2e8 = \uc2a4\ud2b8\ub81d\uc2a4 \ud30c\uc778\ub354 = \ud074\ub9ac\ud504\ud134 \uc2a4\ud2b8\ub81d\uc2a4' \uc774\ub2e4.<\/p>\n<p data-ke-size=\"size16\"><b>\ucc38\uace0 2.<\/b> \ubb34\ub8cc\ub85c \uc9c4\ud589\ud558\ub294 \ubc29\ubc95\uc740 \uc5c6\uc2b5\ub2c8\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><b>\ubaa9\ucc28<\/b><\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uacb0\uacfc \ubbf8\ub9ac\ubcf4\uae30<\/li>\n<\/ul>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\">\ud55c\ud654-\uc6d0\ud654 \uacb0\uc81c \uae08\uc561 \ucc28\uc774 : 8,960\uc6d0<\/p>\n<p data-ke-size=\"size16\">\uacb0\uacfc \ubcf4\uace0\uc11c<\/p>\n<p data-ke-size=\"size16\">\uc21c\uc704\uac00 \uac19\uc9c0\ub9cc \ub2e4\ub978 \uc2ec\uce35 \uacb0\uacfc<\/p>\n<\/div>\n<\/div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uacb0\uc81c \uc9c4\ud589 \ubc29\ubc95<\/li>\n<\/ul>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\">1. \uac24\ub7fd \uac15\uc810 \uc9c4\ub2e8 \ub9c1\ud06c\uc5d0 \ub4e4\uc5b4\uac04\ub2e4.<\/p>\n<p data-ke-size=\"size16\">2. \uc624\ub978\ucabd \uc0c1\ub2e8\uc758 'SIGN IN'\uc744 \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">3. '\uacc4\uc815\uc744 \uc0dd\uc131\ud558\uc2ed\uc2dc\uc624'\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">4. \uc815\ubcf4\ub97c \uc785\ub825\ud558\uace0 \uacc4\uc815\uc744 \uc0dd\uc131\ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\">5. \ucd94\uac00 \uc815\ubcf4\ub97c \uc785\ub825\ud558\uace0 \uc124\uc815\uc744 \uc644\ub8cc\ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\">6. \uc774\uba54\uc77c\ub85c \uc804\uc1a1\ub41c \uc778\uc99d \ucf54\ub4dc\ub97c \uc81c\ucd9c\ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\">7. \ub85c\uadf8\uc778\ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\">8. \uc67c\ucabd \uc704 \uba54\ub274\uc5d0\uc11c \uad6c\ub9e4\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">9. \uc624\ub978\ucabd \uc704 South Korea\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">10. \ub098\ub77c\ub97c United States of America\ub85c \ubc14\uafbc\ub2e4.<\/p>\n<p data-ke-size=\"size16\">11. 'Stay in the United States of America store'\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">12. 'LEARN MORE'\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">13. 'ADD TO CART'\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">14. \uc624\ub978\ucabd \uc704 \uce74\ud2b8 \ubaa8\uc591\uc744 \ub204\ub974\uace0 'CHECKOUT'\uc744 \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">15. \uacb0\uc81c \ucc3d\uc5d0\uc11c \uc6d0\ud558\ub294 \uc218\ub2e8\uc744 \uc120\ud0dd\ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\">16. \ud544\uc694\ud55c \ub0b4\uc6a9\uc744 \uc785\ub825\ud558\uace0 'REVIEW ORDER'\uc744 \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">17. \uc8fc\uc18c \ud655\uc778\uc774 \uc548 \ub420 \ub54c - 'USE SELECTED'\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">18. \ucd5c\uc885 \uc815\ubcf4 \ud655\uc778 \ud6c4 'PLACE ORDER'\uc744 \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">19. 'MANAGE MY DIGITAL REPORTS'\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">20. '\ucf54\ub4dc \ubc30\ud3ec'\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">21. \ubc30\ud3ec\ud560 \ubc88\ub4e4\uc744 \uc120\ud0dd\ud558\uace0 '\ub2e4\uc74c'\uc744 \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">22. \ubc30\ud3ec \ubc29\ubc95\uc5d0\uc11c '\uc774\uba54\uc77c \ucf54\ub4dc'\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">23. \uc774\uba54\uc77c \uc8fc\uc18c\ub97c \uc785\ub825\ud558\uace0 \uc5b8\uc5b4\ub97c 'Korean'\uc73c\ub85c \uc120\ud0dd\ud558\uace0 '\uc800\uc7a5'\uc744 \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">24. \uc5b8\uc5b4\ub97c 'Korean'\uc73c\ub85c \uc120\ud0dd\ud558\uace0 '\uc800\uc7a5'\uc744 \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">25. \ub0b4\uc6a9\uc744 \ud655\uc778 \ud6c4 '\uc774\uba54\uc77c \ucf54\ub4dc'\ub97c \ub204\ub978\ub2e4.<\/p>\n<p data-ke-size=\"size16\">26. \ucd08\ub300\uc7a5\uc774 \ubcf4\ub0b4\uc9c4 \uac83\uc744 \ud655\uc778\ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\">27. \ud574\ub2f9 \uba54\uc77c\uc5d0\uc11c '\uc2dc\uc791\ud558\uae30'\ub97c \ub20c\ub7ec \ud14c\uc2a4\ud2b8\ub97c \uc9c4\ud589\ud55c\ub2e4.<\/p>\n<\/div>\n<\/div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud301<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h2 data-ke-size=\"size26\">\uacb0\uacfc \ubbf8\ub9ac\ubcf4\uae30<\/h2>\n<h3 data-ke-size=\"size23\">\ud55c\ud654-\uc6d0\ud654 \uacb0\uc81c \uae08\uc561 \ucc28\uc774 : 8,960\uc6d0<\/h3>\n<p><figure class=\"imagegridblock\">\n  <div class=\"image-container\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/bCKU9m\/btrXIP1c6ww\/Jj0dsLzgiK42peieBzw2Mk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/bCKU9m\/btrXIP1c6ww\/Jj0dsLzgiK42peieBzw2Mk\/img.png\" data-origin-width=\"840\" data-origin-height=\"740\" data-is-animation=\"false\" style=\"width: 48.8444%; margin-right: 10px;\" data-widthpercent=\"49.42\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/bCKU9m\/btrXIP1c6ww\/Jj0dsLzgiK42peieBzw2Mk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FbCKU9m%2FbtrXIP1c6ww%2FJj0dsLzgiK42peieBzw2Mk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"840\" height=\"740\"\/><\/span><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/rOVOt\/btrXJUVgM00\/7hHZ8PY7C8uzXcj9jJSoQk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/rOVOt\/btrXJUVgM00\/7hHZ8PY7C8uzXcj9jJSoQk\/img.png\" data-origin-width=\"840\" data-origin-height=\"723\" data-is-animation=\"false\" style=\"width: 49.9928%;\" data-widthpercent=\"50.58\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/rOVOt\/btrXJUVgM00\/7hHZ8PY7C8uzXcj9jJSoQk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FrOVOt%2FbtrXJUVgM00%2F7hHZ8PY7C8uzXcj9jJSoQk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"840\" height=\"723\"\/><\/span><\/div>\n  <figcaption>\uc88c: \uc6d0\ud654 \uacb0\uc81c, \uc6b0: \ub2ec\ub7ec \uacb0\uc81c. \uc57d 9\ucc9c\uc6d0 \uac00\ub7c9 \ucc28\uc774\ub09c\ub2e4.<\/figcaption>\n<\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\uacb0\uacfc \ubcf4\uace0\uc11c<\/h3>\n<p><figure class=\"imagegridblock\">\n  <div class=\"image-container\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/TXwoZ\/btrXImZgyE7\/Qq4ENEiO6HIDylVZWAAlVK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/TXwoZ\/btrXImZgyE7\/Qq4ENEiO6HIDylVZWAAlVK\/img.png\" data-origin-width=\"1809\" data-origin-height=\"1417\" data-is-animation=\"false\" style=\"width: 53.1053%; margin-right: 10px;\" data-widthpercent=\"53.73\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/TXwoZ\/btrXImZgyE7\/Qq4ENEiO6HIDylVZWAAlVK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FTXwoZ%2FbtrXImZgyE7%2FQq4ENEiO6HIDylVZWAAlVK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1809\" height=\"1417\"\/><\/span><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/SHeds\/btrXJWFzocv\/6xyntd5wZYnaUiFFP77DV0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/SHeds\/btrXJWFzocv\/6xyntd5wZYnaUiFFP77DV0\/img.png\" data-origin-width=\"1792\" data-origin-height=\"1630\" data-is-animation=\"false\" style=\"width: 45.7319%;\" data-widthpercent=\"46.27\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/SHeds\/btrXJWFzocv\/6xyntd5wZYnaUiFFP77DV0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FSHeds%2FbtrXJWFzocv%2F6xyntd5wZYnaUiFFP77DV0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1792\" height=\"1630\"\/><\/span><\/div>\n  <figcaption>\uc88c: \ud14c\uc2a4\ud2b8 \uc9c4\ud589 \ud6c4 \ub098\uc624\ub294 \ud654\uba74 \uc6b0: \ubc1b\uc744 \uc218 \uc788\ub294 \ubcf4\uace0\uc11c<\/figcaption>\n<\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\uc21c\uc704\uac00 \uac19\uc9c0\ub9cc \ub2e4\ub978 \uc2ec\uce35 \uacb0\uacfc<\/h3>\n<p data-ke-size=\"size16\">\uc544\ub798\ub294 INTP 2\uba85\uc774 \ub3d9\uc2dc\uc5d0 \uc9c4\ud589\ud55c \uacb0\uacfc\uc774\ub2e4. \ub458 \ub2e4 \uac1c\ubcc4\ud654 \ud14c\ub9c8\uac00 1\ub4f1\uc73c\ub85c \ub098\uc654\uc9c0\ub9cc, \uacb0\uacfc\uac00 \uc870\uae08\uc529 \ub2e4\ub974\ub2e4. \ud615\uad11\uc0c9\uce60\ud55c \uac83\uc740 \uac01\uc790\ub97c \uc798 \uc124\uba85\ud558\uba74\uc11c, \ub458\uc774 \uc11c\ub85c \uac00\uc7a5 \ud06c\uac8c \ub2e4\ub974\ub2e4\uace0 \ub290\ub080 \uc9c0\uc810\uc774\ub2e4.<\/p>\n<p><figure class=\"imagegridblock\">\n  <div class=\"image-container\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/tVBPo\/btrXFuYzX3A\/dgeE2kAJ1PFZ8GOqiv8l2k\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/tVBPo\/btrXFuYzX3A\/dgeE2kAJ1PFZ8GOqiv8l2k\/img.png\" data-origin-width=\"996\" data-origin-height=\"1303\" data-is-animation=\"false\" style=\"width: 48.8192%; margin-right: 10px;\" data-widthpercent=\"49.39\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/tVBPo\/btrXFuYzX3A\/dgeE2kAJ1PFZ8GOqiv8l2k\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FtVBPo%2FbtrXFuYzX3A%2FdgeE2kAJ1PFZ8GOqiv8l2k%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"996\" height=\"1303\"\/><\/span><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/nF8sz\/btrXIiWU4wD\/u183lHprdQME3R4eVQyQuK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/nF8sz\/btrXIiWU4wD\/u183lHprdQME3R4eVQyQuK\/img.png\" data-origin-width=\"986\" data-origin-height=\"1259\" data-is-animation=\"false\" style=\"width: 50.018%;\" data-widthpercent=\"50.61\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/nF8sz\/btrXIiWU4wD\/u183lHprdQME3R4eVQyQuK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FnF8sz%2FbtrXIiWU4wD%2Fu183lHprdQME3R4eVQyQuK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"986\" height=\"1259\"\/><\/span><\/div>\n<\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h2 data-ke-size=\"size26\">\uacb0\uc81c \uc9c4\ud589 \ubc29\ubc95<\/h2>\n<h3 data-ke-size=\"size23\">1. \uac24\ub7fd \uac15\uc810 \uc9c4\ub2e8 \ub9c1\ud06c\uc5d0 \ub4e4\uc5b4\uac04\ub2e4.<\/h3>\n<p data-ke-size=\"size16\"><a href=\"https:\/\/www.gallup.com\/cliftonstrengths\/en\/home.aspx\" target=\"_blank\" rel=\"noopener\">https:\/\/www.gallup.com\/cliftonstrengths\/en\/home.aspx<\/a><\/p>\n<figure id=\"og_1675090862146\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"CliftonStrengths\" data-og-description=\"Learn how the CliftonStrengths assessment -- formerly StrengthsFinder -- empowers organizations, managers and millions of people to succeed.\" data-og-host=\"www.gallup.com\" data-og-source-url=\"https:\/\/www.gallup.com\/cliftonstrengths\/en\/home.aspx\" data-og-url=\"https:\/\/www.gallup.com\/cliftonstrengths\/en\/252137\/home.aspx\" data-og-image=\"https:\/\/scrap.kakaocdn.net\/dn\/biZjmp\/hyRstaTozl\/cERxdQLL1jKhKjKkCEfDO0\/img.png?width=1600&amp;height=840&amp;face=0_0_1600_840,https:\/\/scrap.kakaocdn.net\/dn\/GSqtR\/hyRspGkL7r\/Qna3kkO62n1Ws2FEZXaYf0\/img.png?width=1600&amp;height=840&amp;face=0_0_1600_840,https:\/\/scrap.kakaocdn.net\/dn\/bqHpzH\/hyRq96bZjL\/ZhHGUgpPfkBmtgtRtFjEk0\/img.png?width=1840&amp;height=920&amp;face=0_0_1840_920\"><a href=\"https:\/\/www.gallup.com\/cliftonstrengths\/en\/home.aspx\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/www.gallup.com\/cliftonstrengths\/en\/home.aspx\">\n<div class=\"og-image\" style=\"background-image: url('https:\/\/scrap.kakaocdn.net\/dn\/biZjmp\/hyRstaTozl\/cERxdQLL1jKhKjKkCEfDO0\/img.png?width=1600&amp;height=840&amp;face=0_0_1600_840,https:\/\/scrap.kakaocdn.net\/dn\/GSqtR\/hyRspGkL7r\/Qna3kkO62n1Ws2FEZXaYf0\/img.png?width=1600&amp;height=840&amp;face=0_0_1600_840,https:\/\/scrap.kakaocdn.net\/dn\/bqHpzH\/hyRq96bZjL\/ZhHGUgpPfkBmtgtRtFjEk0\/img.png?width=1840&amp;height=920&amp;face=0_0_1840_920');\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">CliftonStrengths<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">Learn how the CliftonStrengths assessment -- formerly StrengthsFinder -- empowers organizations, managers and millions of people to succeed.<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">www.gallup.com<\/p>\n<\/div>\n<\/a><\/figure>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">2. \uc624\ub978\ucabd \uc0c1\ub2e8\uc758 'SIGN IN'\uc744 \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1619\" data-origin-height=\"1636\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/lLcAl\/btrXCZbCliJ\/550ovOqWQIJydJhegrRsF0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/lLcAl\/btrXCZbCliJ\/550ovOqWQIJydJhegrRsF0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/lLcAl\/btrXCZbCliJ\/550ovOqWQIJydJhegrRsF0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FlLcAl%2FbtrXCZbCliJ%2F550ovOqWQIJydJhegrRsF0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1619\" height=\"1636\" data-origin-width=\"1619\" data-origin-height=\"1636\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">3. '\uacc4\uc815\uc744 \uc0dd\uc131\ud558\uc2ed\uc2dc\uc624'\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"465\" data-origin-height=\"567\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/tktrg\/btrXyjJAny8\/DKtgKLX3cTkUKqDVNhPRp0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/tktrg\/btrXyjJAny8\/DKtgKLX3cTkUKqDVNhPRp0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/tktrg\/btrXyjJAny8\/DKtgKLX3cTkUKqDVNhPRp0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Ftktrg%2FbtrXyjJAny8%2FDKtgKLX3cTkUKqDVNhPRp0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"465\" height=\"567\" data-origin-width=\"465\" data-origin-height=\"567\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">4. \uc815\ubcf4\ub97c \uc785\ub825\ud558\uace0 \uacc4\uc815\uc744 \uc0dd\uc131\ud55c\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1128\" data-origin-height=\"1175\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/d3jtTM\/btrXBlmp9ce\/Ynq5DPkrAIDIb1FXroxRR0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/d3jtTM\/btrXBlmp9ce\/Ynq5DPkrAIDIb1FXroxRR0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/d3jtTM\/btrXBlmp9ce\/Ynq5DPkrAIDIb1FXroxRR0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fd3jtTM%2FbtrXBlmp9ce%2FYnq5DPkrAIDIb1FXroxRR0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1128\" height=\"1175\" data-origin-width=\"1128\" data-origin-height=\"1175\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">\uc774\ub984, \uc131\uc5d0\ub294 \ud55c\uae00 \uc785\ub825\uc774 \uac00\ub2a5\ud558\ub2e4.<\/p>\n<p data-ke-size=\"size16\">\uc774\uba54\uc77c \uc8fc\uc18c\uc640 \uc544\uc774\ub514\ub97c \ub2e4\ub974\uac8c \uc785\ub825\ud560 \uc218 \uc788\ub294\ub370, \ub85c\uadf8\uc778\ud560 \ub54c \uc544\uc774\ub514\uac00 \uc4f0\uc778\ub2e4. \uae30\uc5b5\ud558\uae30 \uc27d\uac8c \ub3d9\uc77c\ud558\uac8c \ub9de\ucd94\uae30\ub97c \ucd94\ucc9c\ud55c\ub2e4.<\/p>\n<p data-ke-size=\"size16\">\uad6d\uac00\uc640 \uc6b0\ud3b8\ubc88\ud638\ub97c \uc785\ub825\ud574\uc57c \ud558\ub294\ub370, \uc2f8\uac8c \uacb0\uc81c\ud558\ub824\uba74 \ubbf8\uad6d\uc758 \uc54c\ub798\uc2a4\uce74 \uc8fc\uc5d0 \uc788\ub294 \uc8fc\uc18c\ub97c \uc0ac\uc6a9\ud558\uba74 \ub41c\ub2e4. \uc544\ub798 \ub9c1\ud06c\uc5d0\uc11c \uc784\uc758 \uc8fc\uc18c\ub97c \uc5bb\uc744 \uc218 \uc788\ub2e4. \uacb0\uc81c\ud560 \ub54c \ub2e4\uc2dc \uc368\uc57c \ud558\ub2c8 <span>\uc8fc\uc18c\ub97c \uc5b4\ub518\uac00\uc5d0 \uba54\ubaa8\ud574\ub450\uc790.<\/span><\/p>\n<p data-ke-size=\"size16\"><a href=\"https:\/\/ak.postcodebase.com\/ko\/randomaddress\" target=\"_blank\" rel=\"noopener\">https:\/\/ak.postcodebase.com\/ko\/randomaddress<\/a><\/p>\n<figure id=\"og_1675171689689\" contenteditable=\"false\" data-ke-type=\"opengraph\" data-ke-align=\"alignCenter\" data-og-type=\"website\" data-og-title=\"\uc54c\ub798\uc2a4\uce74\uc758 \uc784\uc758 \uc8fc\uc18c\" data-og-description=\"\uc774 \uc784\uc758 \uc8fc\uc18c \ub3c4\uad6c\ub294 \ubbf8\uad6d\uc5d0\uc11c \uac70\ub9ac, \ub3c4\uc2dc, \uc8fc, \uc6b0\ud3b8 \ubc88\ud638 5 \ubc0f \uc6b0\ud3b8 \ubc88\ud638 5 + 4\ub97c \ud3ec\ud568\ud55c \uc784\uc758 \uc8fc\uc18c\ub97c \uc81c\uacf5\ud569\ub2c8\ub2e4. \ub610\ud55c\uc774 \ubc84\ud2bc\uc744 \ud074\ub9ad\ud558\uc5ec \uc0c8 \uc784\uc758 \uc8fc\uc18c\ub97c \uc0dd\uc131 \ud560 \uc218\ub3c4 \uc788\uc2b5\ub2c8\ub2e4.\" data-og-host=\"ak.postcodebase.com\" data-og-source-url=\"https:\/\/ak.postcodebase.com\/ko\/randomaddress\" data-og-url=\"https:\/\/ak.postcodebase.com\/ko\/randomaddress\" data-og-image=\"https:\/\/scrap.kakaocdn.net\/dn\/bRc6qc\/hyRsvGI9Xu\/Alak1Qi4x0LJ3llfxmV87k\/img.png?width=1200&amp;height=675&amp;face=0_0_1200_675,https:\/\/scrap.kakaocdn.net\/dn\/lrAwV\/hyRsqZHQBL\/A85tjPKD8jyKk05VcRXen0\/img.png?width=1200&amp;height=675&amp;face=0_0_1200_675,https:\/\/scrap.kakaocdn.net\/dn\/joQNf\/hyRst9YJwT\/tvrgIK6qb3zbUotQ48o7v0\/img.png?width=640&amp;height=320&amp;face=0_0_640_320\"><a href=\"https:\/\/ak.postcodebase.com\/ko\/randomaddress\" target=\"_blank\" rel=\"noopener\" data-source-url=\"https:\/\/ak.postcodebase.com\/ko\/randomaddress\">\n<div class=\"og-image\" style=\"background-image: url('https:\/\/scrap.kakaocdn.net\/dn\/bRc6qc\/hyRsvGI9Xu\/Alak1Qi4x0LJ3llfxmV87k\/img.png?width=1200&amp;height=675&amp;face=0_0_1200_675,https:\/\/scrap.kakaocdn.net\/dn\/lrAwV\/hyRsqZHQBL\/A85tjPKD8jyKk05VcRXen0\/img.png?width=1200&amp;height=675&amp;face=0_0_1200_675,https:\/\/scrap.kakaocdn.net\/dn\/joQNf\/hyRst9YJwT\/tvrgIK6qb3zbUotQ48o7v0\/img.png?width=640&amp;height=320&amp;face=0_0_640_320');\">&nbsp;<\/div>\n<div class=\"og-text\">\n<p class=\"og-title\" data-ke-size=\"size16\">\uc54c\ub798\uc2a4\uce74\uc758 \uc784\uc758 \uc8fc\uc18c<\/p>\n<p class=\"og-desc\" data-ke-size=\"size16\">\uc774 \uc784\uc758 \uc8fc\uc18c \ub3c4\uad6c\ub294 \ubbf8\uad6d\uc5d0\uc11c \uac70\ub9ac, \ub3c4\uc2dc, \uc8fc, \uc6b0\ud3b8 \ubc88\ud638 5 \ubc0f \uc6b0\ud3b8 \ubc88\ud638 5 + 4\ub97c \ud3ec\ud568\ud55c \uc784\uc758 \uc8fc\uc18c\ub97c \uc81c\uacf5\ud569\ub2c8\ub2e4. \ub610\ud55c\uc774 \ubc84\ud2bc\uc744 \ud074\ub9ad\ud558\uc5ec \uc0c8 \uc784\uc758 \uc8fc\uc18c\ub97c \uc0dd\uc131 \ud560 \uc218\ub3c4 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n<p class=\"og-host\" data-ke-size=\"size16\">ak.postcodebase.com<\/p>\n<\/div>\n<\/a><\/figure>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">5. \ucd94\uac00 \uc815\ubcf4\ub97c \uc785\ub825\ud558\uace0 \uc124\uc815\uc744 \uc644\ub8cc\ud55c\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"951\" data-origin-height=\"1089\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/bca7lx\/btrXDg5pKVF\/lLdGjbOrXPTzvQEqJFjGvk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/bca7lx\/btrXDg5pKVF\/lLdGjbOrXPTzvQEqJFjGvk\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/bca7lx\/btrXDg5pKVF\/lLdGjbOrXPTzvQEqJFjGvk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fbca7lx%2FbtrXDg5pKVF%2FlLdGjbOrXPTzvQEqJFjGvk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"951\" height=\"1089\" data-origin-width=\"951\" data-origin-height=\"1089\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">\uc870\uc9c1\uc5d0\ub294 \uc790\ub3d9\uc644\uc131\uc5d0 \ub098\uc640 \uc788\uc9c0 \uc54a\uc740 \uac83\uc744 \uc785\ub825\ud558\uc5ec\ub3c4 \ub118\uc5b4\uac00\uc9c4\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">6. \uc774\uba54\uc77c\ub85c \uc804\uc1a1\ub41c \uc778\uc99d \ucf54\ub4dc\ub97c \uc81c\ucd9c\ud55c\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"780\" data-origin-height=\"714\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/buSeDv\/btrXIO810A2\/FKeZkTjsvxWtvrGA89bpN0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/buSeDv\/btrXIO810A2\/FKeZkTjsvxWtvrGA89bpN0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/buSeDv\/btrXIO810A2\/FKeZkTjsvxWtvrGA89bpN0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FbuSeDv%2FbtrXIO810A2%2FFKeZkTjsvxWtvrGA89bpN0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"780\" height=\"714\" data-origin-width=\"780\" data-origin-height=\"714\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">7. \ub85c\uadf8\uc778\ud55c\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1620\" data-origin-height=\"791\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/cRCeBB\/btrXEBafMYV\/W2WPd76HukSS6GpkpUrYek\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/cRCeBB\/btrXEBafMYV\/W2WPd76HukSS6GpkpUrYek\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/cRCeBB\/btrXEBafMYV\/W2WPd76HukSS6GpkpUrYek\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FcRCeBB%2FbtrXEBafMYV%2FW2WPd76HukSS6GpkpUrYek%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1620\" height=\"791\" data-origin-width=\"1620\" data-origin-height=\"791\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">8. \uc67c\ucabd \uc704 \uba54\ub274\uc5d0\uc11c \uad6c\ub9e4\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1920\" data-origin-height=\"561\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/ylYIO\/btrXBMKHWNY\/ogxa5ivz8purvyjPKXVcW0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/ylYIO\/btrXBMKHWNY\/ogxa5ivz8purvyjPKXVcW0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/ylYIO\/btrXBMKHWNY\/ogxa5ivz8purvyjPKXVcW0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FylYIO%2FbtrXBMKHWNY%2Fogxa5ivz8purvyjPKXVcW0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1920\" height=\"561\" data-origin-width=\"1920\" data-origin-height=\"561\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">9. \uc624\ub978\ucabd \uc704 South Korea\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1281\" data-origin-height=\"868\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/GPJWZ\/btrXEC1iHlT\/sfuuWrc2dMIs7TvufvS8lK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/GPJWZ\/btrXEC1iHlT\/sfuuWrc2dMIs7TvufvS8lK\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/GPJWZ\/btrXEC1iHlT\/sfuuWrc2dMIs7TvufvS8lK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FGPJWZ%2FbtrXEC1iHlT%2FsfuuWrc2dMIs7TvufvS8lK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1281\" height=\"868\" data-origin-width=\"1281\" data-origin-height=\"868\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">10. \ub098\ub77c\ub97c United States of America\ub85c \ubc14\uafbc\ub2e4.<\/h3>\n<p data-ke-size=\"size16\">united \uc785\ub825 \ud6c4 \uc120\ud0dd\ud558\uc790.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1278\" data-origin-height=\"633\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/r4wHo\/btrXCio433T\/5XwBaL69yjbY0KeAZrhaMk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/r4wHo\/btrXCio433T\/5XwBaL69yjbY0KeAZrhaMk\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/r4wHo\/btrXCio433T\/5XwBaL69yjbY0KeAZrhaMk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fr4wHo%2FbtrXCio433T%2F5XwBaL69yjbY0KeAZrhaMk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1278\" height=\"633\" data-origin-width=\"1278\" data-origin-height=\"633\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">11. 'Stay in the United States of America store'\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"681\" data-origin-height=\"563\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/Ehp3s\/btrXBk8VYEE\/Zzw8ckxkkvKuCrmYvjg3k1\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/Ehp3s\/btrXBk8VYEE\/Zzw8ckxkkvKuCrmYvjg3k1\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/Ehp3s\/btrXBk8VYEE\/Zzw8ckxkkvKuCrmYvjg3k1\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FEhp3s%2FbtrXBk8VYEE%2FZzw8ckxkkvKuCrmYvjg3k1%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"681\" height=\"563\" data-origin-width=\"681\" data-origin-height=\"563\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">12. 'LEARN MORE'\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1261\" data-origin-height=\"877\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/MA4le\/btrXwiRwdIU\/1JEvgL4GHAjy8lPVKk2qg0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/MA4le\/btrXwiRwdIU\/1JEvgL4GHAjy8lPVKk2qg0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/MA4le\/btrXwiRwdIU\/1JEvgL4GHAjy8lPVKk2qg0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FMA4le%2FbtrXwiRwdIU%2F1JEvgL4GHAjy8lPVKk2qg0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1261\" height=\"877\" data-origin-width=\"1261\" data-origin-height=\"877\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">13. 'ADD TO CART'\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p data-ke-size=\"size16\">\ub2ec\ub7ec\ub85c \uacb0\uc81c\ud574\ub3c4 \ud55c\uae00\ub85c \ud14c\uc2a4\ud2b8\ub97c \uc9c4\ud589\ud560 \uc218 \uc788\uace0, \uacb0\uacfc\ub3c4 \ud55c\uae00\ub85c \ubcfc \uc218 \uc788\ub2e4.<\/p>\n<p data-ke-size=\"size16\">\ud55c\ud654\ub85c\ub294 82,500\uc6d0\uc778 \uac83\uc774 \ub2ec\ub7ec\ub85c\ub294 59.99$\uc774\ub2e4. \ud55c\ud654\uac00 \uc138\uae08\uc774 \ud3ec\ud568\ub418\uc5b4 \uc788\uc9c0\ub9cc, \uc138\uae08\uc744 \uc801\uc6a9\ud574\ub3c4 \uc57d 1\ub9cc\uc6d0 \uac00\ub7c9 \ub354 \uc2f8\ub2e4.<\/p>\n<p><figure class=\"imagegridblock\">\n  <div class=\"image-container\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/dB1lID\/btrXEBafX3V\/VVqOcx52NmxYcH14qc9rZk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/dB1lID\/btrXEBafX3V\/VVqOcx52NmxYcH14qc9rZk\/img.png\" data-origin-width=\"991\" data-origin-height=\"474\" data-filename=\"blob\" data-is-animation=\"false\" style=\"width: 47.5791%; margin-right: 10px;\" data-widthpercent=\"48.14\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/dB1lID\/btrXEBafX3V\/VVqOcx52NmxYcH14qc9rZk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FdB1lID%2FbtrXEBafX3V%2FVVqOcx52NmxYcH14qc9rZk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"991\" height=\"474\"\/><\/span><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/Sm4II\/btrXAjboXm8\/vxCH5ZwwILnbmSg6NkkmEK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/Sm4II\/btrXAjboXm8\/vxCH5ZwwILnbmSg6NkkmEK\/img.png\" data-origin-width=\"946\" data-origin-height=\"420\" data-is-animation=\"false\" style=\"width: 51.2581%;\" data-widthpercent=\"51.86\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/Sm4II\/btrXAjboXm8\/vxCH5ZwwILnbmSg6NkkmEK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&amp;fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FSm4II%2FbtrXAjboXm8%2FvxCH5ZwwILnbmSg6NkkmEK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"946\" height=\"420\"\/><\/span><\/div>\n  <figcaption>\uc88c: \ub2ec\ub7ec \uac00\uaca9, \uc6b0: \uc6d0\ud654 \uac00\uaca9<\/figcaption>\n<\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">14. \uc624\ub978\ucabd \uc704 \uce74\ud2b8 \ubaa8\uc591\uc744 \ub204\ub974\uace0 'CHECKOUT'\uc744 \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"419\" data-origin-height=\"329\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/dPY029\/btrXBAKrPLW\/fTYP7ZJKnbyGKp7aoGHew1\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/dPY029\/btrXBAKrPLW\/fTYP7ZJKnbyGKp7aoGHew1\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/dPY029\/btrXBAKrPLW\/fTYP7ZJKnbyGKp7aoGHew1\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FdPY029%2FbtrXBAKrPLW%2FfTYP7ZJKnbyGKp7aoGHew1%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"419\" height=\"329\" data-origin-width=\"419\" data-origin-height=\"329\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">15. \uacb0\uc81c \ucc3d\uc5d0\uc11c \uc6d0\ud558\ub294 \uc218\ub2e8\uc744 \uc120\ud0dd\ud55c\ub2e4.<\/h3>\n<p data-ke-size=\"size16\">\ubcf8\uc778\uc740 PayPal \ub300\uc2e0 \uce74\ub4dc\ubc88\ud638\ub97c \uc9c1\uc811 \uc785\ub825\ud558\uc5ec \uacb0\uc81c\ud558\ub294 'CHECKOUT'\uc744 \uc120\ud0dd\ud588\ub2e4.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1080\" data-origin-height=\"617\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/4avEU\/btrXEBnNqrC\/lw7ApeO2sOCAfjhGQOPvmK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/4avEU\/btrXEBnNqrC\/lw7ApeO2sOCAfjhGQOPvmK\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/4avEU\/btrXEBnNqrC\/lw7ApeO2sOCAfjhGQOPvmK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2F4avEU%2FbtrXEBnNqrC%2Flw7ApeO2sOCAfjhGQOPvmK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1080\" height=\"617\" data-origin-width=\"1080\" data-origin-height=\"617\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">16. \ud544\uc694\ud55c \ub0b4\uc6a9\uc744 \uc785\ub825\ud558\uace0 'REVIEW ORDER'\uc744 \ub204\ub978\ub2e4.<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc774\ub984, \uc131: \uc0ac\uc6a9\ud560 \uce74\ub4dc\uc640 \ub3d9\uc77c\ud558\uac8c \uc601\uc5b4 \uc774\ub984\uc73c\ub85c \uc785\ub825\ud558\uc600\ub2e4(\ud55c\uae00 \uc785\ub825 \uac00\ub2a5 \uc5ec\ubd80\ub294 \ubaa8\ub984).<\/li>\n<li>\uc8fc\uc18c: \uc704 \uc784\uc758 \uc8fc\uc18c\uc5d0\uc11c \uc5bb\uc740 \uc8fc\uc18c\ub97c \uc785\ub825\ud558\uc600\ub2e4.<\/li>\n<li>\ud578\ub4dc\ud3f0 \ubc88\ud638: \ubcf8\uc778 \uac83\uc744 \uc785\ub825\ud558\uc600\ub294\ub370, \uc6d0\ud654\ub85c \uacb0\uc81c\ud588\uc5b4\ub3c4 \uc544\ubb34\ub7f0 \uc5f0\ub77d\uc774 \uc5c6\uc5c8\ub2e4. \uc815\ubcf4\ub294 \ubaa8\ub450 \uc774\uba54\uc77c\uc744 \ud1b5\ud574\uc11c \uc804\ub2ec\ub418\uc5c8\ub2e4.<\/li>\n<\/ul>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1119\" data-origin-height=\"1775\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/MObKD\/btrXA9l99k5\/VkcpDKV4486kEeSNKcd1rk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/MObKD\/btrXA9l99k5\/VkcpDKV4486kEeSNKcd1rk\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/MObKD\/btrXA9l99k5\/VkcpDKV4486kEeSNKcd1rk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FMObKD%2FbtrXA9l99k5%2FVkcpDKV4486kEeSNKcd1rk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1119\" height=\"1775\" data-origin-width=\"1119\" data-origin-height=\"1775\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">17. \uc8fc\uc18c \ud655\uc778\uc774 \uc548 \ub420 \ub54c - 'USE SELECTED'\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p data-ke-size=\"size16\">\ud55c\uad6d \uc8fc\uc18c\ub85c \ud588\uc744 \ub54c\ub294 \ud574\ub2f9 \uba54\uc2dc\uc9c0\uac00 \ub098\uc624\uc9c0 \uc54a\uc558\ub2e4.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"610\" data-origin-height=\"436\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/bAP989\/btrXCZo9B6k\/oFhiUqwygVjTDOjqTKqJ2K\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/bAP989\/btrXCZo9B6k\/oFhiUqwygVjTDOjqTKqJ2K\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/bAP989\/btrXCZo9B6k\/oFhiUqwygVjTDOjqTKqJ2K\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FbAP989%2FbtrXCZo9B6k%2FoFhiUqwygVjTDOjqTKqJ2K%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"610\" height=\"436\" data-origin-width=\"610\" data-origin-height=\"436\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">18. \ucd5c\uc885 \uc815\ubcf4 \ud655\uc778 \ud6c4 'PLACE ORDER'\uc744 \ub204\ub978\ub2e4.<\/h3>\n<p data-ke-size=\"size16\">\uc138\uae08\uc774 0\uc6d0\uc73c\ub85c \ucc45\uc815\ub41c \uac83\uc744 \ud655\uc778\ud560 \uc218 \uc788\ub2e4.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1090\" data-origin-height=\"829\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/tDlOR\/btrXDgRRB4P\/ArpfVUBb9dj4CQfylKSvaK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/tDlOR\/btrXDgRRB4P\/ArpfVUBb9dj4CQfylKSvaK\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/tDlOR\/btrXDgRRB4P\/ArpfVUBb9dj4CQfylKSvaK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FtDlOR%2FbtrXDgRRB4P%2FArpfVUBb9dj4CQfylKSvaK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1090\" height=\"829\" data-origin-width=\"1090\" data-origin-height=\"829\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">19. 'MANAGE MY DIGITAL REPORTS'\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1058\" data-origin-height=\"676\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/qCWgb\/btrXIy6kpP3\/YbgujV9FB47xHNPHLGVsF0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/qCWgb\/btrXIy6kpP3\/YbgujV9FB47xHNPHLGVsF0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/qCWgb\/btrXIy6kpP3\/YbgujV9FB47xHNPHLGVsF0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FqCWgb%2FbtrXIy6kpP3%2FYbgujV9FB47xHNPHLGVsF0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1058\" height=\"676\" data-origin-width=\"1058\" data-origin-height=\"676\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">20. '\ucf54\ub4dc \ubc30\ud3ec'\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1338\" data-origin-height=\"996\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/Sdiie\/btrXIQsihyq\/q0fAwZ0wiGKU2YL2qqB6E0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/Sdiie\/btrXIQsihyq\/q0fAwZ0wiGKU2YL2qqB6E0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/Sdiie\/btrXIQsihyq\/q0fAwZ0wiGKU2YL2qqB6E0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FSdiie%2FbtrXIQsihyq%2Fq0fAwZ0wiGKU2YL2qqB6E0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1338\" height=\"996\" data-origin-width=\"1338\" data-origin-height=\"996\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">21. \ubc30\ud3ec\ud560 \ubc88\ub4e4\uc744 \uc120\ud0dd\ud558\uace0 '\ub2e4\uc74c'\uc744 \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"764\" data-origin-height=\"434\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/cpcE0Y\/btrXAkBk4e9\/OKps1aw3LxjjWHloZCnrr0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/cpcE0Y\/btrXAkBk4e9\/OKps1aw3LxjjWHloZCnrr0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/cpcE0Y\/btrXAkBk4e9\/OKps1aw3LxjjWHloZCnrr0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FcpcE0Y%2FbtrXAkBk4e9%2FOKps1aw3LxjjWHloZCnrr0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"764\" height=\"434\" data-origin-width=\"764\" data-origin-height=\"434\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">22. \ubc30\ud3ec \ubc29\ubc95\uc5d0\uc11c '\uc774\uba54\uc77c \ucf54\ub4dc'\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p data-ke-size=\"size16\">\uc774\uba54\uc77c \ucf54\ub4dc\ub97c \ub204\ub974\uba74, \ud574\ub2f9 \uc774\uba54\uc77c\ub85c \ud14c\uc2a4\ud2b8 \uc9c4\ud589 \ucd08\uccad \uba54\uc77c\uc774 \ubcf4\ub0b4\uc9c4\ub2e4.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"735\" data-origin-height=\"487\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/cV05Au\/btrXBk19Sfc\/kLRfAvPzlnYWghhRlUFR3k\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/cV05Au\/btrXBk19Sfc\/kLRfAvPzlnYWghhRlUFR3k\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/cV05Au\/btrXBk19Sfc\/kLRfAvPzlnYWghhRlUFR3k\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FcV05Au%2FbtrXBk19Sfc%2FkLRfAvPzlnYWghhRlUFR3k%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"735\" height=\"487\" data-origin-width=\"735\" data-origin-height=\"487\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">23. \uc774\uba54\uc77c \uc8fc\uc18c\ub97c \uc785\ub825\ud558\uace0 \uc5b8\uc5b4\ub97c 'Korean'\uc73c\ub85c \uc120\ud0dd\ud558\uace0 '\uc800\uc7a5'\uc744 \ub204\ub978\ub2e4.<\/h3>\n<p data-ke-size=\"size16\">\uc800\uc7a5\uc744 \ub204\ub974\uba74 \uc544\ub798\uc5d0 \uc218\uc2e0\uc790\uac00 \ucd94\uac00\ub41c \uac83\uc744 \ud655\uc778\ud560 \uc218 \uc788\ub2e4. \ubcf8\uc778\uc740 \ud0c0\uc778\uc5d0\uac8c\ub3c4 \ubcf4\ub0b4\ubcf4\uc558\uace0, \ud14c\uc2a4\ud2b8\ub97c \uc798 \uc9c4\ud589\ud560 \uc218 \uc788\ub294 \uac83\uc744 \ud655\uc778\ud558\uc600\ub2e4.<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"928\" data-origin-height=\"711\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/yjnbJ\/btrXChw30VX\/vN2GBxT7LaG22OqGvu2kGk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/yjnbJ\/btrXChw30VX\/vN2GBxT7LaG22OqGvu2kGk\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/yjnbJ\/btrXChw30VX\/vN2GBxT7LaG22OqGvu2kGk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FyjnbJ%2FbtrXChw30VX%2FvN2GBxT7LaG22OqGvu2kGk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"928\" height=\"711\" data-origin-width=\"928\" data-origin-height=\"711\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">24. \uc5b8\uc5b4\ub97c 'Korean'\uc73c\ub85c \uc120\ud0dd\ud558\uace0 '\uc800\uc7a5'\uc744 \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"928\" data-origin-height=\"1351\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/3Y9Es\/btrXD7NUSb6\/1sOR5Bhd2JHWfETyBKrkI1\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/3Y9Es\/btrXD7NUSb6\/1sOR5Bhd2JHWfETyBKrkI1\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/3Y9Es\/btrXD7NUSb6\/1sOR5Bhd2JHWfETyBKrkI1\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2F3Y9Es%2FbtrXD7NUSb6%2F1sOR5Bhd2JHWfETyBKrkI1%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"928\" height=\"1351\" data-origin-width=\"928\" data-origin-height=\"1351\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">25. \ub0b4\uc6a9\uc744 \ud655\uc778 \ud6c4 '\uc774\uba54\uc77c \ucf54\ub4dc'\ub97c \ub204\ub978\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"909\" data-origin-height=\"405\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/VR29O\/btrXD6OZ4u2\/Vw8pL50AWbj9zeakc0uea0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/VR29O\/btrXD6OZ4u2\/Vw8pL50AWbj9zeakc0uea0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/VR29O\/btrXD6OZ4u2\/Vw8pL50AWbj9zeakc0uea0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FVR29O%2FbtrXD6OZ4u2%2FVw8pL50AWbj9zeakc0uea0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"909\" height=\"405\" data-origin-width=\"909\" data-origin-height=\"405\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">26. \ucd08\ub300\uc7a5\uc774 \ubcf4\ub0b4\uc9c4 \uac83\uc744 \ud655\uc778\ud55c\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"665\" data-origin-height=\"226\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/b8E8z7\/btrXBMxeKbl\/vyKo5kdxEhycp6lI6Ya00k\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/b8E8z7\/btrXBMxeKbl\/vyKo5kdxEhycp6lI6Ya00k\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/b8E8z7\/btrXBMxeKbl\/vyKo5kdxEhycp6lI6Ya00k\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fb8E8z7%2FbtrXBMxeKbl%2FvyKo5kdxEhycp6lI6Ya00k%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"665\" height=\"226\" data-origin-width=\"665\" data-origin-height=\"226\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">27. \ud574\ub2f9 \uba54\uc77c\uc5d0\uc11c '\uc2dc\uc791\ud558\uae30'\ub97c \ub20c\ub7ec \ud14c\uc2a4\ud2b8\ub97c \uc9c4\ud589\ud55c\ub2e4.<\/h3>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1036\" data-origin-height=\"1144\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/b46lY0\/btrXJgEhR4w\/XHxJIlrYsqdsJwBXuTRKxK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/b46lY0\/btrXJgEhR4w\/XHxJIlrYsqdsJwBXuTRKxK\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/b46lY0\/btrXJgEhR4w\/XHxJIlrYsqdsJwBXuTRKxK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fb46lY0%2FbtrXJgEhR4w%2FXHxJIlrYsqdsJwBXuTRKxK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1036\" height=\"1144\" data-origin-width=\"1036\" data-origin-height=\"1144\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h2 data-ke-size=\"size26\">\ud301<\/h2>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc704 \uba54\uc77c\uc5d0\uc11c '\uc2dc\uc791\ud558\uae30'\ub97c \ub204\ub974\uba74 \uc790\ub3d9\uc73c\ub85c \ub9c1\ud06c\uac00 \uc5f0\uacb0\ub41c\ub2e4. \uc544\uc27d\uac8c\ub3c4 \ud574\ub2f9 \uc774\ud6c4 \uacfc\uc815\uc740 \ucea1\ucc98\ud558\uc9c0 \ubabb\ud588\uc73c\ub098, \uc601\uc5b4\ub85c \ub41c \uac83\uc744 \ud55c\uad6d\uc5b4\ub85c \uc124\uc815\ud558\uace0 \ud14c\uc2a4\ud2b8\ub97c \uc9c4\ud589\ud558\uba74 \ub41c\ub2e4. \ud06c\uac8c \uc5b4\ub824\uc6c0\uc740 \uc5c6\uc744 \uac83\uc774\ub2e4.<\/li>\n<li>\ud68c\uc6d0\uac00\uc785\uc744 \uc774\ubbf8 \ud55c \uc0ac\ub78c\uc740 \uac00\uc785\ud560 \ub54c \uc4f4 \uc544\uc774\ub514\ub97c \uc0ac\uc6a9\ud558\uc5ec \ub85c\uadf8\uc778\ud558\uba74 \ub41c\ub2e4. \uc544\ub2cc \uc0ac\ub78c\uc740 \uc0c8\ub85c \uac00\uc785\ud558\uc5ec\uc57c \ud55c\ub2e4.<\/li>\n<li>\ud14c\uc2a4\ud2b8\ub294 \ud558\ub098\uc758 \uc774\uba54\uc77c\uc5d0 \ud55c \ubc88\ub9cc \ud560 \uc218 \uc788\ub3c4\ub85d \ub418\uc5b4 \uc788\ub2e4. \uc0c8\ub85c\uc6b4 \ud14c\uc2a4\ud2b8\ub97c \uc9c4\ud589\ud558\uace0 \uc2f6\uc73c\uba74 \uc0c8\ub85c\uc6b4 \uacc4\uc815\uc744 \uc0dd\uc131\ud558\uc5ec \uc9c4\ud589\ud558\uc790.<\/li>\n<li>\ud14c\uc2a4\ud2b8\ub294 177\ubb38\ud56d\uc774\uba70 20\ucd08\uc758 \uc81c\ud55c\uc2dc\uac04\uc774 \uc9c0\ub098\uac00\uba74 \uc751\ub2f5\uc774 \uc790\ub3d9\uc73c\ub85c \uae30\ub85d\ub418\uace0 \ub2e4\uc74c \ubb38\uc81c\ub85c \ub118\uc5b4\uac04\ub2e4\ub294 \uac83\uc744 \uc8fc\uc758\ud558\uc790.<\/li>\n<li>\ud14c\uc2a4\ud2b8\ub97c \uc9c4\ud589\ud560 \ub54c, \uc790\uc2e0\uc774 \ub418\uace0\uc790 \ud558\ub294 \uc0ac\ub78c\uc774\ub77c\uace0 \uc0dd\uac01\ud558\ub294 \uac83\ubcf4\ub2e4, \uc9c0\ub09c 1\ub144\uc744 \ub3cc\uc544\ubcf4\uc558\uc744 \ub54c, \uc790\uc2e0\uc774 \uc5b4\ub5bb\uac8c \ud589\ub3d9\ud588\ub294\uc9c0\ub97c \uadfc\uac70\ub85c \uc120\ud0dd\ud558\uba74 \uc870\uae08 \ub354 \uc26c\uc6b8 \uac83\uc774\ub2e4.<\/li>\n<li>\uc2e4\ud5d8 \uc911\uac04\uc5d0 \ub4a4\ub85c\uac00\uae30 \ub4f1\uc744 \ub204\ub974\uac70\ub098 \ucc3d\uc774 \uaebc\uc9c0\uba74 \ub3c8\uc774 \uadf8\ub300\ub85c \ub0a0\ub77c\uac11\ub2c8\ub2e4.<\/li>\n<li>\ubc18\ub4dc\uc2dc \ubc29\ud574 \ubc1b\uc9c0 \uc54a\ub294 \ud658\uacbd\uc5d0\uc11c \ub9c8\uc74c \ud3b8\ud558\uac8c \ubcf4\ub294 \uac83\uc744 \ucd94\ucc9c\ud569\ub2c8\ub2e4.<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<p data-ke-size=\"size16\">\uc774\ub807\uac8c \uac24\ub7fd\uc758 \ud074\ub9ac\ud504\ud134 \uc2a4\ud2b8\ub81d\uc2a4\ub97c \uc9c4\ud589\ud574\ubcf4\uc558\ub2e4. \uba87 \ub144 \uc804\ubd80\ud130 \uc9c0\uae08\uae4c\uc9c0 \uacc4\uc18d MBTI\uac00 \uc720\ud589\uc744 \ud0c0\uace0 \uc788\uc9c0\ub9cc, \uc6b0\ub9ac \ubaa8\ub450 \uacbd\ud5d8\uc801\uc73c\ub85c \uc2e0\ub8b0\ub3c4\uac00 \ubd80\uc871\ud558\ub2e4\ub294 \uac83\uc744 \uc54c \uc218 \uc788\ub2e4. \uc624\ud788\ub824 \ud544\uc790\ub294 \ud574\ub2f9 \uac80\uc0ac \uacb0\uacfc\uc5d0 \uc790\uc2e0\uc744 \ub9de\ucd94\ub824\ub294 \uc0ac\ub78c\ub4e4\uc744 \ub9ce\uc774 \ubcf4\uc558\ub2e4. \ubc18\uba74, \uac24\ub7fd \uac15\uc810 \uc9c4\ub2e8\uc740 \ube44\uc2f8\ub2e4\uace0 \uc0dd\uac01\ud560 \uc218 \uc788\uc9c0\ub9cc \ucda9\ubd84\ud788 \uc790\uc2e0\uc744 \uac1d\uad00\uc801\uc73c\ub85c \ud310\ub2e8\ud574\uc11c \ub9de\ucda4\ud615\uc73c\ub85c \ubd84\uc11d\ud574\uc904 \uc218 \uc788\ub294 \ud14c\uc2a4\ud2b8\ub77c\uace0 \uc0dd\uac01\uc774 \ub41c\ub2e4. \ub2e4\uc74c \uae00\uc5d0\ub294 \uac15\uc810\uc774 \ub2e8\uc810\uc774 \ub418\ub294 \uacbd\uc6b0, \ud074\ub9ac\ud504\ud134 \uc2a4\ud2b8\ub81d\uc2a4\uc5d0\uc11c \ubc14\ub77c\ubcf4\ub294 \uac15\uc810\uacfc \uc57d\uc810\uc758 \uc758\ubbf8, \uac80\uc0ac \uacb0\uacfc\uc9c0 \ubd84\uc11d \ubc29\ubc95\uc5d0 \ub300\ud574 \uc124\uba85\ud560 \uac83\uc774\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\uc774 \uae00\uc774 \ub3c4\uc6c0\uc774 \ub418\uc5c8\uae30\ub97c \ubc14\ub78d\ub2c8\ub2e4.<\/p>","category":["\uae30\ud0c0","cliftonstrengths","gallup","strengthsfinder","\uac24\ub7fd","\uac24\ub7fd \uac15\uc810","\uac24\ub7fd \uac15\uc810 \ubb34\ub8cc","\uac24\ub7fd \uac15\uc810 \uc2f8\uac8c","\uac24\ub7fd \uac15\uc810 \uc9c4\ub2e8","\uc2a4\ud2b8\ub81d\uc2a4\ud30c\uc778\ub354","\ud074\ub9ac\ud504\ud134\uc2a4\ud2b8\ub81d\uc2a4"],"author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/136","comments":"https:\/\/hoohaha.tistory.com\/136#entry136comment","pubDate":"Wed, 1 Feb 2023 00:18:28 +0900"},{"title":"[\ud074\ub77c\uc774\ubc0d] \ud074\ub77c\uc774\ubc0d \uc804 \ud544\uc694\ud55c \uc2a4\ud2b8\ub808\uce6d \uc815\ub9ac","link":"https:\/\/hoohaha.tistory.com\/135","description":"<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><b>\ubaa9\ucc28<\/b><\/p>\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"decimal\">\n<li>\uad00\uc808 \ub3cc\ub9ac\uae30<\/li>\n<li>\uc0c1\uccb4 \uc2a4\ud2b8\ub808\uce6d<\/li>\n<li>\ud558\uccb4 \uc2a4\ud2b8\ub808\uce6d<\/li>\n<li>\ud3fc\ub864\ub7ec \ub9c8\uc0ac\uc9c0<\/li>\n<li>\ub9c8\ubb34\ub9ac \uc2a4\ud2b8\ub808\uce6d \ubc0f \ubab8\ud480\uae30 \uc6b4\ub3d9<\/li>\n<\/ol>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h3 data-ke-size=\"size23\">\uad00\uc808 \ub3cc\ub9ac\uae30 - \ubc1c\ubaa9\uc5d0\uc11c \ubaa9\uae4c\uc9c0(<a href=\"https:\/\/m.blog.naver.com\/PostView.naver?isHttpsRedirect=true&amp;blogId=giftsmile4u&amp;logNo=220478891590\" target=\"_blank\" rel=\"noopener\">\ub9c1\ud06c<\/a>)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc190\ubaa9 \ubc1c\ubaa9 \ub3cc\ub9ac\uae30<\/li>\n<li>\ubb34\ub98e \ub3cc\ub9ac\uae30<\/li>\n<li>\ud5c8\ub9ac \ub3cc\ub9ac\uae30<\/li>\n<li>\uc5b4\uae68 \ub3cc\ub9ac\uae30<\/li>\n<li>\ubaa9 \ub3cc\ub9ac\uae30<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"592\" data-origin-height=\"389\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/bfzaG7\/btrXo5X0FKJ\/53P2D1PyjjLXTvN8y0w4nk\/img.webp\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/bfzaG7\/btrXo5X0FKJ\/53P2D1PyjjLXTvN8y0w4nk\/img.webp\" data-alt=\"\ucd9c\ucc98:&amp;nbsp;https:\/\/m.blog.naver.com\/glddldi2\/220531338777\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/bfzaG7\/btrXo5X0FKJ\/53P2D1PyjjLXTvN8y0w4nk\/img.webp\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FbfzaG7%2FbtrXo5X0FKJ%2F53P2D1PyjjLXTvN8y0w4nk%2Fimg.webp\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"592\" height=\"389\" data-origin-width=\"592\" data-origin-height=\"389\"\/><\/span><figcaption>\ucd9c\ucc98:&nbsp;https:\/\/m.blog.naver.com\/glddldi2\/220531338777<\/figcaption>\n<\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h3 data-ke-size=\"size23\">\uc0c1\uccb4 \uc2a4\ud2b8\ub808\uce6d(<a href=\"https:\/\/allaboutexhealth.com\/%EC%95%84%EC%B9%A8-5%EB%B6%84-%EC%8A%A4%ED%8A%B8%EB%A0%88%EC%B9%AD\/\" target=\"_blank\" rel=\"noopener\">\ub9c1\ud06c<\/a>)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ubaa9 \uc704\ub85c \ub4e4\uae30(\uc190 \uae4d\uc9c0 \ud6c4 \uc5c4\uc9c0\ub85c \ud131 \ubc00\uae30)<\/li>\n<li>\ubaa9 \uc544\ub798\ub85c \ub0b4\ub9ac\uae30(\uc591\uc190\uc73c\ub85c \uba38\ub9ac \uac10\uc2f8\uae30)<\/li>\n<li>\ubaa9 \uc0ac\uc120\uc73c\ub85c \ub298\ub824\uc8fc\uae30(\uc2b9\ubaa8\uadfc\uc774 \ub298\uc5b4\ub098\ub3c4\ub85d)<\/li>\n<li>\ud314 \uaef4\uc548\uc544 \ub2f9\uae30\uae30(\uc0bc\ub450 \ub298\ub824\uc8fc\uae30)<\/li>\n<li>\uc190\ubaa9 \ub298\ub824\uc8fc\uae30(<a href=\"http:\/\/www.soo365.co.kr\/mediinfo-20191104-1\/\" target=\"_blank\" rel=\"noopener\">http:\/\/www.soo365.co.kr\/mediinfo-20191104-1\/<\/a>)<\/li>\n<li><img src=\"https:\/\/blog.kakaocdn.net\/dn\/lQ7xt\/btrXmxgB1bL\/wkcXBCOjiKAsYIjO2qW1H1\/img.jpg\" data-image-src=\"https:\/\/blog.kakaocdn.net\/dn\/lQ7xt\/btrXmxgB1bL\/wkcXBCOjiKAsYIjO2qW1H1\/img.jpg\" data-origin-width=\"600\" data-origin-height=\"596\" \/><\/li>\n<li><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\ud314\uafc8\uce58 \ub4a4\ub85c \ub04c\uc5b4 \ub2f9\uae30\uae30(\uad11\ubc30\uadfc \uc0c1\ubd80 \ub298\ub9ac\uae30)<\/span><\/li>\n<li>\uc0c1\uccb4\ub97c \uc88c\uc6b0\ub85c \uae30\uc6b8\uc774\uae30(\uad11\ubc30\uadfc \ud558\ubd80 \ub298\ub9ac\uae30)<\/li>\n<\/ul>\n<p style=\"position: absolute;\" data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h3 data-ke-size=\"size23\">\ud558\uccb4 \uc2a4\ud2b8\ub808\uce6d - pt \uc120\uc0dd\ub2d8\uc740 \uc0c1\uccb4\ubcf4\ub2e4 \ud558\uccb4 \uc2a4\ud2b8\ub808\uce6d\uc774 \uc5b4\ub835\uace0 \uc911\uc694\ud558\uace0 \uc624\ub798 \uac78\ub9b0\ub2e4 \ud558\uc600\uc74c<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ub2e4\ub9ac \uc9e7\uac8c \ubc8c\ub9ac\uae30(\ub0b4\uc804\uadfc \ub298\ub9ac\uae30)<\/li>\n<li>\ub2e4\ub9ac \ud06c\uac8c \ubc8c\ub9ac\uae30(\ub0b4\uc804\uadfc \ub298\ub9ac\uae30, \uc548\ucabd\ubc1c\ub85c \uc9c0\uc9c0)<\/li>\n<li>\uc885\uc544\ub9ac \ub298\ub9ac\uae30(\ub113\uac8c \ubc8c\ub9b0 \uc0c1\ud0dc\uc5d0\uc11c \ub4b7\uafc8\uce58\ub85c \uc9c0\uc9c0)<\/li>\n<li><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\ud584\uc2a4\ud2b8\ub9c1 \ub298\ub9ac\uae30(\ub4b7\ubc85\uc9c0 \ub298\ub9ac\uae30)<\/span><\/li>\n<li>\uc678\uce21\uad11\uadfc \ub298\ub9ac\uae30(\ub4b7\ubc85\uc9c0-\uc606\ubc85\uc9c0 \uc0ac\uc774)<\/li>\n<li>\ub0b4\uc804\uadfc \ub298\ub9ac\uae30(\uc548\ubc85\uc9c0 \ub298\ub9ac\uae30, <a href=\"https:\/\/kr.pixtastock.com\/illustration\/81603017\" target=\"_blank\" rel=\"noopener\">\ub9c1\ud06c<\/a>)<\/li>\n<li>\ub300\ud1f4\uc0ac\ub450 \ub298\ub9ac\uae30(\uc55e\ubc85\uc9c0 \ub298\ub9ac\uae30) - \ub7f0\uc9c0 \uc790\uc138\uc5d0\uc11c \uc190\uc73c\ub85c \ud3fc\ub864\ub7ec \uc9c0\uc9c0 \ud6c4 \ub298\ub9ac\uae30<\/li>\n<li>\uc7a5\uc694\uadfc \ub298\ub9ac\uae30 - \ud3fc\ub864\ub7ec\ub85c \uc9c0\uc9c0 \ud6c4 \ub4b7\ucabd \ubc1c\uc744 \uc190\uc73c\ub85c \uc7a1\uc544\uc11c \ub298\ub9ac\uae30(<a href=\"https:\/\/m.blog.naver.com\/PostView.naver?isHttpsRedirect=true&amp;blogId=ptmaster14&amp;logNo=220290668849\" target=\"_blank\" rel=\"noopener\">\ub9c1\ud06c<\/a>)<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h3 data-ke-size=\"size23\"><span>\ud3fc\ub864\ub7ec \ub9c8\uc0ac\uc9c0 - \ubd80\uc704 \ub2f9 30~50\ubc88 \ucd94\ucc9c<\/span><\/h3>\n<h4 data-ke-size=\"size20\">\ud558\uccb4<\/h4>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud5c8\ubc85\uc9c0 \uc804\uccb4(\uc55e, \uc606, \uc548, \ub4a4) \ud070 \ubd80\ubd84 \uc704\uc8fc\ub85c<\/li>\n<li>\uc885\uc544\ub9ac \uc804\uccb4(\uc55e, \uc606, \ub4a4)<\/li>\n<li>\ud5c8\ubc85\uc9c0-\ubb34\ub98e \uc5f0\uacb0 \ubd80\ubd84<\/li>\n<li>\ub4b7\ubc85\uc9c0-\uc885\uc544\ub9ac\uc5d0 \ud3fc\ub864\ub7ec \ub07c\uace0 \uc549\uae30<\/li>\n<\/ul>\n<h4 data-ke-size=\"size20\">\uc0c1\uccb4<\/h4>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ub4f1 \uc0c1\ubd80 - \ud314\uc744 \uba85\uce58\ub85c \ubaa8\uc73c\uae30 - \uc911\ubd80, \ud558\ubd80 \uc2b9\ubaa8\uadfc \uc704\uc8fc<\/li>\n<li>\ub4f1 \uc0c1\ubd80 - \uba38\ub9ac \ub4a4 \uc190\uae4d\uc9c0 - \ub0a0\uac1c\ubf08&nbsp;<\/li>\n<li>\uad11\ubc30\uadfc - \uc0c1\ubd80 \uc704\uc8fc<\/li>\n<li>\uad11\ubc30\uadfc - \ud558\ubd80 \uc704\uc8fc<\/li>\n<li>\uad11\ubc30\uadfc-\uc0bc\ub450\uadfc<\/li>\n<li>\uac00\uc2b4 \uadfc\uc721<\/li>\n<li>\uc0bc\uac01\uadfc<\/li>\n<li>\uc0c1\ubd80\uc2b9\ubaa8\uadfc-\ubaa9<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h3 data-ke-size=\"size23\">\ub9c8\ubb34\ub9ac \uc2a4\ud2b8\ub808\uce6d \ubc0f \ubab8\ud480\uae30 \uc6b4\ub3d9<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc549\uc558\ub2e4 \uc77c\uc5b4\uc11c\uae30<\/li>\n<li>\uc549\uc544\uc11c \ub2e4\ub9ac \uc591\uc606\uc73c\ub85c \ubc8c\ub9ac\uae30<\/li>\n<li>\uc606\uc73c\ub85c \ubb34\ub98e \uafc7\uc740 \uc0c1\ud0dc\uc5d0\uc11c \ubb34\ub98e-\ubc1c \uc811\ucd09\uc2dc\ud0a4\uace0 \uc0c1\uccb4 \ub3cc\ub824\uc11c \uad11\ubc30\uadfc \ub298\ub824\uc8fc\uae30<\/li>\n<li>\uae30\ud0c0 \ub35c \ud480\ub838\ub2e4\uace0 \uc0dd\uac01\ub418\ub294 \ubd80\ubd84 \ud480\uc5b4\uc8fc\uae30<\/li>\n<\/ul>","category":"\uae30\ud0c0","author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/135","comments":"https:\/\/hoohaha.tistory.com\/135#entry135comment","pubDate":"Sun, 29 Jan 2023 22:16:24 +0900"},{"title":"[PyTorch] \uacf5\uc2dd \ubb38\uc11c Learn the Basics \uc694\uc57d - 8. The full model building process \ubc0f \uc804\uccb4 \uc694\uc57d","link":"https:\/\/hoohaha.tistory.com\/134","description":"<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"444\" data-origin-height=\"246\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/b9Xu6o\/btrWRkO9HBl\/MMgbG0otZYch6jyRU8QfT0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/b9Xu6o\/btrWRkO9HBl\/MMgbG0otZYch6jyRU8QfT0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/b9Xu6o\/btrWRkO9HBl\/MMgbG0otZYch6jyRU8QfT0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fb9Xu6o%2FbtrWRkO9HBl%2FMMgbG0otZYch6jyRU8QfT0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"444\" height=\"246\" data-origin-width=\"444\" data-origin-height=\"246\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\"><b>\ubaa9\ucc28<\/b><\/p>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\">\uc804\uccb4 \ubaa8\ub378 \uad6c\ucd95 \ud504\ub85c\uc138\uc2a4(The full moel building process)<\/p>\n<p data-ke-size=\"size16\">\ub370\uc774\ud130 \uc791\uc5c5(Working with data)<\/p>\n<p data-ke-size=\"size16\">\ubaa8\ub378 \uc0dd\uc131(Creating models)<\/p>\n<p data-ke-size=\"size16\">\ubaa8\ub378 \ub9e4\uac1c\ubcc0\uc218 \ucd5c\uc801\ud654(Optimizing the Model Parameters)<\/p>\n<p data-ke-size=\"size16\">\ubaa8\ub378 \uc800\uc7a5(Saving Models)<\/p>\n<p data-ke-size=\"size16\">\ubaa8\ub378 \ubd88\ub7ec\uc624\uae30(Loading Models)<\/p>\n<p data-ke-size=\"size16\">\uc694\uc57d(Summary)<\/p>\n<\/div>\n<\/div>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h2 data-ke-size=\"size26\">\uc804\uccb4 \ubaa8\ub378 \uad6c\ucd95 \ud504\ub85c\uc138\uc2a4(The full moel building process)<\/h2>\n<p data-ke-size=\"size16\">\uc5ec\uae30\uc5d0\uc11c\ub294 \uae30\uacc4 \ud559\uc2b5\uc758 \uc77c\ubc18\uc801\uc778 \uc791\uc5c5\uc744 \uc704\ud574 API\ub97c \ud1b5\ud574 \uc2e4\ud589 \uc608\uc815<\/p>\n<p data-ke-size=\"size16\">\uc790\uc138\ud55c \uac83\uc740 \uac01 \uc138\uc158\uc758 \ub9c1\ud06c \ucc38\uc870<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ub370\uc774\ud130 \uc791\uc5c5(Working with data)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>Pytorch:<br \/>\ub370\uc774\ud130 \uc791\uc5c5\uc744 \uc704\ud55c \ub450 \uac00\uc9c0 \ud504\ub9ac\ubbf8\ud2f0\ube0c(torch.utils.data.DataLoader, torch.utils.data.Dataset) \uc874\uc7ac<\/li>\n<li>torch.utils.data.DataLoader: Dataset\uc744 iterable\uc744 \ub9e4\ud551<\/li>\n<li>torch.utils.data.Dataset: \uc0d8\ud50c \ubc0f \ub77c\ubca8 \uc874\uc7ac<\/li>\n<\/ul>\n<pre id=\"code_1674293805278\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>%matplotlib inline\nimport torch\nfrom torch import nn\nfrom torch.utils.data import DataLoader\nfrom torchvision import datasets\nfrom torchvision.transforms import ToTensor, Lambda, Compose\nimport matplotlib.pyplot as plt<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>Pytorch:<br \/>\ub370\uc774\ud130\uc14b\uc744 \ud3ec\ud568\ud558\uc5ec \ub3c4\uba54\uc778\ubcc4 \ub77c\uc774\ube0c\ub7ec\ub9ac \uc81c\uacf5: TorchText, TorchVision, TorchAudio<\/li>\n<li>torchvision.datasets module:<br \/>CIFAR, COCO \ub4f1\uc758 \uc2e4\uc81c \ube44\uc804 \ub370\uc774\ud130\uc5d0 \ub300\ud55c Dataset \uac1c\uccb4 \ud3ec\ud568\ub418\uc5b4 \uc788\uc74c<\/li>\n<li>\ubaa8\ub4e0 TorchVision Dataset:<br \/>\uc0d8\ud50c\uacfc \ub77c\ubca8\uc744 \uc218\uc815\ud558\uae30 \uc704\ud55c trasform, target_transform \uc774\ub77c\ub294 argument \uc874\uc7ac<\/li>\n<\/ul>\n<pre id=\"code_1674293815334\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code># Download training data from open datasets.\ntraining_data = datasets.FashionMNIST(\n    root=\"data\",\n    train=True,\n    download=True,\n    transform=ToTensor(),\n)\n\n# Download test data from open datasets.\ntest_data = datasets.FashionMNIST(\n    root=\"data\",\n    train=False,\n    download=True,\n    transform=ToTensor(),\n)\n\n\nDownloading http:\/\/fashion-mnist.s3-website.eu-central-1.amazonaws.com\/train-images-idx3-ubyte.gz\nUsing downloaded and verified file: data\/FashionMNIST\/raw\/train-images-idx3-ubyte.gz\nExtracting data\/FashionMNIST\/raw\/train-images-idx3-ubyte.gz to data\/FashionMNIST\/raw\n\nDownloading http:\/\/fashion-mnist.s3-website.eu-central-1.amazonaws.com\/train-labels-idx1-ubyte.gz\nUsing downloaded and verified file: data\/FashionMNIST\/raw\/train-labels-idx1-ubyte.gz\nExtracting data\/FashionMNIST\/raw\/train-labels-idx1-ubyte.gz to data\/FashionMNIST\/raw\n\nDownloading http:\/\/fashion-mnist.s3-website.eu-central-1.amazonaws.com\/t10k-images-idx3-ubyte.gz\nUsing downloaded and verified file: data\/FashionMNIST\/raw\/t10k-images-idx3-ubyte.gz\nExtracting data\/FashionMNIST\/raw\/t10k-images-idx3-ubyte.gz to data\/FashionMNIST\/raw\n\nDownloading http:\/\/fashion-mnist.s3-website.eu-central-1.amazonaws.com\/t10k-labels-idx1-ubyte.gz\nUsing downloaded and verified file: data\/FashionMNIST\/raw\/t10k-labels-idx1-ubyte.gz\nExtracting data\/FashionMNIST\/raw\/t10k-labels-idx1-ubyte.gz to data\/FashionMNIST\/raw<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>DataLoader\uc5d0 Dataset\uc744 \uc778\uc218(argument)\ub85c \uc804\ub2ec, \uc544\ub798\uc640 \uac19\uc740 \uae30\ub2a5 \uc9c0\uc6d0<br \/>dataset\uc5d0 \ub300\ud55c iterable\uc744 wrapping<br \/>\uc790\ub3d9 \ubc30\uce58, \uc0d8\ud50c\ub9c1, \uc154\ud50c\ub9c1, \uba40\ud2f0\ud504\ub85c\uc138\uc2a4 \ub370\uc774\ud130 \ub85c\ub4dc<\/li>\n<li>batch size = 64 -&gt; dataloader iterable\uc740 \uac01 \ubc30\uce58 \ub2f9 64\uac1c\uc758 \ud53c\uccd0 \ubc0f \ub77c\ubca8 \uc81c\uacf5<\/li>\n<\/ul>\n<pre id=\"code_1674293866588\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>batch_size = 64\n\n# Create data loaders.\ntrain_dataloader = DataLoader(training_data, batch_size=batch_size)\ntest_dataloader = DataLoader(test_data, batch_size=batch_size)\n\nfor X, y in test_dataloader:\n    print(\"Shape of X [N, C, H, W]: \", X.shape)\n    print(\"Shape of y: \", y.shape, y.dtype)\n    break\n    \n# Display sample data\nfigure = plt.figure(figsize=(10, 8))\ncols, rows = 5, 5\nfor i in range(1, cols * rows + 1):\n    idx = torch.randint(len(test_data), size=(1,)).item()\n    img, label = test_data[idx]\n    figure.add_subplot(rows, cols, i)\n    plt.title(label)\n    plt.axis(\"off\")\n    plt.imshow(img.squeeze(), cmap=\"gray\")\nplt.show()\n\nShape of X [N, C, H, W]:  torch.Size([64, 1, 28, 28])\nShape of y:  torch.Size([64]) torch.int64<\/code><\/pre>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"767\" data-origin-height=\"656\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/chV9dL\/btrWR8gg62l\/fAVrif5nri1EAR81lGzKhK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/chV9dL\/btrWR8gg62l\/fAVrif5nri1EAR81lGzKhK\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/chV9dL\/btrWR8gg62l\/fAVrif5nri1EAR81lGzKhK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FchV9dL%2FbtrWR8gg62l%2FfAVrif5nri1EAR81lGzKhK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"767\" height=\"656\" data-origin-width=\"767\" data-origin-height=\"656\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ubaa8\ub378 \uc0dd\uc131(Creating models)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc2e0\uacbd\ub9dd \uc815\uc758 \uc704\ud574 nn.Module\uc744 \uc0c1\uc18d\ud558\ub294 \ud074\ub798\uc2a4 \uc0dd\uc131<\/li>\n<li>__init__: \ub124\ud2b8\uc6cc\ud06c \uacc4\uce35 \uc815\uc758<\/li>\n<li>forward: \uc778\ud48b\uc774 \ub124\ud2b8\uc6cc\ud06c\ub97c \ud1b5\uacfc\ud558\ub294 \ubc29\ubc95 \uc815\uc758<\/li>\n<li>GPU \uc0ac\uc6a9 \uac00\ub2a5\ud55c \uacbd\uc6b0 \uc2e0\uacbd\ub9dd \uc5f0\uc0b0 \uac00\uc18d<\/li>\n<\/ul>\n<pre id=\"code_1674293934496\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code># Get cpu or gpu device for training.\ndevice = \"cuda\" if torch.cuda.is_available() else \"cpu\"\nprint(\"Using {} device\".format(device))\n\n# Define model\nclass NeuralNetwork(nn.Module):\n    def __init__(self):\n        super(NeuralNetwork, self).__init__()\n        self.flatten = nn.Flatten()\n        self.linear_relu_stack = nn.Sequential(\n            nn.Linear(28*28, 512),\n            nn.ReLU(),\n            nn.Linear(512, 512),\n            nn.ReLU(),\n            nn.Linear(512, 10),\n            nn.ReLU()\n        )\n\n    def forward(self, x):\n        x = self.flatten(x)\n        logits = self.linear_relu_stack(x)\n        return logits\n\nmodel = NeuralNetwork().to(device)\nprint(model)\n\n\nUsing cuda device\nNeuralNetwork(\n  (flatten): Flatten(start_dim=1, end_dim=-1)\n  (linear_relu_stack): Sequential(\n    (0): Linear(in_features=784, out_features=512, bias=True)\n    (1): ReLU()\n    (2): Linear(in_features=512, out_features=512, bias=True)\n    (3): ReLU()\n    (4): Linear(in_features=512, out_features=10, bias=True)\n    (5): ReLU()\n  )\n)<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ubaa8\ub378 \ub9e4\uac1c\ubcc0\uc218 \ucd5c\uc801\ud654(Optimizing the Model Parameters)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ubaa8\ub378 \ud559\uc2b5\uc5d0 \ud544\uc694\ud55c \ub85c\uc2a4 \ud568\uc218 \ubc0f \uc635\ud2f0\ub9c8\uc774\uc800 \uc815\uc758<br \/>loss function: nn.CrossEntropyLoss<br \/>optimizer: Stochastic Gradient Descent<\/li>\n<\/ul>\n<pre id=\"code_1674293977713\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>loss_fn = nn.CrossEntropyLoss()\nlearning_rate = 1e-3\noptimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud558\ub098\uc758 \ud559\uc2b5 \ub8e8\ud504 \uc548\uc5d0\uc11c\uc758 \ubaa8\ub378\uc758 \ub3d9\uc791:<\/li>\n<\/ul>\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"decimal\">\n<li>\ubc30\uce58\ub85c \uc81c\uacf5\ub41c \ud559\uc2b5 \ub370\uc774\ud130 \uc14b\uc744 \uc774\uc6a9\ud558\uc5ec \uc608\uce21 \uc218\ud589<\/li>\n<li>\uc624\ub958 \uc5ed\uc804\ud30c<\/li>\n<li>\ubaa8\ub378 \ud559\uc2b5<\/li>\n<\/ol>\n<pre id=\"code_1674293984722\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>def train(dataloader, model, loss_fn, optimizer):\n    size = len(dataloader.dataset)\n    for batch, (X, y) in enumerate(dataloader):\n        X, y = X.to(device), y.to(device)\n        \n        # Compute prediction error\n        pred = model(X)\n        loss = loss_fn(pred, y)\n        \n        # Backpropagation\n        optimizer.zero_grad()\n        loss.backward()\n        optimizer.step()\n\n        if batch % 100 == 0:\n            loss, current = loss.item(), batch * len(X)\n            print(f\"loss: {loss:&gt;7f}  [{current:&gt;5d}\/{size:&gt;5d}]\")<\/code><\/pre>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ubaa8\ub378 \uc798 \ud559\uc2b5\ud558\uace0 \uc788\ub294\uc9c0 \ud655\uc778:<br \/>\ud14c\uc2a4\ud2b8 \ub370\uc774\ud130\uc14b \uc774\uc6a9<\/li>\n<\/ul>\n<pre id=\"code_1674294001802\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>def test(dataloader, model):\n    size = len(dataloader.dataset)\n    model.eval()\n    test_loss, correct = 0, 0\n    with torch.no_grad():\n        for X, y in dataloader:\n            X, y = X.to(device), y.to(device)\n            pred = model(X)\n            test_loss += loss_fn(pred, y).item()\n            correct += (pred.argmax(1) == y).type(torch.float).sum().item()\n    test_loss \/= size\n    correct \/= size\n    print(f\"Test Error: \\n Accuracy: {(100*correct):&gt;0.1f}%, Avg loss: {test_loss:&gt;8f} \\n\")<\/code><\/pre>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud559\uc2b5 \ud504\ub85c\uc138\uc2a4: \uc5ec\ub7ec \uc5d0\ud3ed\uc5d0 \uac78\uccd0 \uc218\ud589<\/li>\n<li>\uac01 \uc5d0\ud3ed \ub0b4\uc5d0\uc11c \ubaa8\ub378\uc740 \ub9e4\uac1c\ubcc0\uc218 \ud559\uc2b5<\/li>\n<li>\ubaa8\ub378\uc758 \uc815\ud655\ub3c4\uc640 \uc190\uc2e4\uc744 \uac01 \uc5d0\ud3ed\uc5d0\uc11c \uc778\uc1c4 -&gt; \ub9e4 \uc5d0\ud3ed\ub9c8\ub2e4 \uc815\ud655\ub3c4\uac00 \uc99d\uac00\ud558\ub294\uc9c0, \uc190\uc2e4\uc774 \uac10\uc18c\ud558\ub294\uc9c0 \ud655\uc778<\/li>\n<\/ul>\n<pre id=\"code_1674294013892\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>epochs = 15\nfor t in range(epochs):\n    print(f\"Epoch {t+1}\\n-------------------------------\")\n    train(train_dataloader, model, loss_fn, optimizer)\n    test(test_dataloader, model)\nprint(\"Done!\")\n\n\nEpoch 1\n-------------------------------\nloss: 2.295773  [    0\/60000]\nloss: 2.297218  [ 6400\/60000]\nloss: 2.283425  [12800\/60000]\nloss: 2.278249  [19200\/60000]\nloss: 2.279926  [25600\/60000]\nloss: 2.260619  [32000\/60000]\nloss: 2.256614  [38400\/60000]\nloss: 2.248361  [44800\/60000]\nloss: 2.234987  [51200\/60000]\nloss: 2.211842  [57600\/60000]\nTest Error: \n Accuracy: 41.4%, Avg loss: 0.034903 \n\nEpoch 2\n-------------------------------\nloss: 2.227791  [    0\/60000]\nloss: 2.239108  [ 6400\/60000]\nloss: 2.196185  [12800\/60000]\nloss: 2.183881  [19200\/60000]\nloss: 2.218200  [25600\/60000]\nloss: 2.170907  [32000\/60000]\nloss: 2.167650  [38400\/60000]\nloss: 2.151818  [44800\/60000]\nloss: 2.121467  [51200\/60000]\nloss: 2.078620  [57600\/60000]\nTest Error: \n Accuracy: 47.1%, Avg loss: 0.033028 \n\nEpoch 3\n-------------------------------\nloss: 2.121974  [    0\/60000]\nloss: 2.144973  [ 6400\/60000]\nloss: 2.049951  [12800\/60000]\nloss: 2.025212  [19200\/60000]\nloss: 2.119183  [25600\/60000]\nloss: 2.028407  [32000\/60000]\nloss: 2.016531  [38400\/60000]\nloss: 1.999072  [44800\/60000]\nloss: 1.939542  [51200\/60000]\nloss: 1.869733  [57600\/60000]\nTest Error: \n Accuracy: 47.2%, Avg loss: 0.030161 \n\nEpoch 4\n-------------------------------\nloss: 1.966496  [    0\/60000]\nloss: 2.005283  [ 6400\/60000]\nloss: 1.842639  [12800\/60000]\nloss: 1.798291  [19200\/60000]\nloss: 1.996712  [25600\/60000]\nloss: 1.860519  [32000\/60000]\nloss: 1.833745  [38400\/60000]\nloss: 1.834117  [44800\/60000]\nloss: 1.752414  [51200\/60000]\nloss: 1.658906  [57600\/60000]\nTest Error: \n Accuracy: 48.4%, Avg loss: 0.027377 \n\nEpoch 5\n-------------------------------\nloss: 1.817958  [    0\/60000]\nloss: 1.873164  [ 6400\/60000]\nloss: 1.658233  [12800\/60000]\nloss: 1.604004  [19200\/60000]\nloss: 1.892859  [25600\/60000]\nloss: 1.728574  [32000\/60000]\nloss: 1.694584  [38400\/60000]\nloss: 1.716881  [44800\/60000]\nloss: 1.620669  [51200\/60000]\nloss: 1.520065  [57600\/60000]\nTest Error: \n Accuracy: 51.3%, Avg loss: 0.025398 \n\nEpoch 6\n-------------------------------\nloss: 1.705050  [    0\/60000]\nloss: 1.772956  [ 6400\/60000]\nloss: 1.524106  [12800\/60000]\nloss: 1.473383  [19200\/60000]\nloss: 1.793820  [25600\/60000]\nloss: 1.624757  [32000\/60000]\nloss: 1.595235  [38400\/60000]\nloss: 1.626289  [44800\/60000]\nloss: 1.523839  [51200\/60000]\nloss: 1.426233  [57600\/60000]\nTest Error: \n Accuracy: 53.5%, Avg loss: 0.023906 \n\nEpoch 7\n-------------------------------\nloss: 1.612589  [    0\/60000]\nloss: 1.694386  [ 6400\/60000]\nloss: 1.419676  [12800\/60000]\nloss: 1.378877  [19200\/60000]\nloss: 1.704812  [25600\/60000]\nloss: 1.536829  [32000\/60000]\nloss: 1.516409  [38400\/60000]\nloss: 1.551574  [44800\/60000]\nloss: 1.448999  [51200\/60000]\nloss: 1.354901  [57600\/60000]\nTest Error: \n Accuracy: 54.4%, Avg loss: 0.022727 \n\nEpoch 8\n-------------------------------\nloss: 1.532547  [    0\/60000]\nloss: 1.630929  [ 6400\/60000]\nloss: 1.336020  [12800\/60000]\nloss: 1.305433  [19200\/60000]\nloss: 1.631139  [25600\/60000]\nloss: 1.462314  [32000\/60000]\nloss: 1.452916  [38400\/60000]\nloss: 1.491249  [44800\/60000]\nloss: 1.388288  [51200\/60000]\nloss: 1.299752  [57600\/60000]\nTest Error: \n Accuracy: 55.3%, Avg loss: 0.021782 \n\nEpoch 9\n-------------------------------\nloss: 1.462481  [    0\/60000]\nloss: 1.578309  [ 6400\/60000]\nloss: 1.268808  [12800\/60000]\nloss: 1.243797  [19200\/60000]\nloss: 1.572057  [25600\/60000]\nloss: 1.400606  [32000\/60000]\nloss: 1.401813  [38400\/60000]\nloss: 1.444131  [44800\/60000]\nloss: 1.337315  [51200\/60000]\nloss: 1.257235  [57600\/60000]\nTest Error: \n Accuracy: 55.8%, Avg loss: 0.021020 \n\nEpoch 10\n-------------------------------\nloss: 1.402163  [    0\/60000]\nloss: 1.532682  [ 6400\/60000]\nloss: 1.213558  [12800\/60000]\nloss: 1.192357  [19200\/60000]\nloss: 1.524502  [25600\/60000]\nloss: 1.349040  [32000\/60000]\nloss: 1.359848  [38400\/60000]\nloss: 1.406721  [44800\/60000]\nloss: 1.294481  [51200\/60000]\nloss: 1.222457  [57600\/60000]\nTest Error: \n Accuracy: 56.7%, Avg loss: 0.020403 \n\nEpoch 11\n-------------------------------\nloss: 1.351255  [    0\/60000]\nloss: 1.496092  [ 6400\/60000]\nloss: 1.168390  [12800\/60000]\nloss: 1.149713  [19200\/60000]\nloss: 1.487448  [25600\/60000]\nloss: 1.306274  [32000\/60000]\nloss: 1.324405  [38400\/60000]\nloss: 1.376118  [44800\/60000]\nloss: 1.258274  [51200\/60000]\nloss: 1.193912  [57600\/60000]\nTest Error: \n Accuracy: 57.4%, Avg loss: 0.019891 \n\nEpoch 12\n-------------------------------\nloss: 1.305900  [    0\/60000]\nloss: 1.466068  [ 6400\/60000]\nloss: 1.130126  [12800\/60000]\nloss: 1.114587  [19200\/60000]\nloss: 1.458147  [25600\/60000]\nloss: 1.270395  [32000\/60000]\nloss: 1.293715  [38400\/60000]\nloss: 1.350162  [44800\/60000]\nloss: 1.227087  [51200\/60000]\nloss: 1.169333  [57600\/60000]\nTest Error: \n Accuracy: 58.3%, Avg loss: 0.019452 \n\nEpoch 13\n-------------------------------\nloss: 1.264979  [    0\/60000]\nloss: 1.439438  [ 6400\/60000]\nloss: 1.097137  [12800\/60000]\nloss: 1.085567  [19200\/60000]\nloss: 1.433970  [25600\/60000]\nloss: 1.239276  [32000\/60000]\nloss: 1.266291  [38400\/60000]\nloss: 1.327267  [44800\/60000]\nloss: 1.199434  [51200\/60000]\nloss: 1.148261  [57600\/60000]\nTest Error: \n Accuracy: 59.0%, Avg loss: 0.019064 \n\nEpoch 14\n-------------------------------\nloss: 1.227596  [    0\/60000]\nloss: 1.414626  [ 6400\/60000]\nloss: 1.067816  [12800\/60000]\nloss: 1.061232  [19200\/60000]\nloss: 1.413360  [25600\/60000]\nloss: 1.212436  [32000\/60000]\nloss: 1.240914  [38400\/60000]\nloss: 1.306429  [44800\/60000]\nloss: 1.174499  [51200\/60000]\nloss: 1.129985  [57600\/60000]\nTest Error: \n Accuracy: 59.8%, Avg loss: 0.018713 \n\nEpoch 15\n-------------------------------\nloss: 1.192792  [    0\/60000]\nloss: 1.391189  [ 6400\/60000]\nloss: 1.041390  [12800\/60000]\nloss: 1.040678  [19200\/60000]\nloss: 1.396070  [25600\/60000]\nloss: 1.188643  [32000\/60000]\nloss: 1.217234  [38400\/60000]\nloss: 1.287394  [44800\/60000]\nloss: 1.151905  [51200\/60000]\nloss: 1.113297  [57600\/60000]\nTest Error: \n Accuracy: 60.6%, Avg loss: 0.018389 \n\nDone!<\/code><\/pre>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ucc98\uc74c\uc5d0 \uc815\ud655\ub3c4 \uc88b\uc9c0 \uc54a\uc744 \uc218 \uc788\uc74c<\/li>\n<li>\uc5d0\ud3ed \uc218 \ub298\ub9ac\uac70\ub098 learning_rate \uc99d\uac00 \uace0\ub824<\/li>\n<li>\uc120\ud0dd\ud55c \ubaa8\ub378\uc774 \ucd5c\uc801\uc758 \ubaa8\ub378\uc774 \uc544\ub2d0 \uc218 \uc788\uc74c<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ubaa8\ub378 \uc800\uc7a5(Saving Models)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ubaa8\ub378\uc744 \uc800\uc7a5\ud558\ub294 \uc77c\ubc18\uc801\uc778 \ubc29\ubc95:<br \/>internal state dictionary\ub97c \uc9c1\ub82c\ud654(serialize)\ud558\ub294 \uac83<\/li>\n<\/ul>\n<pre id=\"code_1674295424961\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>torch.save(model.state_dict(), \"data\/model.pth\")\nprint(\"Saved PyTorch Model State to model.pth\")\n\n\nSaved PyTorch Model State to model.pth<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ubaa8\ub378 \ubd88\ub7ec\uc624\uae30(Loading Models)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ubaa8\ub378 \ub85c\ub4dc\ud558\ub294 \ud504\ub85c\uc138\uc2a4:<\/li>\n<\/ul>\n<ol style=\"list-style-type: decimal;\" data-ke-list-type=\"decimal\">\n<li>\ubaa8\ub378\uc758 \uad6c\uc870 \uc0dd\uc131<\/li>\n<li>\uad6c\uc870\uc5d0 internal state dictionary \ub85c\ub4dc<\/li>\n<\/ol>\n<pre id=\"code_1674295537411\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>model = NeuralNetwork()\nmodel.load_state_dict(torch.load(\"data\/model.pth\"))\n\n\n&lt;All keys matched successfully&gt;<\/code><\/pre>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc704 \ubaa8\ub378 \uc0ac\uc6a9\ud558\uc5ec \uc608\uce21 \uac00\ub2a5<\/li>\n<\/ul>\n<pre id=\"code_1674295573503\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>classes = [\n    \"T-shirt\/top\",\n    \"Trouser\",\n    \"Pullover\",\n    \"Dress\",\n    \"Coat\",\n    \"Sandal\",\n    \"Shirt\",\n    \"Sneaker\",\n    \"Bag\",\n    \"Ankle boot\",\n]\n\nmodel.eval()\nx, y = test_data[0][0], test_data[0][1]\nwith torch.no_grad():\n    pred = model(x)\n    predicted, actual = classes[pred[0].argmax(0)], classes[y]\n    print(f'Predicted: \"{predicted}\", Actual: \"{actual}\"')\n\n\nPredicted: \"Sandal\", Actual: \"Ankle boot\"<\/code><\/pre>\n<p data-ke-size=\"size16\">\ud29c\ud1a0\ub9ac\uc5bc \uc644\ub8cc<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h2 data-ke-size=\"size26\">\uc694\uc57d(Summary)<\/h2>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud604\uc7ac\uae4c\uc9c0 \uc2e0\uacbd\ub9dd \uc0ac\uc6a9\ud558\uc5ec \uae30\uacc4 \ud559\uc2b5 \ubaa8\ub378 \uad6c\ucd95\ud558\ub294 \ud575\uc2ec \uac1c\ub150 \uacf5\ubd80 \ubc0f pytorch\ub85c \uad6c\ud604<\/li>\n<li>FashionMNIST \uc774\uc6a9\ud55c \uc774\ubbf8\uc9c0 \ubd84\ub958 \ubaa8\ub378 \uad6c\ucd95\ud568<\/li>\n<li>\ub2e4\uc74c\uacfc \uac19\uc740 \ubd84\uc57c \ubc30\uc6c0<\/li>\n<li>Tensor\ub97c CPU, GPU\uc640 \ud568\uaed8 \uc0ac\uc6a9\ud558\ub294 \ubc29\ubc95<\/li>\n<li>\ub370\uc774\ud130\uc14b manage, scale, normalize \ubc29\ubc95<\/li>\n<li>\uc2e0\uacbd\ub9dd\uc744 \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc744 \uad6c\ucd95\ud558\ub294 \ubc29\ubc95<\/li>\n<li>\ubaa8\ub378 \ucd5c\uc801\ud654 \ubc29\ubc95<\/li>\n<li>\ubaa8\ub378 \ucd94\ub860 \ucd5c\uc801\ud654 \ubc29\ubc95(model.eval(), with torch.no_grad(), require_grad=False, tensor.detach() \ub4f1)<\/li>\n<\/ul>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>require_grad=False: <br \/>\ud150\uc11c\uc5d0 \ub300\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8\uac00 \uc5ed\ubc29\ud5a5 \ud328\uc2a4 \uc911\uc5d0 \uacc4\uc0b0\ub418\uc5b4\uc57c \ud558\ub294\uc9c0 \uc5ec\ubd80\ub97c \ub098\ud0c0\ub0b4\ub294 PyTorch \ud150\uc11c\uc758 \uc18d\uc131<br \/>\uae30\ubcf8\uc801\uc73c\ub85c PyTorch\uc5d0 \uc758\ud574 \uc0dd\uc131\ub41c \ubaa8\ub4e0 \ud150\uc11c: requires_grad=True<br \/>False\uc778 \uacbd\uc6b0 \ud150\uc11c\uc5d0 \ub300\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8\ub294 \uc5ed\ubc29\ud5a5 \uc804\ub2ec \uc911\uc5d0 \uacc4\uc0b0\ub418\uc9c0 \uc54a\uc73c\uba70 \ud150\uc11c\uc758 grad \uc18d\uc131\uc740 \uc5c5\ub370\uc774\ud2b8\ub418\uc9c0 \uc54a\uc74c<\/li>\n<li>torch.no_grad():<br \/>\ube14\ub85d \ub0b4\uc758 \ubaa8\ub4e0 \ud150\uc11c\uc5d0 \ub300\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8 \ucd94\uc801\uc744 \uc77c\uc2dc\uc801\uc73c\ub85c \ube44\ud65c\uc131\ud654\ud558\ub294 PyTorch\uc5d0\uc11c \uc81c\uacf5\ud558\ub294 \ucee8\ud14d\uc2a4\ud2b8 \uad00\ub9ac\uc790<br \/>\uadf8\ub798\ub514\uc5b8\ud2b8 \ucd94\uc801\uc774 \ube44\ud65c\uc131\ud654\ub418\uba74 \ud3ec\uc6cc\ub4dc \ud328\uc2a4\uac00 \ud3c9\uc18c\uc640 \uac19\uc774 \uc218\ud589\ub418\uc9c0\ub9cc \uadf8\ub798\ub514\uc5b8\ud2b8\uac00 \uacc4\uc0b0\ub418\uc9c0 \uc54a\uace0 \ud150\uc11c\uc758 \uadf8\ub798\ub4dc \uc18d\uc131\uc774 \uc5c5\ub370\uc774\ud2b8\ub418\uc9c0 \uc54a\uc74c<br \/>\uadf8\ub798\ub514\uc5b8\ud2b8\uac00 \ud544\uc694\ud558\uc9c0 \uc54a\uc744 \ub54c \uacc4\uc0b0 \uc18d\ub3c4\ub97c \ub192\uc774\uace0 \uba54\ubaa8\ub9ac\ub97c \uc808\uc57d\ud558\ub294 \ub370 \uc720\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<li>detach(): <br \/>\uc6d0\ub798 \ud150\uc11c\uc640 \ub3d9\uc77c\ud55c \uc800\uc7a5\uc18c\ub97c \uacf5\uc720\ud558\uc9c0\ub9cc \uadf8\ub798\ub514\uc5b8\ud2b8 \ucd94\uc801\uc774 \ube44\ud65c\uc131\ud654\ub41c \uc0c8 \ud150\uc11c\ub97c \ubc18\ud658\ud558\ub294 PyTorch\uc5d0\uc11c \uc81c\uacf5\ud558\ub294 \uba54\uc11c\ub4dc<br \/>\uc6d0\ub798 \ud150\uc11c\ub294 \uc218\uc815\ub418\uc9c0 \uc54a\uc9c0\ub9cc \uc0c8 \ud150\uc11c\ub294 \uae30\uc6b8\uae30\uac00 \uacc4\uc0b0\ub418\uc9c0 \uc54a\uc73c\uba70 grad \uc18d\uc131\uc774 \uc5c5\ub370\uc774\ud2b8\ub418\uc9c0 \uc54a\uc74c<br \/>\uc0ac\uc6a9\uc790\uac00 \ucd94\uac00 \uacc4\uc0b0\uc744 \uc704\ud574 \uc6d0\ub798 \ud150\uc11c\ub97c \uc720\uc9c0\ud558\uace0 \uc2f6\uc9c0\ub9cc \uae30\uc6b8\uae30\uac00 \ud750\ub974\ub294 \uac83\uc744 \ubc29\uc9c0\ud558\ub824\ub294 \uacbd\uc6b0\uc5d0 \uc720\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">\uc694\uc57d:<br \/>requires_grad=False: \ud150\uc11c\uc5d0 \ub300\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8 \ucd94\uc801\uc744 \uc601\uad6c\uc801\uc73c\ub85c \ube44\ud65c\uc131\ud654\ud558\ub294 \ubc29\ubc95<br \/>torch.no_grad(): \uadf8\ub798\ub514\uc5b8\ud2b8 \ucd94\uc801\uc744 \uc77c\uc2dc\uc801\uc73c\ub85c \ube44\ud65c\uc131\ud654\ud558\ub294 \ucee8\ud14d\uc2a4\ud2b8 \uad00\ub9ac\uc790<br \/>detach(): \ud150\uc11c\ub97c \uacc4\uc0b0 \uae30\ub85d\uc5d0\uc11c \ubd84\ub9ac\ud558\ub294 \ubc29\ubc95\uc73c\ub85c \uc720\uc6a9\ud560 \uc218 \uc788\uc74c, \uadf8\ub798\ub514\uc5b8\ud2b8 \ud750\ub974\ub294 \uac83 \ubc29\uc9c0<\/p>\n<\/div>\n<\/div>","category":"Pytorch\/\ud29c\ud1a0\ub9ac\uc5bc","author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/134","comments":"https:\/\/hoohaha.tistory.com\/134#entry134comment","pubDate":"Sat, 21 Jan 2023 19:20:25 +0900"},{"title":"[PyTorch] \uacf5\uc2dd \ubb38\uc11c Learn the Basics \uc694\uc57d - 7. Save and load the model","link":"https:\/\/hoohaha.tistory.com\/133","description":"<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"444\" data-origin-height=\"246\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/2kSyA\/btrWRRk1GF9\/I7NqNGfbxef650x9gEKsOk\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/2kSyA\/btrWRRk1GF9\/I7NqNGfbxef650x9gEKsOk\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/2kSyA\/btrWRRk1GF9\/I7NqNGfbxef650x9gEKsOk\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2F2kSyA%2FbtrWRRk1GF9%2FI7NqNGfbxef650x9gEKsOk%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"444\" height=\"246\" data-origin-width=\"444\" data-origin-height=\"246\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">\ubaa9\ucc28<\/p>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\">\ubaa8\ub378 \uc800\uc7a5\ud558\uace0 \ubd88\ub7ec\uc624\uae30(Save and load the model)<\/p>\n<p data-ke-size=\"size16\">\ubaa8\ub378 \uac00\uc911\uce58 \uc800\uc7a5\ud558\uace0 \ubd88\ub7ec\uc624\uae30(Saving and Loading Model Weights)<\/p>\n<p data-ke-size=\"size16\">\ubaa8\ub378\uc758 \ud615\ud0dc\ub97c \ud3ec\ud568\ud558\uc5ec \uc800\uc7a5\ud558\uace0 \ubd88\ub7ec\uc624\uae30(Saving and Loading Models with Shapes)<\/p>\n<p data-ke-size=\"size16\">\ubaa8\ub378 \ucd94\ub860(Model Inference)<\/p>\n<p data-ke-size=\"size16\">\ubaa8\ub378\uc744 ONNX\ub85c \ub0b4\ubcf4\ub0b4\uae30(Exporting the model to ONNX)<\/p>\n<\/div>\n<\/div>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">keyword:<\/p>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\">torch.save, model.state_dict(), model.load_state_dict(), torch.load(), pickle module, \uc9c1\ub82c\ud654(serialize), \uc5ed\uc9c1\ub82c\ud654(deserialize), model.eval(), \ub4dc\ub86d\uc544\uc6c3(dropout), \ubc30\uce58 \uc815\uaddc\ud654(batch normalization), %matplotlib inline, ONNX(Open Neural Network Exchange), onnxruntime, torch.onnx, Java, JavaScript, C#, ML.NET, \uc2e4\ud589 \uadf8\ub798\ud504(execution graph), ONNX \ub0b4\ubcf4\ub0b4\uae30(ONNX export), Persist, model.onnx, onnx.export, onnxruntime.InferenceSession, onnxruntime.InferenceSession.get_inputs(), onnxruntime.InferenceSession.run()<\/p>\n<\/div>\n<\/div>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h2 data-ke-size=\"size26\">\ubaa8\ub378 \uc800\uc7a5\ud558\uace0 \ubd88\ub7ec\uc624\uae30(Save and load the model)<\/h2>\n<p data-ke-size=\"size16\">\uc774\ubc88 \uc7a5\uc5d0\uc11c\ub294 \uc800\uc7a5\ud558\uae30\ub098 \ubd88\ub7ec\uc624\uae30\ub97c \ud1b5\ud574 \ubaa8\ub378\uc758 \uc0c1\ud0dc\ub97c \uc720\uc9c0(persist)\ud558\uace0 \ubaa8\ub378\uc758 \uc608\uce21\uc744 \uc2e4\ud589\ud558\ub294 \ubc29\ubc95\uc744 \uc54c\uc544\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n<pre id=\"code_1674229421512\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>import torch\nimport torchvision.models as models<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ubaa8\ub378 \uac00\uc911\uce58 \uc800\uc7a5\ud558\uace0 \ubd88\ub7ec\uc624\uae30(Saving and Loading Model Weights)<\/h3>\n<p data-ke-size=\"size16\">PyTorch \ubaa8\ub378\uc740 \ud559\uc2b5\ud55c \ub9e4\uac1c\ubcc0\uc218\ub97c<span>&nbsp;<\/span><span style=\"background-color: #f3f4f7; color: #6c6c6d;\">state_dict<\/span>\ub77c\uace0 \ubd88\ub9ac\ub294 \ub0b4\ubd80 \uc0c1\ud0dc \uc0ac\uc804(internal state dictionary)\uc5d0 \uc800\uc7a5\ud569\ub2c8\ub2e4. \uc774 \uc0c1\ud0dc \uac12\ub4e4\uc740<span>&nbsp;<\/span><span style=\"background-color: #f3f4f7; color: #6c6c6d;\">torch.save<\/span><span>&nbsp;<\/span>\uba54\uc18c\ub4dc\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc800\uc7a5(persist)\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n<pre id=\"code_1674229498292\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>model = models.vgg16(pretrained=True)\ntorch.save(model.state_dict(), 'model_weights.pth')<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\"><span style=\"color: #333333; font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; font-size: 16px; letter-spacing: 0px;\">\ubaa8\ub378 \uac00\uc911\uce58\ub97c \ubd88\ub7ec\uc624\uae30 \uc704\ud574\uc11c\ub294, \uba3c\uc800 \ub3d9\uc77c\ud55c \ubaa8\ub378\uc758 \uc778\uc2a4\ud134\uc2a4(instance)\ub97c \uc0dd\uc131\ud55c \ub2e4\uc74c\uc5d0<\/span><span style=\"color: #333333; font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; font-size: 16px; letter-spacing: 0px;\">&nbsp;<\/span><span style=\"background-color: #f3f4f7; color: #6c6c6d;\">load_state_dict()<\/span><span style=\"color: #333333; font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; font-size: 16px; letter-spacing: 0px;\">&nbsp;<\/span><span style=\"color: #333333; font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; font-size: 16px; letter-spacing: 0px;\">\uba54\uc18c\ub4dc\ub97c \uc0ac\uc6a9\ud558\uc5ec \ub9e4\uac1c\ubcc0\uc218\ub4e4\uc744 \ubd88\ub7ec\uc635\ub2c8\ub2e4.<\/span><\/h3>\n<div>\n<div>\n<pre id=\"code_1674229662272\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>model = models.vgg16() # \uae30\ubcf8 \uac00\uc911\uce58\ub97c \ubd88\ub7ec\uc624\uc9c0 \uc54a\uc73c\ubbc0\ub85c pretrained=True\ub97c \uc9c0\uc815\ud558\uc9c0 \uc54a\uc2b5\ub2c8\ub2e4.\nmodel.load_state_dict(torch.load('model_weights.pth'))\nmodel.eval()<\/code><\/pre>\n<\/div>\n<\/div>\n<div id=\"id4\">\n<div>\n<p data-ke-size=\"size16\">\ucd94\ub860(inference)\uc744 \ud558\uae30 \uc804\uc5d0<span>&nbsp;<\/span><span style=\"background-color: #ffffff; color: #6c6c6d;\">model.eval()<\/span><span>&nbsp;<\/span>\uba54\uc18c\ub4dc\ub97c \ud638\ucd9c\ud558\uc5ec \ub4dc\ub86d\uc544\uc6c3(dropout)\uacfc \ubc30\uce58 \uc815\uaddc\ud654(batch normalization)\ub97c \ud3c9\uac00 \ubaa8\ub4dc(evaluation mode)\ub85c \uc124\uc815\ud574\uc57c \ud569\ub2c8\ub2e4. \uadf8\ub807\uc9c0 \uc54a\uc73c\uba74 \uc77c\uad00\uc131 \uc5c6\ub294 \ucd94\ub860 \uacb0\uacfc\uac00 \uc0dd\uc131\ub429\ub2c8\ub2e4.<\/p>\n<\/div>\n<\/div>\n<div id=\"id5\">&nbsp;<\/div>\n<div>\n<h3 data-ke-size=\"size23\">\ubaa8\ub378\uc758 \ud615\ud0dc\ub97c \ud3ec\ud568\ud558\uc5ec \uc800\uc7a5\ud558\uace0 \ubd88\ub7ec\uc624\uae30(Saving and Loading Models with Shapes)<\/h3>\n<p data-ke-size=\"size16\">\ubaa8\ub378\uc758 \uac00\uc911\uce58\ub97c \ubd88\ub7ec\uc62c \ub54c, \uc2e0\uacbd\ub9dd\uc758 \uad6c\uc870\ub97c \uc815\uc758\ud558\uae30 \uc704\ud574 \ubaa8\ub378 \ud074\ub798\uc2a4\ub97c \uba3c\uc800 \uc0dd\uc131(instantiate)\ud574\uc57c \ud588\uc2b5\ub2c8\ub2e4. \uc774 \ud074\ub798\uc2a4\uc758 \uad6c\uc870\ub97c \ubaa8\ub378\uacfc \ud568\uaed8 \uc800\uc7a5\ud558\uace0 \uc2f6\uc73c\uba74, (<span style=\"background-color: #f3f4f7; color: #6c6c6d;\">model.state_dict()<\/span>\uac00 \uc544\ub2cc)<span>&nbsp;<\/span><span style=\"background-color: #f3f4f7; color: #6c6c6d;\">model<\/span><span>&nbsp;<\/span>\uc744 \uc800\uc7a5 \ud568\uc218\uc5d0 \uc804\ub2ec\ud569\ub2c8\ub2e4:<\/p>\n<div>\n<div>\n<pre id=\"code_1674233285014\" class=\"gams\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>torch.save(model, 'model.pth')<\/code><\/pre>\n<\/div>\n<\/div>\n<p data-ke-size=\"size16\">\ub2e4\uc74c\uacfc \uac19\uc774 \ubaa8\ub378\uc744 \ubd88\ub7ec\uc62c \uc218 \uc788\uc2b5\ub2c8\ub2e4:<\/p>\n<div>\n<div>\n<pre id=\"code_1674233285014\" class=\"ini\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>model = torch.load('model.pth')<\/code><\/pre>\n<\/div>\n<\/div>\n<div id=\"id5\">\n<div>\n<p data-ke-size=\"size16\">\uc774 \uc811\uadfc \ubc29\uc2dd\uc740 Python<span>&nbsp;<\/span><a href=\"https:\/\/docs.python.org\/3\/library\/pickle.html\">pickle<\/a><span>&nbsp;<\/span>\ubaa8\ub4c8\uc744 \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc744 \uc9c1\ub82c\ud654(serialize)\ud558\ubbc0\ub85c, \ubaa8\ub378\uc744 \ubd88\ub7ec\uc62c \ub54c \uc2e4\uc81c \ud074\ub798\uc2a4 \uc815\uc758(definition)\ub97c \uc801\uc6a9(rely on)\ud569\ub2c8\ub2e4.<\/p>\n<\/div>\n<\/div>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">keyword: torch.save, model.state_dict(), model.load_state_dict(), torch.load(), pickle module, \uc9c1\ub82c\ud654(serialize), \uc5ed\uc9c1\ub82c\ud654(deserialize), model.eval(), \ub4dc\ub86d\uc544\uc6c3(dropout), \ubc30\uce58 \uc815\uaddc\ud654(batch normalization)<\/p>\n<div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>torch.save:<br \/>\ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\uc640 \uc635\ud2f0\ub9c8\uc774\uc800\uc758 \uc0c1\ud0dc\ub97c \ud3ec\ud568\ud558\uc5ec PyTorch \ubaa8\ub378\uc758 \uc0c1\ud0dc\ub97c \uc800\uc7a5\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \ud568\uc218<br \/>\ub098\uc911\uc5d0 torch.load \uae30\ub2a5\uc744 \uc0ac\uc6a9\ud558\uc5ec \ub85c\ub4dc\ud560 \uc218 \uc788\ub294 \ud30c\uc77c\uc5d0 \ubaa8\ub378\uc744 \uc800\uc7a5<br \/>\ub610\ud55c \uc635\ud2f0\ub9c8\uc774\uc800\uc758 state_dict\ub97c \uc800\uc7a5\ud558\ubbc0\ub85c \uc635\ud2f0\ub9c8\uc774\uc800\ub294 \uc911\ub2e8\ub41c \uc704\uce58\uc5d0\uc11c \ud559\uc2b5\uc744 \ub2e4\uc2dc \uc2dc\uc791\ud560 \uc218 \uc788\uc74c<br \/>\ubaa8\ub378\uc744 \ud568\uaed8 \uc800\uc7a5\ud560 \uc218 \uc788\uc73c\uba70, \uc774\ub54c\ub294 \ub9e4\uac1c\ubcc0\uc218\ub85c model\uc744 \uc785\ub825\ud558\uba74 \ub428<\/li>\n<li>model.state_dict():<br \/>\ubaa8\ub378\uc758 \ud559\uc2b5 \uac00\ub2a5\ud55c \ubaa8\ub4e0 \ub9e4\uac1c\ubcc0\uc218\ub97c \ud3ec\ud568\ud558\ub294 \uc815\ub82c\ub41c \uc0ac\uc804(ordered dictionary)\uc744 \ubc18\ud658\ud558\ub294 \ud568\uc218<br \/>\ub9ac\ud134 \uac12\uc778 \uc815\ub82c\ub41c \uc0ac\uc804\uc740 torch.save \ud568\uc218\ub97c \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc758 \ud604\uc7ac \uc0c1\ud0dc\ub97c \uc800\uc7a5\ud558\uac70\ub098 model.load_state_dict() \ud568\uc218\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc0ac\uc804 \ud6c8\ub828\ub41c \ubaa8\ub378\uc744 \ub85c\ub4dc\ud558\ub294 \ub370 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<br \/>\ud0a4: \ub9e4\uac1c\ubcc0\uc218\uc758 \uc774\ub984<br \/>\uac12: \ud574\ub2f9 \ub9e4\uac1c\ubcc0\uc218 \uac12\uc744 \ud3ec\ud568\ud558\ub294 \ud150\uc11c<\/li>\n<li>model.load_state_dict():<br \/>\uc774\uc804\uc5d0 \uc800\uc7a5\ub41c state_dict\uc5d0\uc11c \ubaa8\ub378\uc758 \uc0c1\ud0dc\ub97c \ub85c\ub4dc\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \ud568\uc218<br \/>state_dict\ub97c \uac00\uc838\uc640 \uc800\uc7a5\ub41c \ub9e4\uac1c\ubcc0\uc218 \uac12\uc744 \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\uc5d0 \ub9e4\ud551\ud568<\/li>\n<li>torch.load():<br \/>\ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\uc640 \uc635\ud2f0\ub9c8\uc774\uc800\uc758 \uc0c1\ud0dc\ub97c \ud3ec\ud568\ud558\uc5ec \uc774\uc804\uc5d0 \uc800\uc7a5\ub41c PyTorch \ubaa8\ub378\uc744 \ub85c\ub4dc\ud558\ub294 \ud568\uc218<br \/>\ud30c\uc77c \uacbd\ub85c\ub97c \uc778\uc218\ub85c \uc0ac\uc6a9\ud558\uace0 model.load_state_dict()\uc5d0 \uc804\ub2ec\ud560 \uc218 \uc788\ub294 \uc0ac\uc804\uc744 \ubc18\ud658<br \/>\ubaa8\ub378\uc744 \ud568\uaed8 \ubd88\ub7ec\uc62c \uc218 \uc788\uc73c\uba70, \uc774\ub54c\ub294 \ub9e4\uac1c\ubcc0\uc218\ub85c \ubaa8\ub378\uc774 \ud568\uaed8 \uc800\uc7a5\ub41c path\ub97c \uc785\ub825\ud558\uba74 \ub428<\/li>\n<li>pickle module:<br \/>Python dictionary \ubc0f list\uc640 \uac19\uc740 \uac1c\uccb4\ub97c \uc9c1\ub82c\ud654(serialize) \ubc0f \uc5ed\uc9c1\ub82c\ud654(deserialize)\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<br \/>PyTorch \ubaa8\ub378\uc744 \uc800\uc7a5\ud558\uace0 \ub85c\ub4dc\ud558\ub294 \ub370\uc5d0\ub3c4 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<br \/>torch.save \ubc0f torch.load \ud568\uc218\ubcf4\ub2e4 \ud6a8\uc728\uc131\uc774 \ub5a8\uc5b4\uc9c0\ubbc0\ub85c \uc0ac\uc6a9\ud558\uc9c0 \uc54a\ub294 \uac83\uc774 \uc88b\uc74c<br \/>--<br \/>\ud53c\ud074 \ubaa8\ub4c8\uc744 \uc0ac\uc6a9\ud558\uc5ec PyTorch \ubaa8\ub378\uc744 \uc800\uc7a5\ud558\uace0 \ub85c\ub4dc\ud558\ub294 \uac83\uc740 \uc77c\ubc18\uc801\uc73c\ub85c \uad8c\uc7a5\ub418\uc9c0 \uc54a\uc74c<br \/>\uc635\ud2f0\ub9c8\uc774\uc800\uc758 \ub0b4\ubd80 \uc0c1\ud0dc(internal state) \ubc0f \ubaa8\ub378\uc758 \uc5ed\ubc29\ud5a5 \uc804\ub2ec \uae30\ub85d(backward pass history of the model)\uacfc \uac19\uc774 \ubaa8\ub378\uc744 \uc62c\ubc14\ub974\uac8c \uc7ac\uad6c\uc131\ud558\ub294 \ub370 \ud544\uc694\ud55c \ubaa8\ub4e0 \uc815\ubcf4\ub97c \ucc98\ub9ac\ud558\uc9c0 \ubabb\ud560 \uc218 \uc788\uae30 \ub54c\ubb38<br \/>\ub300\uc2e0 PyTorch \ubaa8\ub378\uc744 \ucc98\ub9ac\ud558\ub3c4\ub85d \ud2b9\ubcc4\ud788 \uc124\uacc4\ub418\uc5c8\uc73c\uba70 \ubaa8\ub378\uc744 \uc62c\ubc14\ub974\uac8c \uc7ac\uad6c\uc131\ud558\ub294 \ub370 \ud544\uc694\ud55c \ubaa8\ub4e0 \uc815\ubcf4\ub97c \uc800\uc7a5\ud558\uace0 \ub85c\ub4dc\ud560 \uc218 \uc788\ub294 torch.save \ubc0f torch.load \ud568\uc218\ub97c \uc0ac\uc6a9\ud558\ub294 \uac83\uc774 \uc88b\uc74c<br \/>--<br \/>\uadf8\ub7ec\ub098 \uc774\uc804\uc5d0 pickle \ubaa8\ub4c8\uc744 \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc744 \uc800\uc7a5\ud55c \uacbd\uc6b0 pickle.load() \ud568\uc218\ub97c \uc0ac\uc6a9\ud558\uc5ec \uac1d\uccb4\ub97c \uc5ed\uc9c1\ub82c\ud654\ud558\uace0 \ubaa8\ub378\uacfc \ud574\ub2f9 \uc18d\uc131\uc744 \ub2e4\uc2dc \uac00\uc838\uc62c \uc218 \uc788\uc74c<br \/>\uadf8\ub7f0 \ub2e4\uc74c model.load_state_dict()\ub97c \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc758 \uc0c1\ud0dc \uc0ac\uc804(state dictionary)\uc744 \ub85c\ub4dc\ud558\uace0 optimizer.load_state_dict()\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc635\ud2f0\ub9c8\uc774\uc800\uc758 \uc0c1\ud0dc \uc0ac\uc804\uc744 \ub85c\ub4dc\ud558\uace0 \uc635\ud2f0\ub9c8\uc774\uc800\ub97c \ubaa8\ub378\ub85c \uc124\uc815\ud560 \uc218 \uc788\uc74c<br \/>pickle\ub85c \ubaa8\ub378\uc744 \uc800\uc7a5\ud55c \ud6c4 PyTorch \ubc84\uc804\uc774 \ubcc0\uacbd\ub418\uba74 \ub85c\ub4dc \ud504\ub85c\uc138\uc2a4\uac00 \uc81c\ub300\ub85c \uc791\ub3d9\ud558\uc9c0 \uc54a\uc744 \uc218 \uc788\ub2e4\ub294 \uc810 \uc720\uc758<\/li>\n<\/ul>\n<\/div>\n<\/div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc9c1\ub82c\ud654(serialize):<br \/>PyTorch \ubaa8\ub378\uacfc \uac19\uc740 \ubcf5\uc7a1\ud55c \uac1c\uccb4 \ub610\ub294 \ub370\uc774\ud130 \uad6c\uc870\ub97c \uc800\uc7a5\ud558\uac70\ub098 \uc804\uc1a1\ud560 \uc218 \uc788\ub294 \ud615\uc2dd\uc73c\ub85c \ubcc0\ud658\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>\uc774 \ud504\ub85c\uc138\uc2a4\ub97c \uac1c\uccb4 \"pickling\" \ub610\ub294 \"flattening\"\ub77c\uace0\ub3c4 \ud568<br \/>torch.save \ud568\uc218\ub294 PyTorch \ubaa8\ub378\uc744 \uc9c1\ub82c\ud654\ud558\ub294 \ub370 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<li>\uc5ed\uc9c1\ub82c\ud654(deserialize):<br \/>PyTorch \ubaa8\ub378\uacfc \uac19\uc740 \uc9c1\ub82c\ud654\ub41c \uac1c\uccb4 \ub610\ub294 \ub370\uc774\ud130 \uad6c\uc870\ub97c \ub2e4\uc2dc \uc6d0\ub798 \ud615\uc2dd\uc73c\ub85c \ubcc0\ud658\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>\uc774 \ud504\ub85c\uc138\uc2a4\ub97c \uac1c\uccb4 \"unpickling\" \ub610\ub294 \"inflating\"\uc774\ub77c\uace0\ub3c4 \ud798<br \/>torch.load \ud568\uc218\ub294 PyTorch \ubaa8\ub378\uc744 \uc5ed\uc9c1\ub82c\ud654\ud558\ub294 \ub370 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<li>model.eval():<br \/>\ubaa8\ub378\uc744 \ud3c9\uac00 \ubaa8\ub4dc\ub85c \uc124\uc815\ud558\ub294 \uba54\uc11c\ub4dc<br \/>\ud14c\uc2a4\ud2b8 \ub370\uc774\ud130\uc5d0 \ub300\ud55c \ubaa8\ub378\uc758 \uc131\ub2a5\uc744 \ud3c9\uac00\ud560 \ub54c \uc0ac\uc6a9\ub428<br \/>\ud3c9\uac00 \ubaa8\ub4dc\ub294 \ub4dc\ub86d\uc544\uc6c3 \ubc0f \ubc30\uce58 \uc815\uaddc\ud654\uc640 \uac19\uc740 \ud2b9\uc815 \uae30\ub2a5\uc744 \ud574\uc81c\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>\ub4dc\ub86d\uc544\uc6c3(dropout):<br \/>\ud6c8\ub828 \uc911\uc5d0 \uc785\ub825 \ub2e8\uc704\uc758 \uc77c\ubd80\ub97c \ubb34\uc791\uc704\ub85c 0\uc73c\ub85c \uc124\uc815\ud558\ub294 \uc815\uaddc\ud654 \uae30\uc220<br \/>\ubaa8\ub378\uc758 \ubcf5\uc7a1\uc131\uc744 \uc904\uc774\uace0 \ud65c\uc131\ud654\uc5d0 \uc57d\uac04\uc758 \ub178\uc774\uc988\ub97c \ub3c4\uc785\ud558\uc5ec \uacfc\uc801\ud569\uc744 \ubc29\uc9c0\ud558\ub294 \ub370 \ub3c4\uc6c0\uc744 \uc90c<br \/>\uc77c\ubc18\uc801\uc73c\ub85c \ud559\uc2b5 \uc911\uc5d0\ub9cc \uc0ac\uc6a9\ub418\uba70 \ud3c9\uac00 \ub610\ub294 \ucd94\ub860 \uc911\uc5d0\ub294 \uc0ac\uc6a9\ub418\uc9c0 \uc54a\uc74c<\/li>\n<li>\ubc30\uce58 \uc815\uaddc\ud654(batch normalization):<br \/>\ud3c9\uade0\uc744 \ube7c\uace0 \ud45c\uc900 \ud3b8\ucc28\ub85c \ub098\ub204\uc5b4 \uacc4\uce35\uc758 \ud65c\uc131\ud654(activations of layer)\ub97c \uc815\uaddc\ud654\ud558\ub294 \uae30\uc220<br \/>\ub0b4\ubd80 \uacf5\ubcc0\ub7c9 \ubcc0\ud654(internal covariate shift)\ub97c \uc904\uc5ec \ud6c8\ub828 \uacfc\uc815\uc744 \uc548\uc815\ud654\ud558\ub294 \ub370 \ub3c4\uc6c0\uc744 \uc90c<br \/>\uc77c\ubc18\uc801\uc73c\ub85c \ud6c8\ub828 \uc911\uc5d0\ub9cc \uc0ac\uc6a9\ub418\uba70 \ud3c9\uac00 \ub610\ub294 \ucd94\ub860 \uc911\uc5d0\ub294 \uc0ac\uc6a9\ub418\uc9c0 \uc54a\uc74c<\/li>\n<\/ul>\n<\/div>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/><\/div>\n<h3 data-ke-size=\"size23\">\ubaa8\ub378 \ucd94\ub860(Model Inference)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ubaa8\ub378 \ucd5c\uc801\ud654: \uc5b4\ub824\uc6b4 \uc791\uc5c5, \ub2e4\uc591\ud55c \ud504\ub808\uc784\uc6cc\ud06c \ubc0f \ud558\ub4dc\uc6e8\uc5b4\uc5d0\uc11c \uc131\ub2a5\uc744 \ucd5c\ub300\ud654\ud558\ub294 \ub370 \uc2dc\uac04\uc774 \ub9ce\uc774 \uc18c\uc694\ub428<\/li>\n<li>ONNX(Open Neural Network Exchange) \ub7f0\ud0c0\uc784: \ubaa8\ub4e0 \ud558\ub4dc\uc6e8\uc5b4, \ud074\ub77c\uc6b0\ub4dc \ub610\ub294 \uc5d0\uc9c0 \uc7a5\uce58\uc5d0\uc11c \ud55c \ubc88 \ud6c8\ub828\ud558\uace0 \ucd94\ub860\uc744 \uac00\uc18d\ud654\ud558\ub294 \uc194\ub8e8\uc158<\/li>\n<li>ONNX: \uc2e0\uacbd\ub9dd \ubc0f \uae30\ud0c0 \uae30\uacc4 \ud559\uc2b5 \ubaa8\ub378\uc744 \uacf5\uc720\ud558\uae30 \uc704\ud574 \uc5ec\ub7ec \uacf5\uae09\uc5c5\uccb4\uc5d0\uc11c \uc9c0\uc6d0\ud558\ub294 \uacf5\ud1b5 \ud615\uc2dd<\/li>\n<li>ONNX \uc774\uc810: Java, JavaScript, C# \ubc0f ML.NET\uacfc \uac19\uc740 \ub2e4\ub978 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4 \ubc0f \ud504\ub808\uc784\uc6cc\ud06c\uc5d0\uc11c \ubaa8\ub378\ub85c \ucd94\ub860 \uac00\ub2a5<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\ub2e4\ub978 \uc608\uc81c \ucf54\ub4dc<\/p>\n<pre id=\"code_1674233858922\" class=\"haskell\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>%matplotlib inline\nimport torch\nimport onnxruntime\nfrom torch import nn\nimport torch.onnx as onnx\nimport torchvision.models as models\nfrom torchvision import datasets\nfrom torchvision.transforms import ToTensor\n\nclass NeuralNetwork(nn.Module):\n    def __init__(self):\n        super(NeuralNetwork, self).__init__()\n        self.flatten = nn.Flatten()\n        self.linear_relu_stack = nn.Sequential(\n            nn.Linear(28*28, 512),\n            nn.ReLU(),\n            nn.Linear(512, 512),\n            nn.ReLU(),\n            nn.Linear(512, 10),\n            nn.ReLU()\n        )\n\n    def forward(self, x):\n        x = self.flatten(x)\n        logits = self.linear_relu_stack(x)\n        return logits\n\n\nmodel = NeuralNetwork()\nmodel.load_state_dict(torch.load('data\/model.pth'))\nmodel.eval()<\/code><\/pre>\n<div>\n<div>\n<div>\n<div>\n<div>&nbsp;<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<div>\n<div>\n<div>\n<div>\n<div>keyword: %matplotlib inline, Open Neural Network Exchange (ONNX), onnxruntime, torch.onnx, <span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">Java, JavaScript, C#, ML.NET<\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>%matplotlib inline:<br \/>Jupyter \ub178\ud2b8\ubd81\uc5d0\uc11c matplotlib \ub77c\uc774\ube0c\ub7ec\ub9ac\ub85c \uc0dd\uc131\ub41c \ud50c\ub86f\uc744 \ubcc4\ub3c4\uc758 \ucc3d\uc774 \uc544\ub2cc \ub178\ud2b8\ubd81 \ub0b4\uc5d0\uc11c \ud45c\uc2dc\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \uba85\ub839<br \/>\uc77c\ubc18\uc801\uc73c\ub85c \uc2dc\uac04 \uacbd\uacfc\uc5d0 \ub530\ub978 \uc190\uc2e4 \ub610\ub294 \uc815\ud655\ub3c4\uc640 \uac19\uc740 \ubaa8\ub378\uc758 \uc131\ub2a5\uc744 \uc2dc\uac01\ud654\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>ONNX(Open Neural Network Exchange):<br \/>\ub525 \ub7ec\ub2dd \ubaa8\ub378\uc744 \ub098\ud0c0\ub0b4\ub294 \uac1c\ubc29\ud615 \ud615\uc2dd<br \/>PyTorch \ubc0f TensorFlow\uc640 \uac19\uc740 \uc11c\ub85c \ub2e4\ub978 \ud504\ub808\uc784\uc6cc\ud06c \uac04\uc758 \uc0c1\ud638 \uc6b4\uc6a9\uc131\uc744 \ud5c8\uc6a9\ud558\uace0 \ubaa8\ub378\uc744 \ub0b4\ubcf4\ub0b4\uace0 \uac00\uc838\uc62c \uc218 \uc788\uc74c<\/li>\n<li>onnxruntime: <br \/>ONNX \ubaa8\ub378\uc744 \uc704\ud55c \uace0\uc131\ub2a5 \ucd94\ub860 \uc5d4\uc9c4<br \/>\ub2e4\uc591\ud55c \ud50c\ub7ab\ud3fc\uacfc \ud558\ub4dc\uc6e8\uc5b4\ub97c \uc9c0\uc6d0\ud558\uba70 \ubaa8\ub378\uc758 \ucd94\ub860 \ud504\ub85c\uc138\uc2a4\ub97c \uac00\uc18d\ud654\ud558\ub294 \ub370 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<li>torch.onnx:<br \/>\ubaa8\ub378\uc744 ONNX \ud615\uc2dd\uc73c\ub85c \ub0b4\ubcf4\ub0bc \uc218 \uc788\ub294 PyTorch \ud328\ud0a4\uc9c0<br \/>PyTorch \ubaa8\ub378\uc744 ONNX\ub85c \ubcc0\ud658\ud558\uae30 \uc704\ud55c \uac04\ub2e8\ud55c API\ub97c \uc81c\uacf5<br \/>\ubaa8\ub378\uc744 ONNX\ub97c \uc9c0\uc6d0\ud558\ub294 \ub2e4\ub978 \ud504\ub808\uc784\uc6cc\ud06c \ub610\ub294 \ud50c\ub7ab\ud3fc\uc73c\ub85c \uc27d\uac8c \ub0b4\ubcf4\ub0bc \uc218 \uc788\uc74c<\/li>\n<li>Java: <br \/>\ubaa8\ubc14\uc77c \ubc0f \uc6f9 \uc560\ud50c\ub9ac\ucf00\uc774\uc158\uc5d0\uc11c \uc5d4\ud130\ud504\ub77c\uc774\uc988 \uc18c\ud504\ud2b8\uc6e8\uc5b4 \ubc0f \uacfc\ud559 \uc2dc\ubbac\ub808\uc774\uc158\uc5d0 \uc774\ub974\uae30\uae4c\uc9c0 \uad11\ubc94\uc704\ud55c \uc560\ud50c\ub9ac\ucf00\uc774\uc158\uc744 \uac1c\ubc1c\ud558\ub294 \ub370 \ub110\ub9ac \uc0ac\uc6a9\ub418\ub294 \ub110\ub9ac \uc0ac\uc6a9\ub418\ub294 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4<br \/>Weka, Deeplearning4j \ubc0f MLlib\uc640 \uac19\uc740 \uae30\uacc4 \ud559\uc2b5\uc5d0 \uc0ac\uc6a9\ud560 \uc218 \uc788\ub294 \ub77c\uc774\ube0c\ub7ec\ub9ac \ubc0f \ud504\ub808\uc784\uc6cc\ud06c\uc758 \ub300\uaddc\ubaa8 \uc5d0\ucf54\uc2dc\uc2a4\ud15c \uc874\uc7ac<\/li>\n<li>JavaScript:<br \/>\uc8fc\ub85c \uc6f9 \uc560\ud50c\ub9ac\ucf00\uc774\uc158 \uac1c\ubc1c\uc5d0 \uc0ac\uc6a9\ub418\ub294 \uc778\uae30 \uc788\ub294 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4<br \/>TensorFlow.js, Brain.js \ubc0f ML.js\uc640 \uac19\uc740 \ub77c\uc774\ube0c\ub7ec\ub9ac\uc758 \ub3c4\uc6c0\uc73c\ub85c \uae30\uacc4 \ud559\uc2b5 \uc560\ud50c\ub9ac\ucf00\uc774\uc158 \uac1c\ubc1c\uc5d0\ub3c4 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<li>C#:<br \/>\uc8fc\ub85c Windows \ub370\uc2a4\ud06c\ud1b1 \uc751\uc6a9 \ud504\ub85c\uadf8\ub7a8, \ubaa8\ubc14\uc77c \uc571 \ubc0f \uac8c\uc784 \uac1c\ubc1c\uc5d0 \uc0ac\uc6a9\ub418\ub294 \ub110\ub9ac \uc0ac\uc6a9\ub418\ub294 \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4<br \/>Accord.NET, ML.NET \ubc0f CNTK\uc640 \uac19\uc740 \ub77c\uc774\ube0c\ub7ec\ub9ac\uc758 \ub3c4\uc6c0\uc73c\ub85c \uae30\uacc4 \ud559\uc2b5 \uc751\uc6a9 \ud504\ub85c\uadf8\ub7a8 \uac1c\ubc1c\uc5d0\ub3c4 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<li>ML.NET: <br \/>.NET \uac1c\ubc1c\uc790\ub97c \uc704\ud55c \uc624\ud508 \uc18c\uc2a4 \uae30\uacc4 \ud559\uc2b5 \ud504\ub808\uc784\uc6cc\ud06c<br \/>\uae30\uacc4 \ud559\uc2b5\uc5d0 \ub300\ud55c \uc804\ubb38 \uc9c0\uc2dd \uc5c6\uc774\ub3c4 C# \ub610\ub294 F#\uc744 \uc0ac\uc6a9\ud558\uc5ec \uae30\uacc4 \ud559\uc2b5 \ubaa8\ub378\uc744 \uac1c\ubc1c\ud560 \uc218 \uc788\uc74c<br \/>\ucd08\ubcf4\uc790\uc640 \uc804\ubb38\uac00 \ubaa8\ub450\ub97c \uc704\ud574 \uc124\uacc4\ub418\uc5c8\uc73c\uba70 \ubd84\ub958, \ud68c\uadc0 \ubc0f \uc774\uc0c1 \ud0d0\uc9c0\uc640 \uac19\uc740 \uad11\ubc94\uc704\ud55c \uae30\uacc4 \ud559\uc2b5 \uc791\uc5c5\uc744 \uc9c0\uc6d0<\/li>\n<\/ul>\n<div>\n<div>&nbsp;<\/div>\n<div>&nbsp;<\/div>\n<\/div>\n<h3 data-ke-size=\"size23\">\ubaa8\ub378\uc744 ONNX\ub85c \ub0b4\ubcf4\ub0b4\uae30(Exporting the model to ONNX)<\/h3>\n<div>\n<div>\n<div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>PyTorch: \uae30\ubcf8 ONNX \ub0b4\ubcf4\ub0b4\uae30 \uc9c0\uc6d0<\/li>\n<li>PyTorch \uc2e4\ud589 \uadf8\ub798\ud504\uc758 \ub3d9\uc801 \ud2b9\uc131(<span style=\"color: #374151;\">Dynamic nature of PyTorch execution graph)<\/span>:<br \/>\ub0b4\ubcf4\ub0b4\uae30 \ud504\ub85c\uc138\uc2a4\ub294 \uc720\uc9c0\ub418\ub294(persisted) ONNX \ubaa8\ub378\uc744 \uc0dd\uc131\ud558\uae30 \uc704\ud574 \uc2e4\ud589 \uadf8\ub798\ud504\ub97c \ud1b5\uacfc\ud574\uc57c \ud568<\/li>\n<li>\ud14c\uc2a4\ud2b8 \ubcc0\uc218: \ub0b4\ubcf4\ub0b4\uae30 \ub8e8\ud2f4\uc5d0 \uc804\ub2ec, \uc62c\ubc14\ub978 \ud06c\uae30\uc758 \ub354\ubbf8 \uc81c\ub85c \ud150\uc11c(\ud6c8\ub828 \ub370\uc774\ud130 \uc138\ud2b8\uc758 shape\uc5d0\uc11c \uc5bb\uc744 \uc218 \uc788\uc74c)<\/li>\n<\/ul>\n<pre id=\"code_1674235747666\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>input_image = torch.zeros((1,28,28))\nonnx_model = 'data\/model.onnx'\nonnx.export(model, input_image, onnx_model)<\/code><\/pre>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud14c\uc2a4\ud2b8 \ub370\uc774\ud130 \uc138\ud2b8: \uc608\uce21\uc744 \uc704\ud574 ONNX \ubaa8\ub378\uc5d0\uc11c \ucd94\ub860\ud558\uae30 \uc704\ud55c \uc0d8\ud50c \ub370\uc774\ud130\ub85c \uc0ac\uc6a9<\/li>\n<li>\ucd94\ub860 \uc138\uc158: onnxruntime.InferenceSession\uc73c\ub85c \uc0dd\uc131\ub428<\/li>\n<li>\ucd94\ub860: \uc6d0\ud558\ub294 \ucd9c\ub825 \ubaa9\ub85d\uacfc \uc785\ub825 \uac12 \ub9f5\uc744 \uc804\ub2ec\ud558\uc5ec run() \uba54\uc11c\ub4dc\ub97c \uc0ac\uc6a9\ud558\uc5ec ONNX \ubaa8\ub378\uc5d0\uc11c \uc218\ud589\ub428<\/li>\n<li>\uacb0\uacfc: \ucd9c\ub825 \ubaa9\ub85d<\/li>\n<\/ul>\n<pre id=\"code_1674234989506\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>session = onnxruntime.InferenceSession(onnx_model, None)\ninput_name = session.get_inputs()[0].name\noutput_name = session.get_outputs()[0].name\n\nresult = session.run([output_name], {input_name: x.numpy()})\npredicted, actual = classes[result[0][0].argmax(0)], classes[y]\nprint(f'Predicted: \"{predicted}\", Actual: \"{actual}\"')<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>ONNX \ubaa8\ub378: \ub2e4\uc591\ud55c \ud50c\ub7ab\ud3fc\uacfc \ub2e4\uc591\ud55c \ud504\ub85c\uadf8\ub798\ubc0d \uc5b8\uc5b4\uc5d0\uc11c \ucd94\ub860\uc744 \uc2e4\ud589\ud558\ub294 \ub370 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">keyword: \uc2e4\ud589 \uadf8\ub798\ud504(execution graph), ONNX \ub0b4\ubcf4\ub0b4\uae30(ONNX export), Persist, model.onnx, onnx.export, onnxruntime.InferenceSession, onnxruntime.InferenceSession.get_inputs(), onnxruntime.InferenceSession.run()<\/p>\n<\/div>\n<\/div>\n<\/div>\n<div>\n<div>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc2e4\ud589 \uadf8\ub798\ud504(execution graph):<br \/>\ubaa8\ub378\uc774 \uc218\ud589\ud558\ub294 \uacc4\uc0b0\uc744 \uadf8\ub798\ud504 \uae30\ubc18\uc73c\ub85c \ud45c\ud604\ud55c \uac83<br \/>\uc785\ub825\uc5d0 \ub300\ud574 \uc218\ud589\ub418\ub294 \uc791\uc5c5\uacfc \uc2e4\ud589 \uc21c\uc11c\ub97c \uc124\uba85\ud568<br \/>\ubaa8\ub378\uc758 \uc131\ub2a5\uc744 \ucd5c\uc801\ud654\ud558\uace0 \ubaa8\ub378\uc744 \ub2e4\ub978 \ud504\ub808\uc784\uc6cc\ud06c\ub098 \ud50c\ub7ab\ud3fc\uc73c\ub85c \ub0b4\ubcf4\ub0bc \uc218 \uc788\ub3c4\ub85d \ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>ONNX \ub0b4\ubcf4\ub0b4\uae30(ONNX export):<br \/>\ud558\ub098\uc758 \ud504\ub808\uc784\uc6cc\ud06c\uc5d0\uc11c \ub525 \ub7ec\ub2dd \ubaa8\ub378\uc744 \ud45c\ud604\ud558\uae30 \uc704\ud55c \uac1c\ubc29\ud615 \ud615\uc2dd\uc778 ONNX \ud615\uc2dd\uc73c\ub85c \ubaa8\ub378\uc744 \ubcc0\ud658\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>\uc774\ub97c \ud1b5\ud574 PyTorch \ubc0f TensorFlow\uc640 \uac19\uc740 \uc11c\ub85c \ub2e4\ub978 \ud504\ub808\uc784\uc6cc\ud06c \uac04\uc758 \uc0c1\ud638 \uc6b4\uc6a9\uc131\uc744 \ud5c8\uc6a9\ud558\uace0 \ubaa8\ub378\uc744 \ub0b4\ubcf4\ub0b4\uace0 \uac00\uc838\uc62c \uc218 \uc788\uc74c<br \/>\uc77c\ubc18\uc801\uc73c\ub85c PyTorch\uc758 torch.onnx \ud328\ud0a4\uc9c0\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc218\ud589\ub428<\/li>\n<li>Persist:<br \/>\ub098\uc911\uc5d0 \ub85c\ub4dc\ud558\uc5ec \uc0ac\uc6a9\ud560 \uc218 \uc788\ub3c4\ub85d \ubaa8\ub378\uc758 \uc0c1\ud0dc\ub97c \ud30c\uc77c\uc5d0 \uc800\uc7a5\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>\uc774\ub97c \ud1b5\ud574 \ud559\uc2b5 \ud6c4 \ubaa8\ub378\uc744 \uc800\uc7a5\ud558\uc5ec \ucd94\ub860 \ub610\ub294 \ucd94\uac00 \ud559\uc2b5\uc5d0 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<br \/>PyTorch\uc5d0\uc11c torch.save() \ubc0f torch.load() \ud568\uc218\ub97c \uc0ac\uc6a9\ud558\uac70\ub098 \ubaa8\ub378\uc5d0\uc11c state_dict() \ubc0f load_state_dict() \uba54\uc11c\ub4dc\ub97c \uc0ac\uc6a9\ud558\uc5ec state_dict\ub97c \uc800\uc7a5 \ubc0f \ub85c\ub4dc\ud568\uc73c\ub85c\uc368 \uc218\ud589\ud560 \uc218 \uc788\uc74c<\/li>\n<\/ol>\n<div>\n<div>\n<div>\n<div>\n<div>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>model.onnx: <br \/>\ud6c8\ub828\ub41c \ubaa8\ub378\uc744 ONNX \ud615\uc2dd\uc73c\ub85c \uc800\uc7a5\ud558\uae30 \uc704\ud55c \ud30c\uc77c \ud615\uc2dd<br \/>\ub525 \ub7ec\ub2dd \ubaa8\ub378\uc744 \ud45c\ud604\ud558\uae30 \uc704\ud55c \uac1c\ubc29\ud615 \ud615\uc2dd\uc73c\ub85c, PyTorch \ubc0f TensorFlow\uc640 \uac19\uc740 \uc11c\ub85c \ub2e4\ub978 \ud504\ub808\uc784\uc6cc\ud06c \uac04\uc758 \uc0c1\ud638 \uc6b4\uc6a9\uc131\uc744 \ud5c8\uc6a9<\/li>\n<li>onnx.export:<br \/>PyTorch\uc758 torch.onnx \ud328\ud0a4\uc9c0\uc5d0\uc11c \uc81c\uacf5\ud558\ub294 \ud568\uc218<br \/>PyTorch \ubaa8\ub378\uc744 ONNX \ud615\uc2dd\uc73c\ub85c \ub0b4\ubcf4\ub0bc \uc218 \uc788\uc74c<br \/>&nbsp;\ubaa8\ub378, \uc608\uc81c \uc785\ub825 \ud150\uc11c \ubc0f \ud30c\uc77c \uacbd\ub85c\ub97c \uc778\uc218\ub85c \uc0ac\uc6a9<br \/>\ubaa8\ub378\uc744 \uc9c0\uc815\ub41c \ud30c\uc77c \uacbd\ub85c\ub85c \ub0b4\ubcf4\ub0c4<br \/>\uc81c\uacf5\ud55c \ucf54\ub4dc \uc2a4\ub2c8\ud3ab\uc5d0\uc11c \ubaa8\ub378\uc740 'data\/model.onnx' \ud30c\uc77c\ub85c \ub0b4\ubcf4\ub0b4\uc9c0\uba70 \uc785\ub825 \ud150\uc11c\ub294 \ubaa8\uc591\uc774 (1, 28, 28)\uc774\uace0 0\uc73c\ub85c \ucc44\uc6cc\uc838 \uc788\u3147,\u3141<\/li>\n<\/ol>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<div>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>onnxruntime.InferenceSession:<br \/>ONNX \ubaa8\ub378 \uc2e4\ud589\uc744 \uc704\ud55c \uc138\uc158\uc744 \ub098\ud0c0\ub0b4\ub294 ONNX \ub7f0\ud0c0\uc784 \ub77c\uc774\ube0c\ub7ec\ub9ac\uc758 \ud074\ub798\uc2a4<br \/>ONNX \ubaa8\ub378\uc744 \ub85c\ub4dc\ud558\uace0, \ucd94\ub860\uc744 \uc704\ud574 \uc900\ube44\ud558\uace0, \ubaa8\ub378\uc5d0\uc11c \ucd94\ub860\uc744 \uc2e4\ud589\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<br \/>ONNX \ubaa8\ub378 \ud30c\uc77c\uc758 \uacbd\ub85c\ub97c \uac00\uc838\uc640\uc11c \ubaa8\ub378\uc5d0\uc11c \ucd94\ub860\uc744 \uc2e4\ud589\ud558\ub294 \ub370 \uc0ac\uc6a9\ud560 \uc218 \uc788\ub294 \uc138\uc158 \uac1c\uccb4\ub97c \ub9cc\ub4ec<\/li>\n<li>onnxruntime.InferenceSession.get_inputs(): <br \/>\ubaa8\ub378\uc758 \uc785\ub825 \ud150\uc11c \ubaa9\ub85d\uc744 \ubc18\ud658\ud558\ub294 onnxruntime.InferenceSession \ud074\ub798\uc2a4\uc758 \uba54\uc11c\ub4dc<br \/>\ubaa8\ub378\uc758 \uc785\ub825 \ud150\uc11c\ub97c \uac00\uc838\uc624\uace0 \uc62c\ubc14\ub978 \ud615\uc2dd\uc73c\ub85c \uc785\ub825 \ub370\uc774\ud130\ub97c \uc900\ube44\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>onnxruntime.InferenceSession.run():<br \/>\ubaa8\ub378\uc5d0\uc11c \ucd94\ub860\uc744 \uc2e4\ud589\ud558\ub294 onnxruntime.InferenceSession \ud074\ub798\uc2a4\uc758 \uba54\uc11c\ub4dc<br \/>\uc785\ub825 \ubc0f \ucd9c\ub825 \ud150\uc11c\uc758 \uc774\ub984\uacfc \uc785\ub825 \ub370\uc774\ud130\ub97c \ubc1b\uc544 \ucd9c\ub825 \ub370\uc774\ud130\ub97c \ubc18\ud658<br \/>\uc8fc\uc5b4\uc9c4 \uc785\ub825 \ub370\uc774\ud130\ub85c \ubaa8\ub378\uc5d0 \ub300\ud55c \ucd94\ub860\uc744 \uc218\ud589\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<br \/>\ubaa8\ub378\uc5d0\uc11c \uc0dd\uc131\ub41c \ucd9c\ub825 \ub370\uc774\ud130\ub97c \ubc18\ud658<\/li>\n<\/ol>\n<\/div>\n<\/div>\n<\/div>\n<\/div>","category":"Pytorch\/\ud29c\ud1a0\ub9ac\uc5bc","author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/133","comments":"https:\/\/hoohaha.tistory.com\/133#entry133comment","pubDate":"Sat, 21 Jan 2023 02:47:59 +0900"},{"title":"[PyTorch] \uacf5\uc2dd \ubb38\uc11c Learn the Basics \uc694\uc57d - 6. Optimization","link":"https:\/\/hoohaha.tistory.com\/132","description":"<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"444\" data-origin-height=\"246\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/bpXyYO\/btrWP9zya9c\/iDeaGeONTsY5eMY6e4cEs0\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/bpXyYO\/btrWP9zya9c\/iDeaGeONTsY5eMY6e4cEs0\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/bpXyYO\/btrWP9zya9c\/iDeaGeONTsY5eMY6e4cEs0\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FbpXyYO%2FbtrWP9zya9c%2FiDeaGeONTsY5eMY6e4cEs0%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"444\" height=\"246\" data-origin-width=\"444\" data-origin-height=\"246\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\">\ubaa9\ucc28<\/p>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\">\ubaa8\ub378 \ub9e4\uac1c\ubcc0\uc218 \ucd5c\uc801\ud654(Optimizing the model parameters)<\/p>\n<p data-ke-size=\"size16\">\uae30\ubcf8 \ucf54\ub4dc(Prerequisite code)<\/p>\n<p data-ke-size=\"size16\">\ud558\uc774\ud37c\ud30c\ub77c\ubbf8\ud130 \uc124\uc815(Setting hyperparameters)<\/p>\n<p data-ke-size=\"size16\">\ucd5c\uc801\ud654 \ub8e8\ud504 \ucd94\uac00(Add an optimization loop)<\/p>\n<p data-ke-size=\"size16\">\uc190\uc2e4 \ud568\uc218 \ucd94\uac00(Add a loss function)<\/p>\n<p data-ke-size=\"size16\">\ucd5c\uc801\ud654 \ud328\uc2a4(Optimization pass)<\/p>\n<p data-ke-size=\"size16\">\uc804\uccb4 \uad6c\ud604(Full implementation)<\/p>\n<\/div>\n<\/div>\n<p data-ke-size=\"size16\"><span style=\"background-color: #fafafa;\">keyword:<\/span><\/p>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\">\ud559\uc2b5\ub960(learning rate), \ucd5c\uc0c1\uc758 \uac00\uc911\uce58(best weights), \uc608\uce21\ud560 \uc218 \uc5c6\ub294 \ub3d9\uc791(unpredictable behavior), \uc9c0\uc5ed \ucd5c\uc18c\uac12(local minima), \uc804\uc5ed \ucd5c\uc18c\uac12(global minima), \ucd5c\uc801\ud654 \ub8e8\ud504(optimization loop), \ud559\uc2b5 \ub8e8\ud504(train loop), \uac80\uc99d \ub8e8\ud504(validation loop), \ud14c\uc2a4\ud2b8 \ub8e8\ud504(test loop), \uc190\uc2e4 \uacc4\uc0b0(loss calculation), \uc635\ud2f0\ub9c8\uc774\uc800 \ub2e8\uacc4(optimizer step), \ud3c9\uade0 \uc81c\uacf1 \uc624\ucc28(Mean Square Error, MSE), \uc74c\uc758 \ub85c\uadf8 \uc6b0\ub3c4(Negative Log Likelihood, NLL), \ud06c\ub85c\uc2a4\uc5d4\ud2b8\ub85c\ud53c \uc190\uc2e4(Cross-Entropy Loss, CE), \ub85c\uc9d3(logit), \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998(optimization algorithms), \uc635\ud2f0\ub9c8\uc774\uc800 \uac1d\uccb4(optimizer object), SGD(Stochastic Gradient Descent), ADAM, RMSProp, torch.save, \ub0b4\ubd80 \uc0c1\ud0dc \uc0ac\uc804(internal state dictionary), state_dict()<\/p>\n<\/div>\n<\/div>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h2 data-ke-size=\"size26\">\ubaa8\ub378 \ub9e4\uac1c\ubcc0\uc218 \ucd5c\uc801\ud654(Optimizing the model parameters)<\/h2>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ubaa8\ub378: \ud6c8\ub828, \uac80\uc99d \ubc0f \ud14c\uc2a4\ud2b8\ub428<\/li>\n<li>\ub370\uc774\ud130: \ucd5c\uc801\ud654\uc5d0 \uc0ac\uc6a9<\/li>\n<li>\ud559\uc2b5: \ubc18\ubcf5\uc801\uc778 \uacfc\uc815 \uac70\uce68<\/li>\n<li>\ubc18\ubcf5(iteration): epoch(\uc5d0\ud3ed)\ub77c\uace0 \ud568<\/li>\n<li>\uac01 \ubc18\ubcf5\uc5d0\uc11c\uc758 \ubaa8\ub378\uc758 \ud559\uc2b5: <br \/>\ucd9c\ub825\uc5d0 \ub300\ud574 \ucd94\uce21<br \/>\uc624\ub958(\ucd94\uce21\uacfc \uc815\ub2f5 \uc0ac\uc774\uc758 \ucc28\uc774(\uc190\uc2e4, loss)) \uacc4\uc0b0<br \/>\ub9e4\uac1c\ubcc0\uc218\uc640 \uad00\ub828\ub41c \uc624\ub958\uc758 <span>\ub3c4\ud568\uc218(derivative) \uc218\uc9d1<br \/>\uacbd\uc0ac\ud558\uac15\ubc95(Gradient Descent) \uc0ac\uc6a9\ud558\uc5ec \ub9e4\uac1c\ubcc0\uc218 \ucd5c\uc801\ud654(optimize)<\/span><\/li>\n<\/ul>\n<p data-ke-size=\"size16\">keyword: model optimization<\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ubaa8\ub378 \ucd5c\uc801\ud654(model optimization):<br \/>\uc190\uc2e4 \ud568\uc218\uc640 \uac19\uc740 \uc2a4\uce7c\ub77c \uac12 \ud568\uc218\ub97c \ucd5c\uc18c\ud654\ud558\uae30 \uc704\ud574 \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc870\uc815\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>\uc190\uc2e4 \ud568\uc218\uc758 \ubcc0\ud654\ub3c4\ub97c \uae30\ubc18\uc73c\ub85c \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \uc54c\uace0\ub9ac\uc998\uc778 \uc635\ud2f0\ub9c8\uc774\uc800\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc218\ud589\ub428<br \/>\ubaa9\ud45c\ub294 \ucd5c\uc0c1\uc758 \ubaa8\ub378 \uc131\ub2a5\uc744 \ub098\ud0c0\ub0b4\ub294 \uac00\ub2a5\ud55c \ucd5c\uc800 \uc190\uc2e4(lowest possible loss)\uc744 \ucd08\ub798\ud558\ub294 \ub9e4\uac1c\ubcc0\uc218 \uc9d1\ud569\uc744 \ucc3e\ub294 \uac83<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\uae30\ubcf8 \ucf54\ub4dc(Prerequisite code)<\/h3>\n<pre id=\"code_1674191797769\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>%matplotlib inline\nimport torch\nfrom torch import nn\nfrom torch.utils.data import DataLoader\nfrom torchvision import datasets\nfrom torchvision.transforms import ToTensor, Lambda\n\ntraining_data = datasets.FashionMNIST(\n    root=\"data\",\n    train=True,\n    download=True,\n    transform=ToTensor()\n)\n\ntest_data = datasets.FashionMNIST(\n    root=\"data\",\n    train=False,\n    download=True,\n    transform=ToTensor()\n)\n\ntrain_dataloader = DataLoader(training_data, batch_size=64)\ntest_dataloader = DataLoader(test_data, batch_size=64)\n\nclass NeuralNetwork(nn.Module):\n    def __init__(self):\n        super(NeuralNetwork, self).__init__()\n        self.flatten = nn.Flatten()\n        self.linear_relu_stack = nn.Sequential(\n            nn.Linear(28*28, 512),\n            nn.ReLU(),\n            nn.Linear(512, 512),\n            nn.ReLU(),\n            nn.Linear(512, 10),\n            nn.ReLU()\n        )\n\n    def forward(self, x):\n        x = self.flatten(x)\n        logits = self.linear_relu_stack(x)\n        return logits\n\nmodel = NeuralNetwork()<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ud558\uc774\ud37c\ud30c\ub77c\ubbf8\ud130 \uc124\uc815(Setting hyperparameters)<\/h3>\n<p data-ke-size=\"size16\">\ud558\uc774\ud37c\ud30c\ub77c\ubbf8\ud130<\/p>\n<div>\n<div>\n<div>\n<div>\n<div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc815\uc758: \ubaa8\ub378 \ucd5c\uc801\ud654 \uacfc\uc815\uc744 \uc81c\uc5b4\ud558\ub294 \u200b\u200b\uc870\uc808 \uac00\ub2a5\ud55c \ub9e4\uac1c\ubcc0\uc218<\/li>\n<li>\uc601\ud5a5(impact): \ubaa8\ub378 \ud559\uc2b5 \ubc0f \uc815\ud655\ub3c4 \uc218\uc900, \uc218\ub834\uc728(convergence rate)<\/li>\n<li>\ud559\uc2b5\uc6a9\uc73c\ub85c \uc815\uc758\ub41c \ud558\uc774\ud37c\ud30c\ub77c\ubbf8\ud130:\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>Epoch \uc218: \uc804\uccb4 \uad50\uc721 \ub370\uc774\ud130 \uc138\ud2b8\uac00 \ub124\ud2b8\uc6cc\ud06c\ub97c \ud1b5\uacfc\ud558\ub294 \ud69f\uc218<\/li>\n<li>batch size: \ub9e4\uac1c\ubcc0\uc218\uac00 \uac31\uc2e0\ub418\uae30 \uc804 \uc2e0\uacbd\ub9dd\uc744 \ud1b5\ud574 \uc804\ud30c\ub41c \ub370\uc774\ud130 \uc0d8\ud50c\uc758 \uc218<\/li>\n<li>iterations(\ubc18\ubcf5\uc218): \ud55c \uc5d0\ud3ed\ub97c \uc644\ub8cc\ud558\ub294 \ub370 \ud544\uc694\ud55c \ubc30\uce58 \uc218(\ud559\uc2b5 \ub370\uc774\ud130 \uc218 \/ \ubc30\uce58 \uc0ac\uc774\uc988)<\/li>\n<li>\ud559\uc2b5\ub960(learning rate): <br \/>\uac01 \ubc30\uce58\/\uc5d0\ud3ed\uc5d0\uc11c \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc870\uc808\ud558\ub294 \ube44\uc728<br \/>\ub354 \uc791\uc740 \ud559\uc2b5\ub960: \ud559\uc2b5 \uc18d\ub3c4 \ub290\ub824\uc9d0. \ubaa8\ub378\uc774 \ucd5c\uc0c1\uc758 \uac00\uc911\uce58\ub97c \ucc3e\ub294 \ub370 \ub354 \uc624\ub798 \uac78\ub9bc. local minima\uc5d0 \ube60\uc9c8 \uc218 \uc788\uc74c<br \/>\ub354 \ud070 \ud559\uc2b5\ub960: \ubaa8\ub378\uc774 \uac74\ub108\ub6f0\uace0 \ucd5c\uc0c1\uc758 \uac00\uc911\uce58\ub97c \ub193\uce60 \uc218 \uc788\uc73c\uba70 \ud559\uc2b5 \uc911\uc5d0 \uc608\uce21\ud560 \uc218 \uc5c6\ub294 \ub3d9\uc791\uc774 \ubc1c\uc0dd\ud560 \uc218 \uc788\uc74c<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<pre id=\"code_1674194629264\" class=\"ini\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>learning_rate = 1e-3\nbatch_size = 64\nepochs = 5<\/code><\/pre>\n<p data-ke-size=\"size16\">keyword: learning rate, best weights, unpredictable behavior, local minima, global minima<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<div>\n<div>\n<div>\n<div>\n<div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud559\uc2b5\ub960(learning rate): <br \/>\ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998\uc758 \ub2e8\uacc4 \ud06c\uae30\ub97c \uc81c\uc5b4\ud558\ub294 \u200b\u200b\ud558\uc774\ud37c\ud30c\ub77c\ubbf8\ud130<br \/>\uac01 \ubc18\ubcf5\uc5d0\uc11c \ub9e4\uac1c\ubcc0\uc218\uac00 \uc5c5\ub370\uc774\ud2b8\ub418\ub294 \uc591\uc744 \uacb0\uc815<br \/>\ud559\uc2b5\ub960\uc774 \ub192\uc73c\uba74 \uc5c5\ub370\uc774\ud2b8\uac00 \ucee4\uc9c0\uace0 \uc218\ub834\uc774 \ube68\ub77c\uc9c0\uc9c0\ub9cc \ubaa8\ub378\uc774 \ucd5c\uc801\uc758 \uc194\ub8e8\uc158\uc744 \ucd08\uacfc\ud560 \uc218\ub3c4 \uc788\uc74c<br \/>\ud559\uc2b5\ub960\uc774 \ub0ae\uc73c\uba74 \uc5c5\ub370\uc774\ud2b8\uac00 \uc791\uc544\uc9c0\uace0 \uc218\ub834\uc774 \ub290\ub824\uc9c0\uc9c0\ub9cc \ub85c\uceec \ubbf8\ub2c8\ub9c8\uc5d0 \uac07\ud790 \uc218 \uc788\uc74c<\/li>\n<li>\ucd5c\uc0c1\uc758 \uac00\uc911\uce58(best weights): <br \/>\ucd5c\uc0c1\uc758 \ubaa8\ub378 \uc131\ub2a5\uc744 \ub098\ud0c0\ub0b4\ub294 \uac00\ub2a5\ud55c \ucd5c\uc800 \uc190\uc2e4(lowest possible loss)\uc744 \ucd08\ub798\ud558\ub294 \ub9e4\uac1c\ubcc0\uc218 \uc9d1\ud569<br \/>\uc635\ud2f0\ub9c8\uc774\uc800\uc758 \ubaa9\ud45c\ub294 \uc190\uc2e4 \ud568\uc218\uc758 \uae30\uc6b8\uae30\uc5d0 \ub530\ub77c \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\uc5ec \ubaa8\ub378\uc5d0 \uac00\uc7a5 \uc801\ud569\ud55c \uac00\uc911\uce58\ub97c \ucc3e\ub294 \uac83<\/li>\n<li>\uc608\uce21\ud560 \uc218 \uc5c6\ub294 \ub3d9\uc791(unpredictable behavior):<br \/>\uc635\ud2f0\ub9c8\uc774\uc800\uac00 \ucd5c\uc0c1\uc758 \uac00\uc911\uce58\ub97c \ucc3e\uc744 \uc218 \uc5c6\ub294 \uacbd\uc6b0 \ubaa8\ub378 \ucd5c\uc801\ud654 \ud504\ub85c\uc138\uc2a4 \uc911\uc5d0 \uc608\uce21\ud560 \uc218 \uc5c6\ub294 \ub3d9\uc791\uc774 \ubc1c\uc0dd\ud560 \uc218 \uc788\uc74c<br \/>\uc774\ub294 \uc798\ubabb\ub41c \ud558\uc774\ud37c \ub9e4\uac1c\ubcc0\uc218 \uc120\ud0dd(poor choice of hyperparameter), \uc798\ubabb \ud655\uc7a5\ub41c \ub370\uc774\ud130(poorly scaled data) \ub610\ub294 \ubcf5\uc7a1\ud558\uace0 \uace0\ucc28\uc6d0\uc801\uc778 \ubb38\uc81c(complex, high-dimonsional problem)\uc640 \uac19\uc740 \uc5ec\ub7ec \uc694\uc778\uc73c\ub85c \uc778\ud574 \ubc1c\uc0dd\ud560 \uc218 \uc788\uc74c<br \/>\uc635\ud2f0\ub9c8\uc774\uc800\uac00 \ucc28\uc120\uc758 \uc194\ub8e8\uc158\uc774 \ub420 \uc218 \uc788\ub294 \ub85c\uceec \ubbf8\ub2c8\uba48 \ub610\ub294 \uc548\uc7a5\uc810(saddle point)\uc5d0 \uba48\ucd98 \uacbd\uc6b0\uc5d0\ub3c4 \ubc1c\uc0dd\ud560 \uc218 \uc788\uc74c<\/li>\n<li>\uc9c0\uc5ed \ucd5c\uc18c\uac12(local minima):<br \/>\uadfc\ucc98\uc758 \ubaa8\ub4e0 \uc9c0\uc810\ubcf4\ub2e4 \ub0ae\uc740 \uc190\uc2e4 \ud568\uc218 \ud658\uacbd\uc758 \ud55c \uc9c0\uc810<br \/>\ub354 \ub0ae\uc740 \uc9c0\uc810\uc774 \uc788\uc744 \uc218 \uc788\uc74c<br \/>\ub85c\uceec \ucd5c\uc18c\uac12\uc73c\ub85c \uc218\ub834\ud558\ub3c4\ub85d \ud6c8\ub828\ub41c \ubaa8\ub378\uc740 \uc804\uc5ed \ucd5c\uc18c\uac12\uc73c\ub85c \uc218\ub834\ud558\ub294 \ubaa8\ub378\ubcf4\ub2e4 \uc131\ub2a5\uc774 \uc88b\uc9c0 \uc54a\uc74c<\/li>\n<li>\uc804\uc5ed \ucd5c\uc18c\uac12(global minima): <br \/>\uc804\uccb4\uc801\uc73c\ub85c \uac00\uc7a5 \ub0ae\uc740 \uc9c0\uc810\uc778 \uc190\uc2e4 \ud568\uc218 \ud658\uacbd\uc758 \uc9c0\uc810<br \/>\uc804\uc5ed \ucd5c\uc18c\uac12\uc73c\ub85c \uc218\ub834\ud558\ub3c4\ub85d \ud6c8\ub828\ub41c \ubaa8\ub378\uc774 \ucd5c\uc0c1\uc758 \uc131\ub2a5\uc744 \ubc1c\ud718<\/li>\n<\/ul>\n<\/div>\n<\/div>\n<\/div>\n<div>&nbsp;<\/div>\n<\/div>\n<\/div>\n<\/div>\n<h3 data-ke-size=\"size23\">\ucd5c\uc801\ud654 \ub2e8\uacc4 \ucd94\uac00(Add an optimization loop)<\/h3>\n<div>\n<div>\n<div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ucd5c\uc801\ud654 \ub2e8\uacc4: \ubaa8\ub378 \ud559\uc2b5 \ubc0f \ucd5c\uc801\ud654\uc5d0 \uc0ac\uc6a9<\/li>\n<li>\uac01 \uc5d0\ud3ed: \ub450 \uac00\uc9c0 \uc8fc\uc694 \ubd80\ubd84\uc73c\ub85c \uad6c\uc131\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud6c8\ub828 \ub8e8\ud504: \ud6c8\ub828 \ub370\uc774\ud130 \uc138\ud2b8\ub97c \ubc18\ubcf5\ud558\uace0 \ucd5c\uc801\uc758 \ub9e4\uac1c\ubcc0\uc218\ub85c \uc218\ub834\ud558\ub824\uace0 \uc2dc\ub3c4<\/li>\n<li>\uac80\uc99d\/\ud14c\uc2a4\ud2b8 \ub2e8\uacc4(validation\/test loop): \ud14c\uc2a4\ud2b8 \ub370\uc774\ud130 \uc138\ud2b8\ub97c \ubc18\ubcf5\ud558\uc5ec \ubaa8\ub378 \uc131\ub2a5\uc774 \uac1c\uc120\ub418\uace0 \uc788\ub294\uc9c0 \ud655\uc778<\/li>\n<\/ul>\n<\/li>\n<li>\uad50\uc721 \ub8e8\ud504\uc5d0\uc11c \uc0ac\uc6a9\ub418\ub294 \uac1c\ub150:\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>forward pass: \ubaa8\ub378\uc774 \uc608\uce21\uc744 \uc218\ud589<\/li>\n<li><b>loss calculation: \uc608\uce21\uc744 \uc2e4\uc81c \ucd9c\ub825\uacfc \ube44\uad50<\/b><\/li>\n<li>backward pass: \ubaa8\ub378 \ub9e4\uac1c\ubcc0\uc218\uc5d0 \ub300\ud55c \uc190\uc2e4\uc758 \uae30\uc6b8\uae30\ub97c \uacc4\uc0b0<\/li>\n<li><b>optimizer step: \uacc4\uc0b0\ub41c \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8<\/b><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\ud0a4\uc6cc\ub4dc: optimization loop, train loop, validation loop, test loop, generalization performance, loss calculation, optimizer step<\/p>\n<\/div>\n<\/div>\n<\/div>\n<div>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ucd5c\uc801\ud654 \ub8e8\ud504(optimization loop):<br \/>\uc2e0\uacbd\ub9dd \ud559\uc2b5 \uacfc\uc815\uc758 \uc8fc\uc694 \ub8e8\ud504<br \/>\ubaa8\ub378\uc744 \ud1b5\ud574 \uc785\ub825 \ub370\uc774\ud130\ub97c \ubc18\ubcf5\uc801\uc73c\ub85c \uc804\ub2ec\ud558\uace0, \uc190\uc2e4\uc744 \uacc4\uc0b0\ud558\uace0, \uc635\ud2f0\ub9c8\uc774\uc800\ub97c \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \uac83\uc73c\ub85c \uad6c\uc131\ub428<br \/>\uc9c0\uc815\ub41c \ubc18\ubcf5 \ud69f\uc218 \ub3d9\uc548 \ub610\ub294 \ubaa8\ub378 \uc131\ub2a5\uc774 \ud2b9\uc815 \uc784\uacc4\uac12\uc5d0 \ub3c4\ub2ec\ud560 \ub54c\uae4c\uc9c0 \ubc18\ubcf5\ub428<\/li>\n<li>\ud559\uc2b5 \ub8e8\ud504(train loop):<br \/>\ud6c8\ub828 \ub370\uc774\ud130\uc5d0\uc11c \ubaa8\ub378\uc744 \ud6c8\ub828\uc2dc\ud0a4\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \ud2b9\uc815 \uc720\ud615\uc758 \ucd5c\uc801\ud654 \ub8e8\ud504<br \/>\ud6c8\ub828 \ub8e8\ud504 \ub3d9\uc548 \ud6c8\ub828 \ub370\uc774\ud130\uc5d0\uc11c \uacc4\uc0b0\ub41c \uc190\uc2e4 \ud568\uc218\uc758 \uae30\uc6b8\uae30\ub97c \uae30\ubc18\uc73c\ub85c \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\uac00 \uc5c5\ub370\uc774\ud2b8\ub428<\/li>\n<li>\uac80\uc99d \ub8e8\ud504(validation loop):<br \/>\ud6c8\ub828 \ub8e8\ud504\uc640 \uc720\uc0ac\ud558\uc9c0\ub9cc \uac80\uc99d \ub370\uc774\ud130 \uc138\ud2b8\uc5d0\uc11c \ubaa8\ub378\uc758 \uc131\ub2a5\uc744 \ud3c9\uac00\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<br \/>\ud6c8\ub828 \uc911 \ubaa8\ub378\uc758 \uc131\ub2a5\uc744 \ubaa8\ub2c8\ud130\ub9c1\ud558\uace0 \uacfc\uc801\ud569\uc744 \uac10\uc9c0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>\ud14c\uc2a4\ud2b8 \ub8e8\ud504(test loop):<br \/>\ud559\uc2b5\uc774 \uc644\ub8cc\ub41c \ud6c4 \ud14c\uc2a4\ud2b8 \ub370\uc774\ud130 \uc138\ud2b8\uc5d0\uc11c \ubaa8\ub378\uc758 \uc131\ub2a5\uc744 \ud3c9\uac00\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<br \/>\ubaa8\ub378\uc758 \uc77c\ubc18\ud654 \uc131\ub2a5\uc5d0 \ub300\ud55c \ucd94\uc815\uce58\ub97c \uc81c\uacf5<\/li>\n<li>\uc77c\ubc18\ud654 \uc131\ub2a5(generalization performance):<br \/>\ubcf8 \uc801\uc774 \uc5c6\ub294 \ub370\uc774\ud130\uc5d0 \ub300\ud574 \ubaa8\ub378\uc774 \uc798 \uc218\ud589\ud560 \uc218 \uc788\ub294 \ub2a5\ub825<\/li>\n<li>\uc190\uc2e4 \uacc4\uc0b0(loss calculation):<br \/>\ubaa8\ub378\uc758 \ucd9c\ub825\uacfc \uc2e4\uc81c \ucd9c\ub825\uc5d0 \ub300\ud55c \uc2a4\uce7c\ub77c \uac12 \ud568\uc218\ub97c \uacc4\uc0b0\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>\ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \ub370 \ud544\uc694\ud55c \uc815\ubcf4\ub97c \uc81c\uacf5\ud558\ubbc0\ub85c \ucd5c\uc801\ud654 \ub8e8\ud504\uc5d0\uc11c \uc911\uc694\ud55c \ub2e8\uacc4<\/li>\n<li>\uc635\ud2f0\ub9c8\uc774\uc800 \ub2e8\uacc4(optimizer step):<br \/>\uc190\uc2e4 \ud568\uc218\uc758 \uae30\uc6b8\uae30\uc5d0 \ub530\ub77c \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \ud504\ub85c\uc138\uc2a4<\/li>\n<\/ol>\n<\/div>\n<div>&nbsp;<\/div>\n<\/div>\n<h4 data-ke-size=\"size20\">\uc190\uc2e4 \ud568\uc218 \ucd94\uac00(Add a loss function)<\/h4>\n<div>\n<div>\n<div>\n<div>\n<div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud559\uc2b5 \ub370\uc774\ud130 \uc81c\uacf5 \uc2dc \ud559\uc2b5\ub418\uc9c0 \uc54a\uc740 \ub124\ud2b8\uc6cc\ud06c\ub294 \uc815\ub2f5\uc744 \uc81c\uacf5\ud558\uc9c0 \uc54a\uc744 \uc218 \uc788\uc74c<\/li>\n<li>\uc190\uc2e4 \ud568\uc218: \ubaa9\ud45c \uac12\uc5d0 \ub300\ud55c \uc5bb\uc740 \uacb0\uacfc\uc758 \ube44 \uc720\uc0ac\uc131(degree of dissimilarity)\uc744 \uce21\uc815, \ud559\uc2b5 \uc911 \ucd5c\uc18c\ud654<\/li>\n<li>\uc190\uc2e4 \uacc4\uc0b0:\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc608\uce21: \uc8fc\uc5b4\uc9c4 \ub370\uc774\ud130 \uc0d8\ud50c\uc744 \ubaa8\ub378\uc5d0 \uc785\ub825<\/li>\n<li>\ube44\uad50: \uc2e4\uc81c \ub370\uc774\ud130 \ub808\uc774\ube14 \uac12\uacfc \ube44\uad50<\/li>\n<\/ul>\n<\/li>\n<li>\uc77c\ubc18\uc801\uc778 \uc190\uc2e4 \ud568\uc218:\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>nn.MSELoss: \ud68c\uadc0 \uc791\uc5c5\uc5d0 \uc0ac\uc6a9\ub418\ub294 Mean Square Error<\/li>\n<li>nn.NLLLoss: \ubd84\ub958\uc5d0 \uc0ac\uc6a9\ub418\ub294 \uc74c\uc758 \ub85c\uadf8 \uc6b0\ub3c4<\/li>\n<li>nn.CrossEntropyLoss: <br \/>nn.LogSoftmax\uc640 nn.NLLLoss\ub97c \ud569\uce5c \ub85c\uc2a4<\/li>\n<\/ul>\n<\/li>\n<li>&nbsp;\uc608\uc81c\uc5d0\uc11c\ub294 nn.CrossEntropyLoss \uc0ac\uc6a9, \ubaa8\ub378\uc758 \ucd9c\ub825 \ub85c\uc9d3\uc744 \uc804\ub2ec\ud558\uc5ec \ub85c\uc9d3\uc744 \uc815\uaddc\ud654\ud558\uace0 \uc608\uce21 \uc624\ub958\ub97c \uacc4\uc0b0<\/li>\n<\/ul>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<pre id=\"code_1674191864866\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code># Initialize the loss function\nloss_fn = nn.CrossEntropyLoss()<\/code><\/pre>\n<div>\n<p data-ke-size=\"size16\">keyword: mean square error, negative log likelihood, cross-entropy loss, logits, normalization<\/p>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ud3c9\uade0 \uc81c\uacf1 \uc624\ucc28(Mean Square Error, MSE): <br \/>\uc608\uce21 \ucd9c\ub825\uacfc \uc2e4\uc81c \ucd9c\ub825 \uac04\uc758 \ud3c9\uade0 \uc81c\uacf1 \ucc28\uc774\ub97c \uce21\uc815\ud558\ub294 \uc2a4\uce7c\ub77c \uac12 \ud568\uc218<br \/>\uc5f0\uc18d \uac12\uc744 \uc608\uce21\ud558\ub294 \uac83\uc774 \ubaa9\ud45c\uc778 \ud68c\uadc0 \ubb38\uc81c\uc758 \uc190\uc2e4 \ud568\uc218\ub85c \uc790\uc8fc \uc0ac\uc6a9\ub428<br \/>\uc774\uc0c1\uce58\uc5d0 \ubbfc\uac10\ud558\uba70 \uc791\uc740 \uc624\ub958\ubcf4\ub2e4 \ud070 \uc624\ub958\ub97c \ucc98\ubc8c(punish)\ud568<\/li>\n<li>\uc74c\uc758 \ub85c\uadf8 \uc6b0\ub3c4(Negative Log Likelihood, NLL):<br \/>\uc608\uce21\ub41c \ucd9c\ub825\uc5d0\uc11c \u200b\u200b\uc2e4\uc81c \ucd9c\ub825\uc758 \uc6b0\ub3c4\ub97c \uce21\uc815\ud558\ub294 \uc2a4\uce7c\ub77c \uac12 \ud568\uc218<br \/>\uc774\uc0b0 \uac12\uc744 \uc608\uce21\ud558\ub294 \uac83\uc774 \ubaa9\ud45c\uc778 \ubd84\ub958 \ubb38\uc81c\uc758 \uc190\uc2e4 \ud568\uc218\ub85c \uc790\uc8fc \uc0ac\uc6a9\ub428<br \/>\uc74c\uc758 \ub85c\uadf8\ub294 \ud568\uc218\ub97c \ubbf8\ubd84 \uac00\ub2a5\ud558\uac8c \ub9cc\ub4dc\ub294 \ub370 \uc0ac\uc6a9\ub428<br \/>\uc74c\uc758 \ubd80\ud638\ub294 \ud568\uc218\ub97c \ucd5c\uc18c\ud654\ud574\uc57c \ud558\ub294 \uc190\uc2e4 \ud568\uc218\ub85c \ub9cc\ub4dc\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>\ud06c\ub85c\uc2a4\uc5d4\ud2b8\ub85c\ud53c \uc190\uc2e4(Cross-Entropy Loss, CE):<br \/>\uc608\uce21\ub41c \ucd9c\ub825\uacfc \uc2e4\uc81c \ucd9c\ub825 \uac04\uc758 \ucc28\uc774\ub97c \uce21\uc815\ud558\ub294 \uc2a4\uce7c\ub77c \uac12 \ud568\uc218<br \/>\uc774\uc0b0 \uac12\uc744 \uc608\uce21\ud558\ub294 \uac83\uc774 \ubaa9\ud45c\uc778 \ubd84\ub958 \ubb38\uc81c\uc758 \uc190\uc2e4 \ud568\uc218\ub85c \uc790\uc8fc \uc0ac\uc6a9\ub428<br \/>NLL\uacfc \ubc00\uc811\ud55c \uad00\ub828\uc774 \uc788\uc73c\uba70 \ucd9c\ub825\uc758 \ud655\ub960\uc744 \uc5bb\uae30 \uc704\ud574 softmax activation function\uacfc \ud568\uaed8 \uc790\uc8fc \uc0ac\uc6a9\ub428<\/li>\n<li>\ub85c\uc9d3(logit):<br \/>\ud65c\uc131\ud654 \ud568\uc218\uac00 \uc801\uc6a9\ub418\uae30 \uc804 \ubaa8\ub378\uc758 \uc6d0\uc2dc \ucd9c\ub825<br \/>\uc2e0\uacbd\ub9dd\uc5d0\uc11c \ub85c\uc9d3\uc740 softmax \ud568\uc218\uac00 \uc801\uc6a9\ub418\uae30 \uc804 \ub9c8\uc9c0\ub9c9 \uc120\ud615 \ub808\uc774\uc5b4\uc758 \ucd9c\ub825<br \/>\ub85c\uc9d3\uc740 \ud06c\ub85c\uc2a4 \uc5d4\ud2b8\ub85c\ud53c \uc190\uc2e4\uc744 \uacc4\uc0b0\ud560 \ub54c \uc18c\ud504\ud2b8\ub9e5\uc2a4 \ud568\uc218\uac00 \u200b\u200b\ub098\uc911\uc5d0 \uc801\uc6a9\ub418\uae30 \ub54c\ubb38\uc5d0 \uc720\uc6a9<\/li>\n<\/ol>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<\/div>\n<h4 data-ke-size=\"size20\">\ucd5c\uc801\ud654 \ud328\uc2a4(Optimization pass)<\/h4>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ucd5c\uc801\ud654: \uac01 \ud559\uc2b5 \ub2e8\uacc4\uc5d0\uc11c \ubaa8\ub378 \uc624\ub958\ub97c \uc904\uc774\uae30 \uc704\ud574 \ubaa8\ub378 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc870\uc815\ud558\ub294 \ud504\ub85c\uc138\uc2a4<\/li>\n<li>\ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998: \ucd5c\uc801\ud654 \ud504\ub85c\uc138\uc2a4 \uc218\ud589 \ubc29\ubc95 \uc815\uc758(\uc774 \uc608\uc5d0\uc11c\ub294 Stochastic Gradient Descent\uac00 \uc0ac\uc6a9\ub428)<\/li>\n<li>\uc635\ud2f0\ub9c8\uc774\uc800 \uac1d\uccb4: \ubaa8\ub4e0 \ucd5c\uc801\ud654 \ub85c\uc9c1\uc744 \ucea1\uc290\ud654<\/li>\n<li>SGD \uc635\ud2f0\ub9c8\uc774\uc800: \uc774 \uc608\uc5d0\uc11c \uc0ac\uc6a9\ub418\uba70 \ub2e4\ub978 \uc635\uc158\uc5d0\ub294 ADAM \ubc0f RMSProp \ub4f1\uc774 \uc788\uc74c<\/li>\n<li>\uc635\ud2f0\ub9c8\uc774\uc800 \ucd08\uae30\ud654: \ud559\uc2b5\ud560 \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218 \ubc0f \ud559\uc2b5\ub960\uc744 \ud558\uc774\ud37c\ud30c\ub77c\ubbf8\ud130\ub85c \uc804\ub2ec\ud558\uc5ec \uc2dc\ud589<\/li>\n<li>\ud559\uc2b5 \ub2e8\uacc4(loop)\uc5d0\uc11c\uc758 \ucd5c\uc801\ud654: 3\ub2e8\uacc4\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>optimizer.zero_grad(): \ubaa8\ub378 \ub9e4\uac1c\ubcc0\uc218\uc758 \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uc7ac\uc124\uc815. \uadf8\ub798\ub514\uc5b8\ud2b8\ub294 \uae30\ubcf8\uc801\uc73c\ub85c \ud569\uc0b0(accumulate)\ub418\uae30\uc5d0 \uba85\uc2dc\uc801\uc73c\ub85c 0\uc73c\ub85c \ucd08\uae30\ud654 \ud544\uc694<\/li>\n<li>loss.backwards(): \uc608\uce21 \uc190\uc2e4(prediction loss)\uc744 \uc5ed\uc804\ud30c. PyTorch\ub294 \uac01 \ub9e4\uac1c\ubcc0\uc218\uc5d0 \ub300\ud574 \uc190\uc2e4 \ubcc0\ud654\ub3c4\ub97c \uc800\uc7a5<\/li>\n<li>optimizer.step(): \uc5ed\uc804\ud30c \ud328\uc2a4\uc5d0\uc11c \uc218\uc9d1\ub41c \uadf8\ub798\ub514\uc5b8\ud2b8\ub85c \ub9e4\uac1c\ubcc0\uc218\ub97c \uc870\uc815<\/li>\n<\/ol>\n<\/li>\n<\/ul>\n<pre id=\"code_1674191874373\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">keyword: optimization algorithms, optimizer object, SGD, ADAM, RMSProp<\/p>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998(optimization algorithms):<br \/>\uc190\uc2e4 \ud568\uc218\uc640 \uac19\uc740 \uc2a4\uce7c\ub77c \uac12 \ud568\uc218\ub97c \ucd5c\uc18c\ud654\ud558\uae30 \uc704\ud574 \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc870\uc815\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \ubc29\ubc95<br \/>\uc77c\ubc18\uc801\uc778 \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998\uc5d0\ub294 SGD(Stochastic Gradient Descent), Adam \ubc0f RMSProp\uc774 \ud3ec\ud568\ub428<br \/>\uac01 \uc54c\uace0\ub9ac\uc998\uc5d0\ub294 \ud559\uc2b5 \ud504\ub85c\uc138\uc2a4\ub97c \uc81c\uc5b4\ud558\ub294 \u200b\u200b\ud559\uc2b5 \uc18d\ub3c4\uc640 \uac19\uc740 \uace0\uc720\ud55c \ud558\uc774\ud37c \ub9e4\uac1c\ubcc0\uc218 \uc138\ud2b8\uac00 \uc788\uc74c<\/li>\n<li>\uc635\ud2f0\ub9c8\uc774\uc800 \uac1d\uccb4(optimizer object):<br \/>\ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 PyTorch\uc758 SGD, Adam \ub610\ub294 RMSProp\uacfc \uac19\uc740 \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998\uc758 \uc778\uc2a4\ud134\uc2a4<br \/>\ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc6d0\ud558\ub294 \ud558\uc774\ud37c \ub9e4\uac1c\ubcc0\uc218\uc640 \ud568\uaed8 \uc54c\uace0\ub9ac\uc998\uc758 \uc0dd\uc131\uc790(constructor)\uc5d0 \uc804\ub2ec\ud558\uc5ec \uc635\ud2f0\ub9c8\uc774\uc800 \uac1d\uccb4\ub97c \uc0dd\uc131\ud560 \uc218 \uc788\uc74c<\/li>\n<li>SGD(Stochastic Gradient Descent):<br \/>\uc190\uc2e4 \ud568\uc218\uc758 \uc74c\uc758 \uae30\uc6b8\uae30 \ubc29\ud5a5\uc73c\ub85c \ub2e8\uacc4\ub97c \uc9c4\ud589\ud558\uc5ec \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \uac04\ub2e8\ud55c \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998<br \/>\uacc4\uc0b0\uc801\uc73c\ub85c \ud6a8\uc728\uc801\uc774\uba70 \uad11\ubc94\uc704\ud55c \ucd5c\uc801\ud654 \ubb38\uc81c\uc5d0 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc9c0\ub9cc \ud559\uc2b5\ub960 \uc120\ud0dd\uc5d0 \ubbfc\uac10<\/li>\n<li>ADAM:<br \/>\uacfc\uac70 \uadf8\ub798\ub514\uc5b8\ud2b8 \uc815\ubcf4\ub97c \uae30\ubc18\uc73c\ub85c \uac01 \ub9e4\uac1c\ubcc0\uc218\uc758 \ud559\uc2b5\ub960\uc744 \uc870\uc815\ud558\ub294 \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998<br \/>\uc774\uc804 \ub2e8\uacc4\uc758 \uadf8\ub798\ub514\uc5b8\ud2b8 \uc815\ubcf4\ub97c \uc0ac\uc6a9\ud558\uc5ec \ub2e8\uacc4 \ud06c\uae30\ub97c \uc870\uc815\ud558\ubbc0\ub85c \ud559\uc2b5\ub960 \uc120\ud0dd\uc5d0 \ub354\uc6b1 \uac15\uac74\ud574\uc9d0<\/li>\n<li>RMSProp:<br \/>&nbsp;\uacfc\uac70 \uadf8\ub798\ub514\uc5b8\ud2b8 \uc815\ubcf4\ub97c \uae30\ubc18\uc73c\ub85c \uac01 \ub9e4\uac1c\ubcc0\uc218\uc758 \ud559\uc2b5\ub960\uc744 \uc870\uc815\ud558\ub294 \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998<br \/>\uadf8\ub798\ub514\uc5b8\ud2b8\uc758 \uc81c\uacf1\uc758 \uc774\ub3d9 \ud3c9\uade0\uc744 \uc0ac\uc6a9\ud558\uc5ec \ub2e8\uacc4 \ud06c\uae30\ub97c \uc870\uc815\ud558\ubbc0\ub85c \ud559\uc2b5\ub960 \uc120\ud0dd\uc5d0 \ub354 \uac15\uac74\ud574\uc9d0<\/li>\n<\/ol>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\uc804\uccb4 \uad6c\ud604(Full implementation)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>train_loop: \ucd5c\uc801\ud654 \ucf54\ub4dc\ub97c \ubc18\ubcf5\ud558\uc5ec \uc218\ud589<\/li>\n<li>test_loop: \ud14c\uc2a4\ud2b8 \ub370\uc774\ud130\ub85c \ubaa8\ub378\uc758 \uc131\ub2a5 \uce21\uc815<\/li>\n<li>\ubaa8\ub378 \ud559\uc2b5 \ub2e8\uacc4\n<ul style=\"list-style-type: circle;\" data-ke-list-type=\"circle\">\n<li>\uc190\uc2e4 \ud568\uc218 \uc815\uc758<\/li>\n<li>\uc635\ud2f0\ub9c8\uc774\uc800 \ucd08\uae30\ud654(\ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218 \ubc0f \ub7ec\ub2dd\ub808\uc774\ud2b8 \uc804\ub2ec)<\/li>\n<li>\uc5d0\ud3ed \uc218 \uc815\uc758 \ud6c4 train_loop, test_loop \uc2e4\ud589<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc131\ub2a5 \uc88b\uc9c0 \uc54a\uc744 \uc2dc epoch, learning_rate \ub4f1 \uc870\uc815 \uac00\ub2a5. \ub610\ub294 \ubaa8\ub378\uc774 \uc801\ud569\ud558\uc9c0 \uc54a\uc744 \uc218 \uc788\uc74c. \uc774\ud6c4 \ube44\uc804 \ubb38\uc81c\uc5d0 \ub300\ud574 \ub2e4\ub8f0 \uac83<\/li>\n<\/ul>\n<pre id=\"code_1674191906618\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>def train_loop(dataloader, model, loss_fn, optimizer):\n    size = len(dataloader.dataset)\n    for batch, (X, y) in enumerate(dataloader):        \n        # Compute prediction and loss\n        pred = model(X)\n        loss = loss_fn(pred, y)\n        \n        # Backpropagation\n        optimizer.zero_grad()\n        loss.backward()\n        optimizer.step()\n\n        if batch % 100 == 0:\n            loss, current = loss.item(), batch * len(X)\n            print(f\"loss: {loss:&gt;7f}  [{current:&gt;5d}\/{size:&gt;5d}]\")\n\n\ndef test_loop(dataloader, model, loss_fn):\n    size = len(dataloader.dataset)\n    test_loss, correct = 0, 0\n\n    with torch.no_grad():\n        for X, y in dataloader:\n            pred = model(X)\n            test_loss += loss_fn(pred, y).item()\n            correct += (pred.argmax(1) == y).type(torch.float).sum().item()\n            \n    test_loss \/= size\n    correct \/= size\n    print(f\"Test Error: \\n Accuracy: {(100*correct):&gt;0.1f}%, Avg loss: {test_loss:&gt;8f} \\n\")\n    \n    loss_fn = nn.CrossEntropyLoss()\noptimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)\n\nepochs = 10\nfor t in range(epochs):\n    print(f\"Epoch {t+1}\\n-------------------------------\")\n    train_loop(train_dataloader, model, loss_fn, optimizer)\n    test_loop(test_dataloader, model, loss_fn)\nprint(\"Done!\")\n\n\nEpoch 1\n-------------------------------\nloss: 2.307260  [    0\/60000]\nloss: 2.305284  [ 6400\/60000]\nloss: 2.293966  [12800\/60000]\nloss: 2.291592  [19200\/60000]\nloss: 2.288022  [25600\/60000]\nloss: 2.259277  [32000\/60000]\nloss: 2.277950  [38400\/60000]\nloss: 2.252569  [44800\/60000]\nloss: 2.238333  [51200\/60000]\nloss: 2.239141  [57600\/60000]\nTest Error: \n Accuracy: 27.5%, Avg loss: 0.035050 \n\nEpoch 2\n-------------------------------\nloss: 2.222609  [    0\/60000]\nloss: 2.244805  [ 6400\/60000]\nloss: 2.209550  [12800\/60000]\nloss: 2.227453  [19200\/60000]\nloss: 2.217051  [25600\/60000]\nloss: 2.162092  [32000\/60000]\nloss: 2.206926  [38400\/60000]\nloss: 2.151579  [44800\/60000]\nloss: 2.117667  [51200\/60000]\nloss: 2.143689  [57600\/60000]\nTest Error: \n Accuracy: 38.9%, Avg loss: 0.033368 \n\nEpoch 3\n-------------------------------\nloss: 2.102783  [    0\/60000]\nloss: 2.154025  [ 6400\/60000]\nloss: 2.076486  [12800\/60000]\nloss: 2.124048  [19200\/60000]\nloss: 2.107713  [25600\/60000]\nloss: 2.014179  [32000\/60000]\nloss: 2.090220  [38400\/60000]\nloss: 1.989485  [44800\/60000]\nloss: 1.933911  [51200\/60000]\nloss: 2.002917  [57600\/60000]\nTest Error: \n Accuracy: 41.2%, Avg loss: 0.030885 \n\nEpoch 4\n-------------------------------\nloss: 1.926293  [    0\/60000]\nloss: 2.019496  [ 6400\/60000]\nloss: 1.888668  [12800\/60000]\nloss: 1.987653  [19200\/60000]\nloss: 1.968171  [25600\/60000]\nloss: 1.838344  [32000\/60000]\nloss: 1.951870  [38400\/60000]\nloss: 1.808960  [44800\/60000]\nloss: 1.749038  [51200\/60000]\nloss: 1.868777  [57600\/60000]\nTest Error: \n Accuracy: 44.4%, Avg loss: 0.028537 \n\nEpoch 5\n-------------------------------\nloss: 1.754023  [    0\/60000]\nloss: 1.889865  [ 6400\/60000]\nloss: 1.724985  [12800\/60000]\nloss: 1.880932  [19200\/60000]\nloss: 1.852289  [25600\/60000]\nloss: 1.703095  [32000\/60000]\nloss: 1.850078  [38400\/60000]\nloss: 1.679640  [44800\/60000]\nloss: 1.618462  [51200\/60000]\nloss: 1.781099  [57600\/60000]\nTest Error: \n Accuracy: 46.4%, Avg loss: 0.026904 \n\nEpoch 6\n-------------------------------\nloss: 1.629323  [    0\/60000]\nloss: 1.794621  [ 6400\/60000]\nloss: 1.609603  [12800\/60000]\nloss: 1.806047  [19200\/60000]\nloss: 1.771073  [25600\/60000]\nloss: 1.610854  [32000\/60000]\nloss: 1.782800  [38400\/60000]\nloss: 1.593032  [44800\/60000]\nloss: 1.530435  [51200\/60000]\nloss: 1.721836  [57600\/60000]\nTest Error: \n Accuracy: 47.5%, Avg loss: 0.025738 \n\nEpoch 7\n-------------------------------\nloss: 1.541017  [    0\/60000]\nloss: 1.723998  [ 6400\/60000]\nloss: 1.525540  [12800\/60000]\nloss: 1.745950  [19200\/60000]\nloss: 1.714844  [25600\/60000]\nloss: 1.542636  [32000\/60000]\nloss: 1.735072  [38400\/60000]\nloss: 1.529822  [44800\/60000]\nloss: 1.467118  [51200\/60000]\nloss: 1.675812  [57600\/60000]\nTest Error: \n Accuracy: 48.3%, Avg loss: 0.024844 \n\nEpoch 8\n-------------------------------\nloss: 1.474333  [    0\/60000]\nloss: 1.669000  [ 6400\/60000]\nloss: 1.460421  [12800\/60000]\nloss: 1.694097  [19200\/60000]\nloss: 1.674764  [25600\/60000]\nloss: 1.487773  [32000\/60000]\nloss: 1.699166  [38400\/60000]\nloss: 1.481064  [44800\/60000]\nloss: 1.419311  [51200\/60000]\nloss: 1.638599  [57600\/60000]\nTest Error: \n Accuracy: 48.7%, Avg loss: 0.024137 \n\nEpoch 9\n-------------------------------\nloss: 1.420322  [    0\/60000]\nloss: 1.625176  [ 6400\/60000]\nloss: 1.408073  [12800\/60000]\nloss: 1.649715  [19200\/60000]\nloss: 1.644693  [25600\/60000]\nloss: 1.443653  [32000\/60000]\nloss: 1.671596  [38400\/60000]\nloss: 1.443777  [44800\/60000]\nloss: 1.382555  [51200\/60000]\nloss: 1.608089  [57600\/60000]\nTest Error: \n Accuracy: 49.1%, Avg loss: 0.023570 \n\nEpoch 10\n-------------------------------\nloss: 1.375013  [    0\/60000]\nloss: 1.588062  [ 6400\/60000]\nloss: 1.364595  [12800\/60000]\nloss: 1.612044  [19200\/60000]\nloss: 1.621220  [25600\/60000]\nloss: 1.407904  [32000\/60000]\nloss: 1.649211  [38400\/60000]\nloss: 1.415225  [44800\/60000]\nloss: 1.353849  [51200\/60000]\nloss: 1.582835  [57600\/60000]\nTest Error: \n Accuracy: 49.5%, Avg loss: 0.023104 \n\nDone!<\/code><\/pre>\n<h3 data-ke-size=\"size23\">\ubaa8\ub378 \uc800\uc7a5(Saving Models)<\/h3>\n<p data-ke-size=\"size16\">\ubaa8\ub378 \uc131\ub2a5 \ub9cc\uc871 \uc2dc torch.save \uc0ac\uc6a9\ud558\uc5ec \uc800\uc7a5 \uac00\ub2a5<\/p>\n<p data-ke-size=\"size16\">pytorch \ubaa8\ub378\uc740 state_dict\ub77c\ub294 internal state dictionary\uc5d0 \ub9e4\uac1c\ubcc0\uc218 \uc800\uc7a5<\/p>\n<p data-ke-size=\"size16\">T<span style=\"background-color: #ffffff; color: #000000;\">hese can be persisted(\uc720\uc9c0) with the<span>&nbsp;<\/span><\/span>torch.save<span style=\"background-color: #ffffff; color: #000000;\"><span>&nbsp;<\/span>method:<\/span><\/p>\n<pre id=\"code_1674229236921\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>torch.save(model.state_dict(), \"data\/model.pth\")\n\nprint(\"Saved PyTorch Model State to model.pth\")<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">keyword: torch.save, internal state dictionary, state_dict<\/p>\n<div>\n<div>\n<div>\n<div>\n<div>\n<div>\n<div>\n<div>\n<div>\n<div>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>torch.save:<br \/>\ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\uc640 \uc635\ud2f0\ub9c8\uc774\uc800\uc758 \uc0c1\ud0dc\ub97c \ud3ec\ud568\ud558\uc5ec PyTorch \ubaa8\ub378\uc758 \uc0c1\ud0dc\ub97c \uc800\uc7a5\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \ud568\uc218<br \/>\ub098\uc911\uc5d0 torch.load \uae30\ub2a5\uc744 \uc0ac\uc6a9\ud558\uc5ec \ub85c\ub4dc\ud560 \uc218 \uc788\ub294 \ud30c\uc77c\uc5d0 \ubaa8\ub378\uc744 \uc800\uc7a5<\/li>\n<li>\ub0b4\ubd80 \uc0c1\ud0dc \uc0ac\uc804(internal state dictionary):<br \/>\ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218 \ubc0f \uc635\ud2f0\ub9c8\uc774\uc800\uc758 \uc0c1\ud0dc\ub97c \ud3ec\ud568\ud558\uc5ec PyTorch \ubaa8\ub378\uc758 \ud604\uc7ac \uc0c1\ud0dc\ub97c \ud3ec\ud568\ud558\ub294 Python \uc0ac\uc804<br \/>\uc774 \uc0ac\uc804\uc740 \ubaa8\ub378\uc774\ub098 \uc635\ud2f0\ub9c8\uc774\uc800\uc758 state_dict() \ud568\uc218\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc561\uc138\uc2a4\ud560 \uc218 \uc788\uc74c<\/li>\n<li>state_dict():<br \/>\ubaa8\ub378 \ub610\ub294 \uc635\ud2f0\ub9c8\uc774\uc800\uc758 \ud559\uc2b5 \uac00\ub2a5\ud55c \ubaa8\ub4e0 \ub9e4\uac1c\ubcc0\uc218\ub97c \ud3ec\ud568\ud558\ub294 \uc815\ub82c\ub41c \uc0ac\uc804\uc744 \ubc18\ud658\ud558\ub294 \ud568\uc218<br \/>torch.save \ud568\uc218\ub97c \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc758 \ud604\uc7ac \uc0c1\ud0dc\ub97c \uc800\uc7a5\ud558\uac70\ub098 torch.load \ud568\uc218\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc0ac\uc804 \ud6c8\ub828\ub41c \ubaa8\ub378\uc744 \ub85c\ub4dc\ud558\ub294 \ub370 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<br \/>\ud0a4\ub294 \ub9e4\uac1c\ubcc0\uc218\uc758 \uc774\ub984<br \/>\uac12\uc740 \ud574\ub2f9 \ub9e4\uac1c\ubcc0\uc218 \uac12\uc744 \ud3ec\ud568\ud558\ub294 \ud150\uc11c<\/li>\n<\/ol>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<p data-ke-size=\"size16\">&nbsp;<\/p>","category":"Pytorch\/\ud29c\ud1a0\ub9ac\uc5bc","author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/132","comments":"https:\/\/hoohaha.tistory.com\/132#entry132comment","pubDate":"Fri, 20 Jan 2023 17:16:14 +0900"},{"title":"[PyTorch] \uacf5\uc2dd \ubb38\uc11c Learn the Basics \uc694\uc57d - 5. Automatic differentiation with torch.autograd","link":"https:\/\/hoohaha.tistory.com\/131","description":"<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"444\" data-origin-height=\"246\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/kWmtN\/btrWNaydhRB\/1iJ2UU0c6M3hVomrY1gDZ1\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/kWmtN\/btrWNaydhRB\/1iJ2UU0c6M3hVomrY1gDZ1\/img.png\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/kWmtN\/btrWNaydhRB\/1iJ2UU0c6M3hVomrY1gDZ1\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FkWmtN%2FbtrWNaydhRB%2F1iJ2UU0c6M3hVomrY1gDZ1%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"444\" height=\"246\" data-origin-width=\"444\" data-origin-height=\"246\"\/><\/span><\/figure>\n<\/p>\n<p data-ke-size=\"size16\"><b>\ubaa9\ucc28<\/b><\/p>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\">torch.autograd\ub97c \uc0ac\uc6a9\ud55c \uc790\ub3d9 \ubbf8\ubd84(Automatic differentiation with torch.autograd)<\/p>\n<p data-ke-size=\"size16\">Tensor, Function\uacfc \uc5f0\uc0b0 \uadf8\ub798\ud504(Tensors, Functions and Computational graph)<\/p>\n<p data-ke-size=\"size16\">\ubcc0\ud654\ub3c4 \uacc4\uc0b0\ud558\uae30(Computing gradients)<\/p>\n<p data-ke-size=\"size16\">\ubcc0\ud654\ub3c4 \ucd94\uc801 \uba48\ucd94\uae30(Disabling gradient tracking)<\/p>\n<p data-ke-size=\"size16\">\uc5f0\uc0b0 \uadf8\ub798\ud504\uc5d0 \ub300\ud55c \ucd94\uac00 \uc815\ubcf4(more on Computational Graphs)<\/p>\n<p data-ke-size=\"size16\">\uc120\ud0dd\uc801\uc73c\ub85c \uc77d\uae30: \ud150\uc11c \ubcc0\ud654\ub3c4\uc640 \uc57c\ucf54\ube44\uc548 \uacf1(Optional Reading: Tensor gradients and jacobian products)<\/p>\n<\/div>\n<\/div>\n<p data-ke-size=\"size16\">Keywords: backpropagation, gradient, loss function, retraining, gradient descent, differentiation engine, computational graph, class Function in pytorch, forward direction, requires_grad property, grad_fn property, forward pass, backward pass, &part;loss\/&part;w, &part;loss\/&part;b, loss.backward(), w.grad, b.grad, leaf nodes, root node, backward call, retain_graph=True, torch.no_grad() block, gradient tracking, frozen parameter, fine tuning, autograd, directed acyclic graph, chain rule, dynamic graph, control flow statements, computation vs operation, cycle vs loop, scalar loss function, Jacobian matrix, vector function, Jacobian product, shape vs size on tensor, scalar-valued function, optimizer<\/p>\n<hr contenteditable=\"false\" data-ke-type=\"horizontalRule\" data-ke-style=\"style5\" \/>\n<h2 data-ke-size=\"size26\">torch.autograd\ub97c \uc0ac\uc6a9\ud55c \uc790\ub3d9 \ubbf8\ubd84(Automatic differentiation with torch.autograd)<\/h2>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc5ed\uc804\ud30c: \uc2e0\uacbd\ub9dd \ud6c8\ub828\uc5d0 \uac00\uc7a5 \uc790\uc8fc \uc0ac\uc6a9\ub418\ub294 \uc54c\uace0\ub9ac\uc998<br \/>\uc190\uc2e4 \ud568\uc218\uc758 <b>\ubcc0\ud654\ub3c4(gradient)<\/b>\uc5d0 \ub530\ub77c \ub9e4\uac1c\ubcc0\uc218(\ubaa8\ub378 \uac00\uc911\uce58)\ub97c \uc870\uc815<br \/><b>\ub124\ud2b8\uc6cc\ud06c\ub97c \ud1b5\ud574 \uc5ed\ubc29\ud5a5\uc73c\ub85c \uc774\ub3d9\ud558\uc5ec \uac00\uc911\uce58\uc640 \ud3b8\ud5a5\uc744 \uc870\uc815\ud558\uc5ec \ubaa8\ub378\uc744 \uc7ac\ud6c8\ub828<\/b><\/li>\n<li>\uc190\uc2e4 \ud568\uc218: \uc2e0\uacbd\ub9dd\uc758 \uc608\uc0c1 \ucd9c\ub825\uacfc \uc2e4\uc81c \ucd9c\ub825 \uac04\uc758 \ucc28\uc774\ub97c \uacc4\uc0b0<br \/>\ubaa9\ud45c: \uc190\uc2e4 \ud568\uc218\ub97c 0\uc73c\ub85c \ucd5c\uc18c\ud654\ud558\ub294 \uac83<\/li>\n<li><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\uacbd\uc0ac\ud558\uac15\ubc95(gradient descent): <b>\uc190\uc2e4\uc744 \uc904\uc774\uae30 \uc704\ud574 \uc2dc\uac04\uc774 \uc9c0\ub0a8\uc5d0 \ub530\ub77c \ubaa8\ub378\uc744 \uc7ac\ud6c8\ub828(retraining)<\/b>\ud558\ub294 \ud504\ub85c\uc138\uc2a4<\/span><\/li>\n<li>PyTorch\uc5d0\ub294 \uadf8\ub798\ub514\uc5b8\ud2b8 \uacc4\uc0b0\uc744 \uc704\ud55c torch.autograd\ub77c\ub294 \ubbf8\ubd84 \uc5d4\uc9c4\uc774 \ub0b4\uc7a5\ub418\uc5b4 \uc788\uc74c<\/li>\n<li>\ub2e4\uc74c\uacfc \uac19\uc774 \uac04\ub2e8\ud55c 1\uacc4\uce35 \uc2e0\uacbd\ub9dd \uc785\ub825 x, \ub9e4\uac1c\ubcc0\uc218 w \ubc0f b, \uc190\uc2e4 \ud568\uc218\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc815\uc758\ud560 \uc218 \uc788\uc74c<\/li>\n<\/ul>\n<pre id=\"code_1674109244986\" class=\"makefile\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>import torch\n\nx = torch.ones(5)  # input tensor\ny = torch.zeros(3)  # expected output\nw = torch.randn(5, 3, requires_grad=True)\nb = torch.randn(3, requires_grad=True)\nz = torch.matmul(x, w)+b\nloss = torch.nn.functional.binary_cross_entropy_with_logits(z, y)<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">Keywords: backpropagation, gradient, loss function, retraining, gradient descent, differentiation engine, computational graph<\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>backpropagation:<br \/>\uc2e0\uacbd\ub9dd\uc744 \ud6c8\ub828\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \uc54c\uace0\ub9ac\uc998\uc73c\ub85c, \uac01 \uacc4\uce35\uc5d0\uc11c \uc624\ub958\ub97c \uacc4\uc0b0\ud558\uace0 \uadf8\uc5d0 \ub530\ub77c \uac00\uc911\uce58\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>gradient:<br \/>gradient descent\uacfc \uac19\uc740 \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998\uc5d0 \uc0ac\uc6a9\ub418\ub294 <b>\ud568\uc218\uc758 \uac00\uc7a5 \uae09\uaca9\ud55c \uc99d\uac00(steepest increse)\uc758 \ubc29\ud5a5\uacfc \ud06c\uae30\ub97c \ub098\ud0c0\ub0b4\ub294 \ubca1\ud130<\/b><\/li>\n<li>loss function:<br \/>\ubaa8\ub378\uc744 \ucd5c\uc801\ud654\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \ud568\uc218<br \/>\uc2e0\uacbd\ub9dd\uc5d0\uc11c \uc608\uce21\ub41c \ucd9c\ub825\uacfc \uc2e4\uc81c \ucd9c\ub825 \uac04\uc758 \uc624\ucc28 \ub610\ub294 \ucc28\uc774\ub97c \uce21\uc815\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \ud568\uc218<\/li>\n<li>retraining:<br \/>\uc131\ub2a5\uc744 \ud5a5\uc0c1\uc2dc\ud0a4\uae30 \uc704\ud574 \uc0c8\ub85c\uc6b4 \ub370\uc774\ud130\uc5d0 \ub300\ud574 \ubaa8\ub378\uc744 \ub2e4\uc2dc \ud6c8\ub828\uc2dc\ud0a4\ub294 \uacfc\uc815<\/li>\n<li>gradient descent:<br \/>gradient\uc758 \ubc18\ub300 \ubc29\ud5a5\uc73c\ub85c weights\ub97c \uc870\uc815\ud558\uc5ec \uc2e0\uacbd\ub9dd\uc5d0\uc11c loss function\uc744 \ucd5c\uc18c\ud654\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998<\/li>\n<li>differentiation engine: <br \/>\uc2e0\uacbd\ub9dd\uc758 \ub2e4\uc591\ud55c \uc5f0\uc0b0\uc5d0 \ub300\ud55c gradients\ub97c \uacc4\uc0b0\ud558\ub294 \uad6c\uc131 \uc694\uc18c(component)<br \/>backpropagation\uc744 \uc704\ud55c gradients\ub97c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>computational graph:<br \/><b>\uc218\ud559\uc801 \uacc4\uc0b0\uc758 \uadf8\ub798\ud504 \ud45c\ud604<br \/><\/b><span>\uc2e0\uacbd\ub9dd\uc5d0\uc11c backpropagation\ud558\ub294 \ub3d9\uc548 gradients\ub97c \ud6a8\uc728\uc801\uc73c\ub85c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/span><\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">Tensor, Function\uacfc \uc5f0\uc0b0 \uadf8\ub798\ud504(Tensors, Functions and Computational graph)<\/h3>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p><figure class=\"imageblock alignCenter\" data-ke-mobileStyle=\"widthOrigin\" data-origin-width=\"1008\" data-origin-height=\"347\"><span data-url=\"https:\/\/blog.kakaocdn.net\/dn\/q3qY7\/btrWIzle7I1\/4MQLZo6qlvVGmrOwXSjbLK\/img.png\" data-phocus=\"https:\/\/blog.kakaocdn.net\/dn\/q3qY7\/btrWIzle7I1\/4MQLZo6qlvVGmrOwXSjbLK\/img.png\" data-alt=\"\uc704 \ucf54\ub4dc\ub85c\ubd80\ud130 \uc815\uc758\ub41c computational graph\"><img src=\"https:\/\/blog.kakaocdn.net\/dn\/q3qY7\/btrWIzle7I1\/4MQLZo6qlvVGmrOwXSjbLK\/img.png\" srcset=\"https:\/\/img1.daumcdn.net\/thumb\/R1280x0\/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2Fq3qY7%2FbtrWIzle7I1%2F4MQLZo6qlvVGmrOwXSjbLK%2Fimg.png\" onerror=\"this.onerror=null; this.src='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png'; this.srcset='\/\/t1.daumcdn.net\/tistory_admin\/static\/images\/no-image-v1.png';\" loading=\"lazy\" width=\"1008\" height=\"347\" data-origin-width=\"1008\" data-origin-height=\"347\"\/><\/span><figcaption>\uc704 \ucf54\ub4dc\ub85c\ubd80\ud130 \uc815\uc758\ub41c computational graph<\/figcaption>\n<\/figure>\n<\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc774 \uc2e0\uacbd\ub9dd\uc5d0\uc11c\ub294 \ub9e4\uac1c\ubcc0\uc218 w\uc640 b\ub97c \ucd5c\uc801\ud654\ud574\uc57c \ud558\uace0 \uae30\uc6b8\uae30\ub97c \uacc4\uc0b0\ud574\uc57c \ud568<\/li>\n<li>\uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0\ud558\ub824\uba74 \ud150\uc11c\uc758 requires_grad \uc18d\uc131 \uc124\uc815 \ud544\uc694<\/li>\n<li>\uacc4\uc0b0 \uadf8\ub798\ud504\ub97c \uad6c\uc131\ud558\uae30 \uc704\ud574 \ud150\uc11c\uc5d0 \uc801\uc6a9\ub418\ub294 \ud568\uc218: Function \ud074\ub798\uc2a4\uc758 \uac1c\uccb4<\/li>\n<li>\ud568\uc218 \uac1c\uccb4\ub294 \uc21c\ubc29\ud5a5\uc73c\ub85c \ud568\uc218\ub97c \uacc4\uc0b0\ud558\ub294 \ubc29\ubc95\uacfc \uc5ed\ubc29\ud5a5 \uc804\ud30c \uc911\uc5d0 \ub3c4\ud568\uc218\ub97c \uacc4\uc0b0\ud558\ub294 \ubc29\ubc95 \uc54c\uace0 \uc788\uc74c<\/li>\n<li>\uc5ed\ubc29\ud5a5 \uc804\ud30c \ud568\uc218\uc5d0 \ub300\ud55c \ucc38\uc870(reference)\ub294 \ud150\uc11c\uc758 grad_fn \uc18d\uc131\uc5d0 \uc800\uc7a5\ub428<\/li>\n<li>\ud150\uc11c\ub97c \uc0dd\uc131\ud560 \ub54c(requires_grad=True \ud30c\ub77c\ubbf8\ud130) \ub610\ub294 \ub098\uc911\uc5d0 x.requires_grad_(True) \uba54\uc11c\ub4dc\ub97c \uc0ac\uc6a9\ud558\uc5ec requires_grad \uac12\uc744 \uc124\uc815\ud560 \uc218 \uc788\uc74c<\/li>\n<\/ul>\n<pre id=\"code_1674110701968\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>print('Gradient function for z =',z.grad_fn)\nprint('Gradient function for loss =', loss.grad_fn)\n\nGradient function for z = &lt;AddBackward0 object at 0x00000280CC630CA0&gt;\nGradient function for loss = &lt;BinaryCrossEntropyWithLogitsBackward object at 0x00000280CC630310&gt;<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">Keywords: class Function in pytorch, forward direction, requires_grad property, grad_fn property, forward pass, backward pass<\/p>\n<div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>class 'Function' in pytorch:<br \/>\uc2e0\uacbd\ub9dd\uc758 \ubaa8\ub4e0 \uc791\uc5c5\uc5d0 \ub300\ud55c \uae30\ubcf8 \ud074\ub798\uc2a4\ub85c \ub2e8\uc77c \uc815\ubc29\ud5a5 \uacc4\uc0b0\uc744 \ub098\ud0c0\ub0c4<\/li>\n<li>forward direction:<br \/>\uc77c\ubc18\uc801\uc73c\ub85c \uc785\ub825\uc5d0\uc11c \ucd9c\ub825\uc73c\ub85c \uc2e0\uacbd\ub9dd\uc5d0\uc11c \uacc4\uc0b0\uc774 \uc218\ud589\ub418\ub294 \ubc29\ud5a5<\/li>\n<li>require_grad property:<br \/>PyTorch \ud150\uc11c\uc758 \uc18d\uc131<br \/>\uc5ed\ubc29\ud5a5 \uc804\ub2ec(backward pass) \uc911\uc5d0 \ud150\uc11c\uc5d0 \ub300\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0\ud574\uc57c \ud558\ub294\uc9c0 \uc5ec\ubd80\ub97c \ub098\ud0c0\ub0c4<\/li>\n<li>grad_fn property:<br \/>\ud150\uc11c\ub97c \uc0dd\uc131\ud55c \ud568\uc218\ub97c \ucc38\uc870\ud558\ub294 PyTorch \ud150\uc11c\uc758 \uc18d\uc131<br \/>\uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0\ud558\uae30 \uc704\ud574 backward pass\uc5d0 \uc0ac\uc6a9\ub428<\/li>\n<li>forward pass:<br \/>\uc785\ub825\uc774 \uc8fc\uc5b4\uc9c4 \uc2e0\uacbd\ub9dd\uc758 \ucd9c\ub825\uc744 \uacc4\uc0b0\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>\uc785\ub825 \ud150\uc11c\ub85c \uc2dc\uc791\ud558\uc5ec \uacc4\uc0b0 \uadf8\ub798\ud504\uc758 Function \ud074\ub798\uc2a4\uc5d0 \uc758\ud574 \uc815\uc758\ub41c \uc77c\ub828\uc758 \uc791\uc5c5\uc744 \uc801\uc6a9\ud558\uc5ec \ud55c \uc791\uc5c5\uc758 \ucd9c\ub825\uc744 \ub2e4\uc74c \uc791\uc5c5\uc758 \uc785\ub825\uc73c\ub85c \uc804\ub2ec<br \/>forward pass\uc758 \uacb0\uacfc\ub294 \uc77c\ubc18\uc801\uc73c\ub85c loss \ud568\uc218\ub97c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \ucd9c\ub825 \ud150\uc11c<\/li>\n<li>backward pass:<br \/>\ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\uc5d0 \ub300\ud55c loss \ud568\uc218\uc758 \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>\uc77c\ubc18\uc801\uc73c\ub85c loss \ud150\uc11c\uc778 \ub8e8\ud2b8 \ub178\ub4dc\uc5d0\uc11c \uc2dc\uc791\ud558\uc5ec chain rule of differentiation\uc744 \uc801\uc6a9\ud558\uc5ec \uacc4\uc0b0 \uadf8\ub798\ud504\uc758 \ubaa8\ub4e0 \ud150\uc11c\uc5d0 \ub300\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0<br \/>\uadf8\ub798\ub514\uc5b8\ud2b8\ub294 \ud150\uc11c\uc758 grad \uc18d\uc131\uc5d0 \uc800\uc7a5\ub418\uba70 \uc635\ud2f0\ub9c8\uc774\uc800\ub97c \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \ub370 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<br \/>backward pass\ub294 \ub8e8\ud2b8 \ud150\uc11c\uc5d0\uc11c backward() \uba54\uc11c\ub4dc\ub97c \ud638\ucd9c\ud558\uc5ec \uc218\ud589\ub428<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\"><b>class '<a href=\"https:\/\/pytorch.org\/docs\/stable\/autograd.html#function\" target=\"_blank\" rel=\"noopener\">Function<\/a>' in pytorch \uc0c1\uc138<\/b><\/p>\n<\/div>\n<div data-ke-type=\"moreLess\" data-text-more=\"\ub354\ubcf4\uae30\" data-text-less=\"\ub2eb\uae30\"><a class=\"btn-toggle-moreless\">\ub354\ubcf4\uae30<\/a>\n<div class=\"moreless-content\">\n<p data-ke-size=\"size16\">PyTorch\uc758 Function \ud074\ub798\uc2a4\ub294 \uc2e0\uacbd\ub9dd\uc5d0\uc11c \uacc4\uc0b0\uc744 \ub9cc\ub4e4\uace0 \ud45c\ud604\ud558\uae30 \uc704\ud55c \uae30\ubcf8 \ube4c\ub529 \ube14\ub85d\uc785\ub2c8\ub2e4. \ud589\ub82c \uacf1\uc148 \ub610\ub294 \ube44\uc120\ud615 \ud65c\uc131\ud654 \ud568\uc218\uc640 \uac19\uc740 \ud2b9\uc815 \uc5f0\uc0b0\uc758 \uc815\ubc29\ud5a5 \uacc4\uc0b0\uc744 \uc815\uc758\ud558\ub294 \ub370 \uc0ac\uc6a9\ub429\ub2c8\ub2e4. \ub610\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0\ud558\uae30 \uc704\ud574 \uc5ed\ubc29\ud5a5 \ud328\uc2a4\uc5d0\uc11c \uc0ac\uc6a9\ub418\ub294 \uacc4\uc0b0\uc758 \uc785\ub825 \ubc0f \ucd9c\ub825 \ud150\uc11c\uc5d0 \ub300\ud55c \uc815\ubcf4\ub97c \ubcf4\uc720\ud569\ub2c8\ub2e4.<br \/>\uc0ac\uc6a9\uc790\ub294 Function\uc744 \uc11c\ube0c\ud074\ub798\uc2f1\ud558\uace0 \uc785\ub825 \ud150\uc11c\uc5d0\uc11c \uc218\ud589\ub418\ub294 \uacc4\uc0b0\uc744 \uc815\uc758\ud558\uace0 \ucd9c\ub825 \ud150\uc11c\ub97c \ubc18\ud658\ud558\ub294 \uc804\ub2ec \uba54\uc11c\ub4dc\ub97c \uad6c\ud604\ud558\uc5ec \uc0c8 \uc791\uc5c5\uc744 \uc815\uc758\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc785\ub825 \ud150\uc11c\uc5d0 \ub300\ud55c \uae30\uc6b8\uae30\ub97c \uacc4\uc0b0\ud558\ub294 \uc5ed\ubc29\ud5a5 \ubc29\ubc95\uc740 autograd \ud328\ud0a4\uc9c0\ub97c \uc0ac\uc6a9\ud558\uc5ec PyTorch\uc5d0 \uc758\ud574 \uc790\ub3d9\uc73c\ub85c \uc815\uc758\ub429\ub2c8\ub2e4.<br \/>\uc2e0\uacbd\ub9dd\uc5d0\uc11c \uacc4\uc0b0\uc774 \uc218\ud589\ub418\uba74 PyTorch\ub294 \ud574\ub2f9 Function \ud074\ub798\uc2a4\uc758 \uc778\uc2a4\ud134\uc2a4\ub97c \uc0dd\uc131\ud558\uace0 \uc804\ub2ec \uba54\uc11c\ub4dc\ub97c \ud638\ucd9c\ud569\ub2c8\ub2e4. \uc21c\ubc29\ud5a5 \uba54\uc11c\ub4dc\uc5d0\uc11c \ubc18\ud658\ub41c \ucd9c\ub825 \ud150\uc11c\ub294 grad_fn \uc18d\uc131\uacfc \uc5f0\uacb0\ub418\uc5b4 \uc788\uc5b4 PyTorch\uac00 \uacc4\uc0b0\uc744 \ucd94\uc801\ud558\uace0 \uc5ed\ubc29\ud5a5 \ud328\uc2a4 \uc911\uc5d0 \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<br \/>\ub610\ud55c \uae30\ubcf8\uc801\uc73c\ub85c PyTorch\uc5d0 \uc758\ud574 \uc0dd\uc131\ub41c \ubaa8\ub4e0 \ud150\uc11c\uc5d0\ub294 require_grad=False\uac00 \uc788\uc2b5\ub2c8\ub2e4. \uc989, \uc774 \ud150\uc11c\uc5d0 \ub300\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8\ub294 \uc5ed\ubc29\ud5a5 \ud328\uc2a4 \uc911\uc5d0 \uacc4\uc0b0\ub418\uc9c0 \uc54a\uc2b5\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \ud150\uc11c\uc5d0 require_grad=True\ub97c \uc124\uc815\ud558\uba74 \uadf8\ub798\ub514\uc5b8\ud2b8\uac00 \uacc4\uc0b0\ub418\uc5b4 \ud150\uc11c\ub97c \uc5ed\ubc29\ud5a5 \uc804\ub2ec\uc5d0 \uc0ac\uc6a9\ud560 \uc218 \uc788\uace0 \uc0ac\uc6a9\uc790\uac00 \ucd5c\uc801\ud654 \ud504\ub85c\uadf8\ub7a8\uc744 \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">The Function class in PyTorch is a fundamental building block for creating and representing computations in a neural network. It is used to define the forward computation of a specific operation, such as a matrix multiplication or a non-linear activation function. It also holds information about the input and output tensors of the computation, which are used in the backward pass to compute gradients.<br \/>A user can define a new operation by subclassing Function and implementing the forward method, which defines the computation performed on the input tensors and returns the output tensors. The backward method, which computes gradients with respect to the input tensors, is automatically defined by PyTorch using the autograd package.<br \/>When a computation is performed in a neural network, PyTorch creates an instance of the corresponding Function class and calls its forward method. The output tensors returned by the forward method are associated with the grad_fn property, which allows PyTorch to trace the computation and compute gradients during the backward pass.<br \/>It is also worth noting that, by default, all tensors created by PyTorch have requires_grad=False, which means that gradients with respect to these tensors will not be computed during the backward pass. However, by setting requires_grad=True on a tensor, gradients will be computed, allowing the tensor to be used in the backward pass and allowing the user to update the model's parameters using the optimizer.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ubcc0\ud654\ub3c4 \uacc4\uc0b0\ud558\uae30(Computing gradients)<\/h3>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc2e0\uacbd\ub9dd\uc5d0\uc11c \ub9e4\uac1c\ubcc0\uc218\uc758 \uac00\uc911\uce58\ub97c \ucd5c\uc801\ud654\ud558\ub824\uba74 \ub9e4\uac1c\ubcc0\uc218\uc5d0 \ub300\ud55c \uc190\uc2e4 \ud568\uc218\uc758 \ub3c4\ud568\uc218 \uacc4\uc0b0 \ud544\uc694<\/li>\n<li>\uc774\ub97c \uc704\ud574\uc11c\ub294 x \ubc0f y\uc758 \uc77c\ubd80 \uace0\uc815 \uac12\uc5d0 \ub300\ud574 &part;loss\/&part;w \ubc0f &part;loss\/&part;b \uacc4\uc0b0 \ud544\uc694<\/li>\n<li>\uc774\ub7ec\ud55c \ub3c4\ud568\uc218\ub97c \uacc4\uc0b0\ud558\uae30 \uc704\ud574 loss.backward() \uba54\uc11c\ub4dc\uac00 \ud638\ucd9c\ub428<\/li>\n<li>\uc774\ud6c4 w.grad \ubc0f b.grad\uc5d0\uc11c \uac12 \uac80\uc0c9 \uac00\ub2a5<\/li>\n<\/ol>\n<pre id=\"code_1674111250888\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>loss.backward()\nprint(w.grad)\nprint(b.grad)\n\ntensor([[0.2739, 0.0490, 0.3279],\n        [0.2739, 0.0490, 0.3279],\n        [0.2739, 0.0490, 0.3279],\n        [0.2739, 0.0490, 0.3279],\n        [0.2739, 0.0490, 0.3279]])\ntensor([0.2739, 0.0490, 0.3279])<\/code><\/pre>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uae30\uc6b8\uae30(gradient, grad \uc18d\uc131)\ub294 require_grad \uc18d\uc131\uc774 True\ub85c \uc124\uc815\ub41c \uacc4\uc0b0 \uadf8\ub798\ud504\uc758 \ub9ac\ud504 \ub178\ub4dc\uc5d0\ub9cc \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<li>\uadf8\ub798\ud504\uc758 \ub2e4\ub978 \ub178\ub4dc\uc5d0\ub294 \uadf8\ub77c\ub514\uc5b8\ud2b8\ub97c \uc0ac\uc6a9\ud560 \uc218 \uc5c6\uc74c<\/li>\n<li>backward\ub97c \uc0ac\uc6a9\ud55c \uae30\uc6b8\uae30 \uacc4\uc0b0\uc740 \uc131\ub2a5\uc0c1\uc758 \uc774\uc720\ub85c \uc9c0\uc815\ub41c \uadf8\ub798\ud504\uc5d0\uc11c \ud55c \ubc88\ub9cc \uc218\ud589\ud560 \uc218 \uc788\uc74c<\/li>\n<li>\ub3d9\uc77c\ud55c \uadf8\ub798\ud504\uc5d0\uc11c \uc5ec\ub7ec \uac1c\uc758 backward \ud638\ucd9c(calls)\uc774 \ud544\uc694\ud55c \uacbd\uc6b0 retain_graph=True \ub9e4\uac1c\ubcc0\uc218\ub97c backward \ud638\ucd9c\uc5d0 \uc804\ub2ec\ud574\uc57c \ud568<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\ud0a4\uc6cc\ub4dc: <span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">&part;loss\/&part;w, &part;loss\/&part;b, loss.backward(), w.grad, b.grad, leaf nodes, root node, backward call, retain_graph=True<\/span><\/p>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>&part;loss\/&part;w, &part;loss\/&part;b:<br \/>\uac01\uac01 weight \ubc0f bias \ubcc0\uc218\uc5d0 \ub300\ud55c \uc190\uc2e4 \ud568\uc218\uc758 \ud3b8\ub3c4\ud568\uc218(partial derivatives)<br \/>weight \ub610\ub294 bias\uc758 \uc791\uc740 \ubcc0\ud654(small change)\uc5d0 \ub300\ud55c \uc190\uc2e4 \ud568\uc218\uc758 \ubcc0\ud654\ub97c \ub098\ud0c0\ub0c4<\/li>\n<li>loss.backward():<br \/>PyTorch\uc5d0\uc11c \uc190\uc2e4 \ubcc0\uc218(loss variable)\ub85c \ud45c\uc2dc\ub418\ub294 \uacc4\uc0b0\uc758 \uc785\ub825 \ud150\uc11c\uc5d0 \ub300\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \uba54\uc11c\ub4dc<br \/>loss variable\uc758 grad_fn \uc18d\uc131\uacfc \uc5f0\uacb0\ub41c Function \ud074\ub798\uc2a4\uc758 backward \uba54\uc11c\ub4dc\ub85c \uacc4\uc0b0\ub41c \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uc0ac\uc6a9\ud568<br \/><span style=\"background-color: #f7f7f8; color: #374151;\"> It uses the gradients computed by the backward method of the <\/span>Function<span style=\"background-color: #f7f7f8; color: #374151;\"> class associated with the <\/span>grad_fn<span style=\"background-color: #f7f7f8; color: #374151;\"> property of the loss variable.<\/span><\/li>\n<li>w.grad, b.grad:<br \/><span>\uac01\uac01 weight \ubc0f bias \ubcc0\uc218\uc758<span>&nbsp;<\/span><\/span>\uae30\uc6b8\uae30<br \/>\uc5ed\ubc29\ud5a5 \ud328\uc2a4 \ud6c4 PyTorch\uc5d0 \uc758\ud574 \uc790\ub3d9\uc73c\ub85c \uacc4\uc0b0\ub428<br \/>\uc635\ud2f0\ub9c8\uc774\uc800\ub97c \uc0ac\uc6a9\ud558\uc5ec \uac00\uc911\uce58 \ubc0f \ud3b8\ud5a5 \ubcc0\uc218\ub97c \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \ub370 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<li>leaf node: <br \/>\ud558\uc704 \ub178\ub4dc\uac00 \uc5c6\ub294 \uacc4\uc0b0 \uadf8\ub798\ud504\uc758 \ud150\uc11c<br \/>\uadf8\ub798\ud504\uc5d0 \ub300\ud55c \uc785\ub825 \ud150\uc11c\uc774\uba70 \uc77c\ubc18\uc801\uc73c\ub85c require_grad=True\ub97c \uac00\uc9d0<\/li>\n<li>root node:<br \/>backward pass\uac00 \uc2dc\uc791\ub418\ub294 \ud150\uc11c<br \/>\uc77c\ubc18\uc801\uc73c\ub85c loss \ud150\uc11c\uc774\uba70 forward pass\uc758 \ub9c8\uc9c0\ub9c9 \uc5f0\uc0b0<\/li>\n<li>backward call: <br \/>backward() call\uc740 \ub8e8\ud2b8 \ub178\ub4dc\uc5d0\uc11c \uc2dc\uc791\ud558\uc5ec \uacc4\uc0b0 \uadf8\ub798\ud504\uc758 \ubaa8\ub4e0 \ud150\uc11c\uc5d0 \ub300\ud55c \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<br \/>\uadf8\ub798\ud504\ub97c \uc5ed\ubc29\ud5a5\uc73c\ub85c \ud0d0\uc0c9\ud558\uace0 \ubbf8\ubd84\uc758 \uc5f0\uc1c4 \ubc95\uce59(chain rule)\uc744 \uc0ac\uc6a9\ud558\uc5ec \uadf8\ub798\ub514\uc5b8\ud2b8\ub97c \uacc4\uc0b0<\/li>\n<li>retain_graph=True:<br \/>backward() \uba54\uc11c\ub4dc\uc5d0 \uc804\ub2ec\ud560 \uc218 \uc788\ub294 \uc120\ud0dd\uc801 \ub9e4\uac1c\ubcc0\uc218<br \/>\uae30\ubcf8\uc801\uc73c\ub85c retain_graph=False\ub294 \ubc31\uc6cc\ub4dc \ud328\uc2a4 \ud6c4\uc5d0 \uadf8\ub798\ud504\uac00 \ud574\uc81c\ub428\uc744 \uc758\ubbf8<br \/>retain_graph=True\uc774\uba74 \uadf8\ub798\ud504\uac00 \uc720\uc9c0\ub418\uace0 \ub2e4\ub978 \uc5ed\ubc29\ud5a5 \uc804\ub2ec\uc5d0 \uc0ac\uc6a9\ub420 \uc218 \uc788\uc74c<\/li>\n<\/ul>\n<\/div>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<h3 data-ke-size=\"size23\">\ubcc0\ud654\ub3c4 \ucd94\uc801 \uba48\ucd94\uae30(Disabling gradient tracking)<\/h3>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uae30\ubcf8\uc801\uc73c\ub85c, require_grad=True\uc778 \ud150\uc11c\ub294 \uacc4\uc0b0 \uae30\ub85d\uc744 \ucd94\uc801\ud558\uace0 \uadf8\ub798\ub514\uc5b8\ud2b8 \uacc4\uc0b0\uc744 \uc9c0\uc6d0<\/li>\n<li>\ud559\uc2b5\ub41c \ubaa8\ub378\uc744 \uc785\ub825 \ub370\uc774\ud130\uc5d0 \uc801\uc6a9\ub9cc \ud558\ub294 \uc21c\uc804\ud30c \uc5f0\uc0b0\ub9cc \ud544\uc694\ud55c \uacbd\uc6b0\uc640 \uac19\uc774 \ucd94\uc801 \ubc0f \uadf8\ub798\ub514\uc5b8\ud2b8 \uacc4\uc0b0\uc774 \ud544\uc694\ud558\uc9c0 \uc54a\uc740 \uacbd\uc6b0 \uc874\uc7ac<\/li>\n<li>\uc774\ub7ec\ud55c \uacbd\uc6b0 \ucf54\ub4dc\ub97c torch.no_grad() \ube14\ub85d\uc73c\ub85c \ub458\ub7ec\uc2f8\uc11c \ucd94\uc801 \uc911\uc9c0\ud560 \uc218 \uc788\uc74c<\/li>\n<li>detach() \uba54\uc18c\ub4dc\ub3c4 \ub3d9\uc77c\ud558\uac8c \ucd94\uc801 \uc911\uc9c0\ud560 \uc218 \uc788\uc74c<\/li>\n<\/ul>\n<pre id=\"code_1674114411661\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>z = torch.matmul(x, w)+b\nprint(z.requires_grad)\n\nwith torch.no_grad():\n    z = torch.matmul(x, w)+b\nprint(z.requires_grad)\n\nz = torch.matmul(x, w)+b\nz_det = z.detach()\nprint(z_det.requires_grad)\n\nTrue\nFalse\nFalse<\/code><\/pre>\n<p data-ke-size=\"size16\">\ubcc0\ud654\ub3c4 \ucd94\uc801\uc744 \uba48\ucdb0\uc57c \ud558\ub294 \uc774\uc720\ub4e4<\/p>\n<ol style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc2e0\uacbd\ub9dd\uc758 \uc77c\ubd80 \ub9e4\uac1c\ubcc0\uc218\ub97c \uace0\uc815\uc73c\ub85c \ud45c\uc2dc(frozen parameter)<br \/>\uc774\ub294 <a href=\"https:\/\/tutorials.pytorch.kr\/beginner\/finetuning_torchvision_models_tutorial.html\" target=\"_blank\" rel=\"noopener\">\uc0ac\uc804 \ud6c8\ub828\ub41c \ub124\ud2b8\uc6cc\ud06c\ub97c \ubbf8\uc138 \uc870\uc815\ud558\uae30 \uc704\ud55c \uc77c\ubc18\uc801\uc778 \uc2dc\ub098\ub9ac\uc624<\/a><\/li>\n<li>\uadf8\ub798\ub514\uc5b8\ud2b8 \ucd94\uc801\uc774 \uc5c6\ub294 \ud150\uc11c\uc5d0 \ub300\ud55c \uacc4\uc0b0\uc774 \ub354 \ud6a8\uc728\uc801\uc774\uae30 \ub54c\ubb38\uc5d0 \uc21c\ubc29\ud5a5 \ud328\uc2a4\ub9cc \uc218\ud589\ud560 \ub54c \uacc4\uc0b0 \uc18d\ub3c4 \ud5a5\uc0c1<\/li>\n<\/ol>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<div>\n<div>\n<p data-ke-size=\"size16\">Keywords: <span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">torch.no_grad() block, gradient tracking, frozen parameter, fine tuning<\/span><\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>torch.no_grad() block:<br \/>\uadf8\ub798\ub514\uc5b8\ud2b8 \ucd94\uc801\uc744 \uc77c\uc2dc\uc801\uc73c\ub85c \ube44\ud65c\uc131\ud654\ud558\ub294 PyTorch\uc5d0\uc11c \uc81c\uacf5\ud558\ub294 \ucee8\ud14d\uc2a4\ud2b8 \uad00\ub9ac\uc790<br \/>\uadf8\ub798\ub514\uc5b8\ud2b8 \ucd94\uc801\uc774 \ube44\ud65c\uc131\ud654\ub418\uba74 \ud3ec\uc6cc\ub4dc \ud328\uc2a4\uac00 \ud3c9\uc18c\uc640 \uac19\uc774 \uc218\ud589\ub428<br \/>\uadf8\ub7ec\ub098 \uadf8\ub798\ub514\uc5b8\ud2b8\uac00 \uacc4\uc0b0\ub418\uc9c0 \uc54a\uace0 \ud150\uc11c\uc758 grad \uc18d\uc131\uc774 \uc5c5\ub370\uc774\ud2b8\ub418\uc9c0 \uc54a\uc74c<br \/>\uadf8\ub798\ub514\uc5b8\ud2b8\uac00 \ud544\uc694\ud558\uc9c0 \uc54a\uc744 \ub54c \uacc4\uc0b0 \uc18d\ub3c4\ub97c \ub192\uc774\uace0 \uba54\ubaa8\ub9ac\ub97c \uc808\uc57d\ud558\ub294 \ub370 \uc720\uc6a9\ud560 \uc218 \uc788\uc74c<\/li>\n<li>gradient tracking:<br \/>backward pass \ub3d9\uc548 \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\uc640 \uad00\ub828\ud558\uc5ec loss \ud568\uc218\uc758 \uae30\uc6b8\uae30\ub97c \uacc4\uc0b0\ud558\uace0 \uc800\uc7a5\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>PyTorch\uc758 autograd \ud328\ud0a4\uc9c0\uc5d0 \uc758\ud574 \uc218\ud589\ub418\uba70 requires_grad=True\uc778 \ubaa8\ub4e0 \ud150\uc11c\uc5d0 \ub300\ud574 \uae30\ubcf8\uc801\uc73c\ub85c \ud65c\uc131\ud654<\/li>\n<li>frozen parameter: <br \/>\uae30\uc6b8\uae30 \ucd94\uc801\uc774 \ube44\ud65c\uc131\ud654\ub41c \ub9e4\uac1c\ubcc0\uc218<br \/>backward pass \uc911\uc5d0 \uc5c5\ub370\uc774\ud2b8\ud560 \uc218 \uc5c6\uc74c<br \/>require_grad=False\ub97c \uc124\uc815\ud558\uac70\ub098 \ub9e4\uac1c\ubcc0\uc218\ub97c torch.no_grad() \ube14\ub85d \ub0b4\ubd80\uc5d0 \ubc30\uce58\ud558\uc5ec \uc218\ud589\ud560 \uc218 \uc788\uc74c<\/li>\n<li>fine tuning:<br \/>\uc0c8\ub85c\uc6b4 \uc791\uc5c5\uc5d0 \ub300\ud574 \uc0ac\uc804 \ud6c8\ub828\ub41c \ubaa8\ub378\uc744 \ud6c8\ub828\ud558\uac70\ub098 \uc5c5\ub370\uc774\ud2b8\ud558\ub294 \ud504\ub85c\uc138\uc2a4<br \/>\uace0\uc815\ub41c \ub9e4\uac1c\ubcc0\uc218\uc758 \uc77c\ubd80 \ub610\ub294 \uc804\ubd80\ub97c \uace0\uc815 \ud574\uc81c\ud558\uace0 \ud6c8\ub828 \uc911\uc5d0 \uc5c5\ub370\uc774\ud2b8\ud558\uc5ec \uc218\ud589\ud560 \uc218 \uc788\uc74c<br \/>\uc0ac\uc804 \ud6c8\ub828\ub41c \ubaa8\ub378\uc774 \uc0c8 \uc791\uc5c5\uc5d0 \uc720\uc6a9\ud55c \uae30\ub2a5\uc744 \ud559\uc2b5\ud588\uace0 \uc774 \uc9c0\uc2dd\uc744 \ud65c\uc6a9\ud558\uc5ec \uc131\ub2a5\uc744 \ud5a5\uc0c1\uc2dc\ud0a4\ub824\ub294 \uacbd\uc6b0\uc5d0 \uc720\uc6a9<\/li>\n<\/ul>\n<div>\n<div>\n<div>\n<div>\n<div>\n<div>&nbsp;<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<div>&nbsp;<\/div>\n<\/div>\n<h3 data-ke-size=\"size23\">\uc5f0\uc0b0 \uadf8\ub798\ud504\uc5d0 \ub300\ud55c \ucd94\uac00 \uc815\ubcf4(more on Computational Graphs)<\/h3>\n<div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>autograd\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>DAG\uc5d0 \ub370\uc774\ud130(\ud150\uc11c) \uae30\ub85d<\/li>\n<li>DAG\uc5d0 \uc2e4\ud589\ub41c \ubaa8\ub4e0 \uc5f0\uc0b0\ub4e4(\uacb0\uacfc \ud150\uc11c \uc0dd\uc131 \ud3ec\ud568) \uae30\ub85d<\/li>\n<li>\ubc29\ud5a5\uc131 \ube44\uc21c\ud658 \uadf8\ub798\ud504(DAG): Function \uac1d\uccb4\ub85c \uad6c\uc131\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc78e(leaves): \uc785\ub825 \ud150\uc11c<\/li>\n<li>\ubfcc\ub9ac(root): \uacb0\uacfc \ud150\uc11c<\/li>\n<li>\uadf8\ub798\ud504 \ucd94\uc801: \uccb4\uc778 \uaddc\uce59\uc744 \uc0ac\uc6a9\ud558\uc5ec \ubcc0\ud654\ub3c4 \uc790\ub3d9 \uacc4\uc0b0<\/li>\n<\/ul>\n<\/li>\n<li>\uc21c\uc804\ud30c \ub2e8\uacc4(forward pass):\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc694\uccad\ub41c \uc5f0\uc0b0 \uc218\ud589: \uacb0\uacfc \ud150\uc11c\ub97c \uacc4\uc0b0&nbsp;<\/li>\n<li>\ubcc0\ud654\ub3c4 \uae30\ub2a5(gradient function): DAG\uc5d0\uc11c \uc720\uc9c0<\/li>\n<\/ul>\n<\/li>\n<li>\uc5ed\uc804\ud30c \ub2e8\uacc4(backward pass):\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>.backward()\uac00 DAG \ub8e8\ud2b8\uc5d0\uc11c \ud638\ucd9c\ub420 \ub54c \uc2dc\uc791<\/li>\n<li>\ubcc0\ud654\ub3c4 \uacc4\uc0b0: \uac01 .grad_fn\uc5d0\uc11c<\/li>\n<li>\ubcc0\ud654\ub3c4 \ub204\uc801(accumulate): \uac01 \ud150\uc11c\uc758 .grad \uc18d\uc131\uc5d0\uc11c<\/li>\n<li>\uc804\ud30c(propagate): \uccb4\uc778 \uaddc\uce59\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc78e(leaf) \ud150\uc11c\uae4c\uc9c0<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<\/li>\n<li>PyTorch\uc758 DAG: \ub3d9\uc801(dynamic)\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uadf8\ub798\ud504 \uc7ac\uc0dd\uc131(recreate) \ubc0f \ucc44\uc6c0(populate): \uac01 .backward() \ud638\ucd9c \ud6c4 \ucc98\uc74c\ubd80\ud130<\/li>\n<li>\ubaa8\ub378\uc5d0\uc11c \uc81c\uc5b4 \ud750\ub984(control flow) \ubb38 \ud5c8\uc6a9\ub428<\/li>\n<li>\ubaa8\uc591(shape), \ud06c\uae30(size), \uc5f0\uc0b0(operation): \ubaa8\ub4e0 \ubc18\ubcf5(iteration)\uc5d0\uc11c \ubcc0\uacbd\ub420 \uc218 \uc788\uc74c<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">\ud0a4\uc6cc\ub4dc: autograd, directed acyclic graph, chain rule, dynamic graph, allowing control flow statements, computation vs operation, cycle vs loop<\/p>\n<\/div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>Autograd: <br \/>\ud150\uc11c\uc758 \ubaa8\ub4e0 \uc791\uc5c5\uc5d0 \ub300\ud55c \uc790\ub3d9 \ubbf8\ubd84\ub97c \uc81c\uacf5\ud558\ub294 PyTorch \ubaa8\ub4c8<br \/>\uc2e0\uacbd\ub9dd\uc758 backward pass \uc911\uc5d0 \ubcc0\ud654\ub3c4\ub97c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\uc5b4 \ud6c8\ub828 \uc911\uc5d0 \ubcc0\ud654\ub3c4\ub97c \ud6a8\uc728\uc801\uc73c\ub85c \uacc4\uc0b0\ud560 \uc218 \uc788\uc74c<\/li>\n<li>Directed Acyclic Graph(DAG):<br \/>edge\uac00 \ubc29\ud5a5\uc744 \uac00\uc838 \ubc29\ud5a5\uc774 \uc788\ub294 \uadf8\ub798\ud504\ub97c \ud615\uc131\ud558\ub294 \uc77c\uc885\uc758 \uadf8\ub798\ud504 \ub370\uc774\ud130 \uad6c\uc870<br \/>cycle\uc774\ub098 loop\uac00 \uc5c6\uc74c<\/li>\n<li>\uc5f0\uc1c4 \ubc95\uce59(chain rule):<br \/>\ud569\uc131 \ud568\uc218\uc758 \ub3c4\ud568\uc218\ub294 \uad6c\uc131 \ud568\uc218\uc758 \ub3c4\ud568\uc218\ub97c \uacf1\ud558\uc5ec \uacc4\uc0b0\ud560 \uc218 \uc788\ub2e4\ub294 \ubbf8\uc801\ubd84\ud559\uc758 \uae30\ubcf8 \uac1c\ub150<br \/>\uc2e0\uacbd\ub9dd\uc5d0\uc11c \uccb4\uc778 \uaddc\uce59\uc740 \ub124\ud2b8\uc6cc\ud06c\uc758 \uac01 \uacc4\uce35\uc5d0 \uaddc\uce59\uc744 \ubc18\ubcf5\uc801\uc73c\ub85c \uc801\uc6a9\ud558\uc5ec \uc5ed\ubc29\ud5a5 \ud1b5\uacfc \uc911\uc5d0 \uae30\uc6b8\uae30\ub97c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>\ub3d9\uc801 \uadf8\ub798\ud504(dynamic graph):<br \/>\ub3d9\uc801\uc73c\ub85c \ubcc0\uacbd\ub420 \uc218 \uc788\ub294 \uadf8\ub798\ud504<br \/>\uc989, \ub7f0\ud0c0\uc784 \uc911\uc5d0 \uadf8\ub798\ud504\uac00 \ubcc0\uacbd\ub420 \uc218 \uc788\uc74c<br \/>\ub354 \ub9ce\uc740 \uc720\uc5f0\uc131\uc744 \ud5c8\uc6a9\ud558\uace0 \ub124\ud2b8\uc6cc\ud06c\uc758 \uc21c\ubc29\ud5a5 \uc804\ub2ec\uc5d0\uc11c \ub8e8\ud504 \ubc0f \uc870\uac74\ubb38\uacfc \uac19\uc740 \uc81c\uc5b4 \ud750\ub984 \ubb38\uc744 \uc0ac\uc6a9\ud560 \uc218 \uc788\uac8c \ud568<\/li>\n<li>\uc81c\uc5b4 \ud750\ub984 \ubb38 \ud5c8\uc6a9(allowing control flow statement):<br \/>Pytorch\ub294 \ub3d9\uc801 \uadf8\ub798\ud504 \uae30\ub2a5\uc744 \uc0ac\uc6a9\ud558\uc5ec \ub124\ud2b8\uc6cc\ud06c\uc758 \uc21c\ubc29\ud5a5 \uc804\ub2ec\uc5d0\uc11c \uc81c\uc5b4 \ud750\ub984 \ubb38\uc744 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc74c<br \/>\uc774\ub97c \ud1b5\ud574 \ub354 \ubcf5\uc7a1\ud55c \uc2e0\uacbd\ub9dd \uc544\ud0a4\ud14d\ucc98\ub97c \ub9cc\ub4e4 \uc218 \uc788\uc9c0\ub9cc \uacc4\uc0b0 \uadf8\ub798\ud504\uac00 \ub354 \ubcf5\uc7a1\ud574\uc9c0\uace0 \uc131\ub2a5\uc5d0 \uc601\ud5a5\uc744 \ubbf8\uce60 \uc218 \uc788\uc74c<\/li>\n<li>&nbsp;computation vs operation:<br \/>computation\uc740 \ucd5c\uc885 \ucd9c\ub825\uc744 \uc0dd\uc131\ud558\uae30 \uc704\ud574 \ud150\uc11c\uc5d0\uc11c \uc218\ud589\ub418\ub294 sequence of operations<br \/>\uc2e0\uacbd\ub9dd\uc5d0\uc11c\uc758 layer \ubc0f activation function \ub4f1\uc744 \ud1b5\ud55c forward pass<br \/>operation\uc740 \ud589\ub82c\uacf1, non-linear activation function\uacfc \uac19\uc740 computation\uc758 \ub2e8\uc77c \ub2e8\uacc4<\/li>\n<li>cycle vs loop<br \/>cycle\uc740 \ub3d9\uc77c\ud55c \ub178\ub4dc\ub97c \uc5ec\ub7ec \ubc88 \ubc29\ubb38\ud560 \uc218 \uc788\ub294 \uadf8\ub798\ud504\uc758 \ub2eb\ud78c \uacbd\ub85c<br \/>loop\ub294 \ud2b9\uc815 \uc870\uac74\uc774 \uc2e4\ud589\ub420 \ub54c\uae4c\uc9c0 \ubc18\ubcf5\uc801\uc73c\ub85c \uc2e4\ud589\ub418\ub294 \uc77c\ub828\uc758 \uba85\ub839<br \/>DAG \ub9e5\ub77d\uc5d0\uc11c cycle\uc740 \ub3d9\uc77c\ud55c \ub178\ub4dc\ub85c\uc758 \uc7ac\ubc29\ubb38\uc744 \uc758\ubbf8, \ube44\uc8fc\uae30\uc801\uc774\ubbc0\ub85c DAG\uc5d0\uc11c \ud5c8\uc6a9\ub418\uc9c0 \uc54a\uc74c -&gt; Q) RNN\uc5d0\uc11c\ub294?<br \/><span>DAG \ub9e5\ub77d\uc5d0\uc11c<span> loop\ub294 forward pass \uc911 \ubc1c\uc0dd\ud558\ub294 \uacc4\uc0b0\uc758 \ubc18\ubcf5\uc744 \uc758\ubbf8<span><span>&nbsp;<\/span>-&gt; Q) RNN\uc5d0\uc11c\ub294?<\/span><\/span><\/span><\/li>\n<\/ul>\n<div>&nbsp;<\/div>\n<\/div>\n<\/div>\n<h3 data-ke-size=\"size23\">\uc120\ud0dd\uc801\uc73c\ub85c \uc77d\uae30: \ud150\uc11c \ubcc0\ud654\ub3c4\uc640 \uc57c\ucf54\ube44\uc548 \uacf1(Optional Reading: Tensor gradients and jacobian products)<\/h3>\n<div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc2a4\uce7c\ub77c \uc190\uc2e4 \ud568\uc218: \uc77c\ubd80 \ub9e4\uac1c\ubcc0\uc218\uc5d0 \ub300\ud55c \uae30\uc6b8\uae30<\/li>\n<li>\ucd9c\ub825 \ud568\uc218\uac00 \uc784\uc758\uc758(arbitrary) \ud150\uc11c\uc778 \uacbd\uc6b0 \uc788\uc74c<\/li>\n<li>PyTorch: Jacobian \uacf1 \uacc4\uc0b0. \uc2e4\uc81c gradient \uc544\ub2d8<\/li>\n<li>\ubca1\ud130 \ud568\uc218: y = f(x), \uc5ec\uae30\uc11c x = &lt;x1, ..., xn&gt; \ubc0f y = &lt;y1, ..., ym&gt;<\/li>\n<li>x\uc5d0 \ub300\ud55c y\uc758 \uae30\uc6b8\uae30: Jacobian \ud589\ub82c, Jij = &part;yi\/&part;xj<\/li>\n<li>Jacobian \ud589\ub82c: \uc9c1\uc811 \uacc4\uc0b0\ub418\uc9c0 \uc54a\uc74c<\/li>\n<li>jacobian \uacf1: \uc8fc\uc5b4\uc9c4 \uc785\ub825 \ubca1\ud130 v = (v1 ... vm)\uc5d0 \ub300\ud55c vT&sdot;J<\/li>\n<li>backward \ud568\uc218: v\ub97c \uc778\uc218\ub85c \uc0ac\uc6a9\ud558\uc5ec \ud638\ucd9c<\/li>\n<li>v\uc758 \ud06c\uae30(size): \uacf1\uc744 \uacc4\uc0b0\ud558\ub824\uace0 \ud558\ub294 \uc6d0\ub798 \ud150\uc11c\uc758 \ud06c\uae30\uc640 \uac19\uc544\uc57c \ud568<\/li>\n<\/ul>\n<pre id=\"code_1674184923503\" class=\"python\" data-ke-language=\"python\" data-ke-type=\"codeblock\"><code>inp = torch.eye(5, requires_grad=True)\nout = (inp+1).pow(2)\nout.backward(torch.ones_like(inp), retain_graph=True)\nprint(f\"First call\\n{inp.grad}\")\nout.backward(torch.ones_like(inp), retain_graph=True)\nprint(f\"\\nSecond call\\n{inp.grad}\")\ninp.grad.zero_()\nout.backward(torch.ones_like(inp), retain_graph=True)\nprint(f\"\\nCall after zeroing gradients\\n{inp.grad}\")\n\n\ntensor([[4., 2., 2., 2., 2.],\n        [2., 4., 2., 2., 2.],\n        [2., 2., 4., 2., 2.],\n        [2., 2., 2., 4., 2.],\n        [2., 2., 2., 2., 4.]])\n\nSecond call\ntensor([[8., 4., 4., 4., 4.],\n        [4., 8., 4., 4., 4.],\n        [4., 4., 8., 4., 4.],\n        [4., 4., 4., 8., 4.],\n        [4., 4., 4., 4., 8.]])\n\nCall after zeroing gradients\ntensor([[4., 2., 2., 2., 2.],\n        [2., 4., 2., 2., 2.],\n        [2., 2., 4., 2., 2.],\n        [2., 2., 2., 4., 2.],\n        [2., 2., 2., 2., 4.]])<\/code><\/pre>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\ub3d9\uc77c\ud55c \uc778\uc790\ub85c backward \ub450 \ucc28\ub840 \ud638\ucd9c\ud558\uba74 \ubcc0\ud654\ub3c4 \uac12 \ub2ec\ub77c\uc9d0<\/li>\n<li>pytorch\uac00 backward propagation \uc218\ud589 \uc2dc \ubcc0\ud654\ub3c4\ub97c \ub204\uc801\ud558\uae30 \ub54c\ubb38\uc5d0 \ubc1c\uc0dd<\/li>\n<li>\uacc4\uc0b0 \uadf8\ub798\ud504\uc758 \ubaa8\ub4e0 \ub9ac\ud504 \ub178\ub4dc\uc758 grad \uc18d\uc131\uc5d0 \ub354\ud574\uc9d0<\/li>\n<li>\uc801\uc808\ud55c \ubcc0\ud654\ub3c4 \uacc4\uc0b0\uc744 \uc704\ud574 grad \uc18d\uc131\uc744 0\uc73c\ub85c \ub9cc\ub4e4\uc5b4\uc57c \ud568<\/li>\n<li>\uc635\ud2f0\ub9c8\uc774\uc800: \uc2e4\uc81c \ud559\uc2b5 \uc2dc grad \uc18d\uc131\uc744 0\uc73c\ub85c \ub9cc\ub4dc\ub294 \ub370 \ub3c4\uc640\uc90c<\/li>\n<\/ul>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>backward() \ud568\uc218: \uc704 \ucf54\ub4dc\uc5d0\uc11c \ub9e4\uac1c\ubcc0\uc218 \uc5c6\uc774 \ud638\ucd9c\ub428<\/li>\n<li>backward(torch.tensor(1.0)) \ud638\ucd9c\uacfc \ub3d9\uc77c<\/li>\n<li>\uc2e0\uacbd\ub9dd \ud6c8\ub828 \uc911 loss\uc640 \uac19\uc740 \uc2a4\uce7c\ub77c \uac12 \ud568\uc218\uc758 \ubcc0\ud654\ub3c4 \uacc4\uc0b0\uc5d0 \uc720\uc6a9\ud568<\/li>\n<li>\ud0a4\uc6cc\ub4dc: \uc5ed\uc804\ud30c, Pytorch, \uadf8\ub798\ub514\uc5b8\ud2b8, \ub204\uc0b0, \uc635\ud2f0\ub9c8\uc774\uc800, \uc2a4\uce7c\ub77c \uac12 \ud568\uc218, \uc2e0\uacbd\ub9dd \ud6c8\ub828.<\/li>\n<\/ul>\n<\/div>\n<p data-ke-size=\"size16\">&nbsp;<\/p>\n<p data-ke-size=\"size16\">Keywords: scalar loss function, <span>Jacobian matrix, vector function,<span>&nbsp;<\/span><\/span>Jacobian product, shape vs size on tensor, scalar-valued function, optimizer<\/p>\n<\/div>\n<div>\n<ul style=\"list-style-type: disc;\" data-ke-list-type=\"disc\">\n<li>\uc2a4\uce7c\ub77c \uc190\uc2e4 \ud568\uc218(Scalar Loss Function): <br \/>&nbsp;\ubaa8\ub378\uc758 \uc608\uce21\uac12\uacfc \uc815\ub2f5\uac12\uc744 \ucde8\ud558\uc5ec \uc774\ub4e4 \uc0ac\uc774\uc758 \ubd88\uc77c\uce58(discrepancy, \ucc28\uc774)\ub97c \uce21\uc815 \ud6c4 \uc2a4\uce7c\ub77c \uac12\uc744 \ubc18\ud658\ud558\ub294 \ud568\uc218<br \/>\uc608) Mean Squared Error, Binary Cross-Entropy<br \/>\ubaa8\ub378\uc758 \uc608\uce21 \ucd9c\ub825\uacfc \uc2e4\uc81c \ucd9c\ub825 \uac04\uc758 \ucc28\uc774\ub97c \ucd5c\uc18c\ud654\ud558\uc5ec \uc2e0\uacbd\ub9dd\uc744 \ud6c8\ub828\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>\uc790\ucf54\ube44\uc548 \ud589\ub82c(Jacobian matrix):<br \/>\ubca1\ud130 \ud568\uc218\uc758 \ubaa8\ub4e0 1\ucc28 \ud3b8\ubbf8\ubd84 \ud589\ub82c<br \/>\uc2e0\uacbd\ub9dd\uc5d0\uc11c\ub294 \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\uc5d0 \ub300\ud55c \uc190\uc2e4 \ud568\uc218\uc758 \uae30\uc6b8\uae30\ub97c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<br \/>\uc2e0\uacbd\ub9dd\uc758 \uc5ed\ubc29\ud5a5 \ud1b5\uacfc \ub3d9\uc548 \uae30\uc6b8\uae30\ub97c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \uc790\ub3d9 \ubbf8\ubd84\uc758 \uae30\ubcf8 \uac1c\ub150<\/li>\n<li><span>\ubca1\ud130 \ud568\uc218(vector function):<br \/>\ud558\ub098 \uc774\uc0c1\uc758 \uc785\ub825\uc744 \ubc1b\uc544 \ubca1\ud130 \ucd9c\ub825\uc744 \ubc18\ud658\ud558\ub294 \ud568\uc218<br \/>\uc2e0\uacbd\ub9dd\uc5d0\uc11c \ubaa8\ub378\uc758 \uc815\ubc29\ud5a5 \uc804\ub2ec(forward pass)\uc740 \uc785\ub825\uc774 \ubca1\ud130\uc774\uace0 \ucd9c\ub825\ub3c4 \ubca1\ud130\uc778 \ubca1\ud130 \ud568\uc218\ub85c \uc0dd\uac01\ud560 \uc218 \uc788\uc74c<\/span><\/li>\n<li>\uc790\ucf54\ube44\uc548 \uacf1(Jacobian Product): <br \/>\uc790\ucf54\ube44\uc548 \ud589\ub82c\uacfc \ubca1\ud130\uc758 \uacf1<br \/>\uc790\ucf54\ube44\uc548 \ud589\ub82c\uc740 \ubca1\ud130 \ud568\uc218\uc758 \ubaa8\ub4e0 1\uacc4 \ud3b8\ub3c4\ud568\uc218\uc758 \ud589\ub82c<br \/>\ubca1\ud130\ub294 \uadf8\ub798\ub514\uc5b8\ud2b8\uac00 \uacc4\uc0b0\ub418\ub294 \ubc29\ud5a5<br \/>\ud2b9\uc815 \ubc29\ud5a5\uc758 \uc785\ub825\uc5d0 \ub300\ud55c \ud568\uc218\uc758 \uae30\uc6b8\uae30\ub97c \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>Tensor\uc758 \ubaa8\uc591 \ub300 \ud06c\uae30(shape vs size):<br \/>PyTorch\uc5d0\uc11c Tensor\ub294 \ub2e4\ucc28\uc6d0 \ubc30\uc5f4<br \/>\ud150\uc11c\uc758 \ubaa8\uc591\uc740 \ud150\uc11c\uc758 \uac01 \ucc28\uc6d0\uc758 \ud06c\uae30\ub97c \ub098\ud0c0\ub0b4\ub294 \ud29c\ud50c<br \/>\ud150\uc11c\uc758 \ud06c\uae30\ub294 \ubaa8\uc591\uc758 \uacf1\uc778 \ud150\uc11c\uc758 \ucd1d \uc694\uc18c \uc218<br \/>\uc608) \ud150\uc11c\uc758 \ubaa8\uc591\uc774 (2,3)\uc774\ub77c\uba74 \ud06c\uae30\ub294 6<br \/>\ubaa8\uc591\uc740 \ub370\uc774\ud130\uc758 \uad6c\uc870\ub97c \uc774\ud574\ud558\ub294 \ub370 \uc720\uc6a9<br \/>\ud06c\uae30\ub294 \ub370\uc774\ud130\uc758 \uc591\uc744 \uc774\ud574\ud558\ub294 \ub370 \uc720\uc6a9<\/li>\n<li>\uc2a4\uce7c\ub77c \uac12 \ud568\uc218(scalar-valued function):<br \/>\ud558\ub098 \uc774\uc0c1\uc758 \uc785\ub825\uc744 \ubc1b\uc544 \ub2e8\uc77c \uc2a4\uce7c\ub77c \ucd9c\ub825\uc744 \ubc18\ud658\ud558\ub294 \ud568\uc218<br \/>\uba38\uc2e0\ub7ec\ub2dd\uc5d0\uc11c cost \ub610\ub294 loss function\uc744 \ub098\ud0c0\ub0b4 \uc2e0\uacbd\ub9dd\uc744 \ud6c8\ub828\ud558\ub294 \ub370 \uc0ac\uc6a9\ub428<\/li>\n<li>\uc635\ud2f0\ub9c8\uc774\uc800(optimizer): <br \/><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\uc190\uc2e4 \ud568\uc218\uc640 \uac19\uc740 \uc2a4\uce7c\ub77c \uac12 \ud568\uc218\ub97c \ucd5c\uc18c\ud654\ud558\uae30 \uc704\ud574 \ubaa8\ub378\uc758 \ub9e4\uac1c\ubcc0\uc218\ub97c \uc870\uc815\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \uc54c\uace0\ub9ac\uc998<br \/><\/span><span style=\"font-family: -apple-system, BlinkMacSystemFont, 'Helvetica Neue', 'Apple SD Gothic Neo', Arial, sans-serif; letter-spacing: 0px;\">\uc608) Stochastic Gradient Descent(SGD), Adam, Adagrad<br \/><\/span>learning rate \ub4f1\uc758 \u200b\u200b\uc790\uccb4 \ud558\uc774\ud37c \ub9e4\uac1c\ubcc0\uc218 \uc138\ud2b8 \uc874\uc7ac<\/li>\n<\/ul>\n<\/div>\n<\/div>","category":"Pytorch\/\ud29c\ud1a0\ub9ac\uc5bc","author":"\ub300\ub450\ucf54\uae30","guid":"https:\/\/hoohaha.tistory.com\/131","comments":"https:\/\/hoohaha.tistory.com\/131#entry131comment","pubDate":"Fri, 20 Jan 2023 12:45:38 +0900"}]}}