{"id":1313,"date":"2023-04-18T16:27:29","date_gmt":"2023-04-18T07:27:29","guid":{"rendered":"http:\/\/edu.ujhb.org\/?p=1313"},"modified":"2023-04-18T16:27:30","modified_gmt":"2023-04-18T07:27:30","slug":"mlflow%e5%85%a5%e9%96%80%ef%bc%88%e6%97%a5%e6%9c%ac%e8%aa%9e%e7%89%88%ef%bc%89","status":"publish","type":"post","link":"https:\/\/edu.ujhb.org\/?p=1313","title":{"rendered":"MLFlow\u5165\u9580\uff08\u65e5\u672c\u8a9e\u7248\uff09"},"content":{"rendered":"\n<p>IT\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u306e\u30d3\u30c7\u30aa\u3092\u3084\u308a\u305f\u3044\u306e\u3067\u3059\u304c\u3001\u30a6\u30a7\u30d6\u30b5\u30a4\u30c8\u306e\u30b3\u30f3\u30c6\u30f3\u30c4\u306b\u5fdc\u3058\u3066\u30d3\u30c7\u30aa\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u306e\u30c6\u30ad\u30b9\u30c8\u30b3\u30f3\u30c6\u30f3\u30c4\u3092\u914d\u7f6e\u3057\u3066\u3044\u305f\u3060\u3051\u307e\u3059\u304b\u3002 MLFLOW\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u304b\u3089\u59cb\u3081\u307e\u3057\u3087\u3046\u3001\u305d\u3057\u3066\u3042\u306a\u305f\u306f\u79c1\u306b\u30ab\u30d0\u30fc\u3059\u308b\u3082\u306e\u306e\u30ea\u30b9\u30c8\u3092\u4e0e\u3048\u308b\u3067\u3057\u3087\u3046\u3002<\/p>\n\n\n\n<p>\u3082\u3061\u308d\u3093\u3067\u3059\uff01 MLflow \u30d9\u30fc\u30b9\u306e\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u306e\u6982\u8981\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u30b3\u30fc\u30b9\u306e\u8aac\u660e\u3068\u76ee\u7684\n<ul class=\"wp-block-list\">\n<li>\u3053\u306e\u30b3\u30fc\u30b9\u306e\u76ee\u7684\u3068\u76ee\u7684<\/li>\n\n\n\n<li>MLflow \u3092\u4f7f\u7528\u3059\u308b\u7406\u7531<\/li>\n\n\n\n<li>MLflow \u306e\u5229\u70b9\u3068\u6a5f\u80fd<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>MLflow \u306e\u6982\u8981\n<ul class=\"wp-block-list\">\n<li>MLflow \u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3<\/li>\n\n\n\n<li>MLflow \u30c8\u30ec\u30fc\u30b9\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3001\u30e2\u30c7\u30eb\u3001\u30ec\u30b8\u30b9\u30c8\u30ea\u306e\u6982\u5ff5\u306e\u6982\u8981<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3068\u69cb\u6210\n<ul class=\"wp-block-list\">\n<li>MLflow \u306e\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u65b9\u6cd5<\/li>\n\n\n\n<li>\u30d0\u30c3\u30af\u30a8\u30f3\u30c9 \u30b9\u30c8\u30ec\u30fc\u30b8\u3092\u4f7f\u7528\u3057\u3066 MLflow \u306e\u74b0\u5883\u3092\u69cb\u6210\u3059\u308b<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>ML\u30d5\u30ed\u30fc\u8ffd\u8de1\n<ul class=\"wp-block-list\">\n<li>MLflow \u8ffd\u8de1\u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3059\u308b<\/li>\n\n\n\n<li>MLflow UI \u3092\u4f7f\u7528\u3057\u305f\u5b9f\u9a13\u7d50\u679c\u306e\u8868\u793a<\/li>\n\n\n\n<li>\u30c6\u30b9\u30c8\u7d50\u679c\u306e\u30af\u30a8\u30ea\u3068\u6bd4\u8f03<\/li>\n\n\n\n<li>\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u30d1\u30e9\u30e1\u30fc\u30bf\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\n<ul class=\"wp-block-list\">\n<li>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f5c\u6210\u3059\u308b<\/li>\n\n\n\n<li>ML\u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30d5\u30a1\u30a4\u30eb\u3092\u5b9a\u7fa9\u3059\u308b<\/li>\n\n\n\n<li>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u7ba1\u7406<\/li>\n\n\n\n<li>MLflow \u3092\u4f7f\u7528\u3057\u3066\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3059\u308b<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>MLflow \u30e2\u30c7\u30eb\n<ul class=\"wp-block-list\">\n<li>MLflow \u30e2\u30c7\u30eb\u3092\u4f5c\u6210\u3059\u308b<\/li>\n\n\n\n<li>\u5225\u306e\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b<\/li>\n\n\n\n<li>MLflow \u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4<\/li>\n\n\n\n<li>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u7ba1\u7406\u3059\u308b<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\n<ul class=\"wp-block-list\">\n<li>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306e\u6982\u8981<\/li>\n\n\n\n<li>\u30e2\u30c7\u30eb\u306e\u767b\u9332\u3068\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406<\/li>\n\n\n\n<li>\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u7ba1\u7406\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>MLflow \u3092\u4ed6\u306e\u30c4\u30fc\u30eb\u3068\u7d71\u5408\u3059\u308b\n<ul class=\"wp-block-list\">\n<li>MLflow \u3092 TensorFlow \u3084 PyTorch \u306a\u3069\u306e\u30c7\u30a3\u30fc\u30d7\u30e9\u30fc\u30cb\u30f3\u30b0 \u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3068\u7d71\u5408\u3059\u308b<\/li>\n\n\n\n<li>\u30b8\u30e5\u30d4\u30bf\u30fc \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3067 MLflow \u3092\u4f7f\u7528\u3059\u308b<\/li>\n\n\n\n<li>AWS SageMaker\u3001Azure ML\u306a\u3069\u306e\u30af\u30e9\u30a6\u30c9\u30b5\u30fc\u30d3\u30b9\u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u3068\u7d71\u5408\u3059\u308b<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>\u904b\u7528\u74b0\u5883\u306e\u30c7\u30d7\u30ed\u30a4\u3068\u30d9\u30b9\u30c8 \u30d7\u30e9\u30af\u30c6\u30a3\u30b9\n<ul class=\"wp-block-list\">\n<li>MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u65b9\u6cd5<\/li>\n\n\n\n<li>\u904b\u7528\u74b0\u5883\u3067\u306e MLflow \u306e\u30d9\u30b9\u30c8 \u30d7\u30e9\u30af\u30c6\u30a3\u30b9<\/li>\n\n\n\n<li>\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u6700\u9069\u5316<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>\u30b3\u30fc\u30b9\u306e\u6982\u8981\u3068\u5c55\u671b<\/li>\n<\/ol>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u3053\u306e\u30b3\u30fc\u30b9\u306e\u5fa9\u7fd2\u3068\u307e\u3068\u3081<\/li>\n\n\n\n<li>MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u3068\u30ea\u30bd\u30fc\u30b9<\/li>\n\n\n\n<li>MLflow \u306e\u672a\u6765<\/li>\n<\/ul>\n\n\n\n<p>\u3053\u306e\u6982\u8981\u306b\u57fa\u3065\u3044\u3066\u3001MLflow \u306e\u3055\u307e\u3056\u307e\u306a\u5074\u9762\u3078\u306e\u8a73\u7d30\u306a\u30a2\u30af\u30bb\u30b9\u3092\u63d0\u4f9b\u3059\u308b\u4e00\u9023\u306e\u30d3\u30c7\u30aa \u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u3092\u4f5c\u6210\u3067\u304d\u307e\u3059\u3002 \u304a\u5f79\u306b\u7acb\u3066\u308c\u3070!<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u3053\u306e\u30b3\u30fc\u30b9\u306e\u76ee\u7684\u3068\u76ee\u7684\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u3053\u306e\u30b3\u30fc\u30b9\u306e\u76ee\u7684\u3068\u76ee\u7684\u306f\u6b21\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>MLflow \u306e\u57fa\u672c\u7684\u306a\u6982\u5ff5\u3068\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3092\u7406\u89e3\u3059\u308b: \u3053\u306e\u30b3\u30fc\u30b9\u3067\u306f\u3001\u30c8\u30ec\u30fc\u30b9\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3001\u30e2\u30c7\u30eb\u3001\u30ec\u30b8\u30b9\u30c8\u30ea\u306a\u3069\u3001MLflow \u306e\u57fa\u672c\u7684\u306a\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u3068\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u306b\u3064\u3044\u3066\u5b66\u7fd2\u3057\u307e\u3059\u3002 \u3053\u308c\u306f\u3001\u5b66\u751f\u304c MLflow \u306e\u30b3\u30a2\u6a5f\u80fd\u3068\u4f7f\u7528\u30b7\u30ca\u30ea\u30aa\u3092\u7fd2\u5f97\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002<\/li>\n\n\n\n<li>MLflow \u306e\u5b9f\u8df5\u3092\u30de\u30b9\u30bf\u30fc\u3059\u308b: \u3053\u306e\u30b3\u30fc\u30b9\u3067\u306f\u3001MLflow \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8ffd\u8de1\u3057\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u6574\u7406\u3057\u3001\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3057\u3066\u30c7\u30d7\u30ed\u30a4\u3057\u3001\u30e2\u30c7\u30eb \u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u7ba1\u7406\u3059\u308b\u65b9\u6cd5\u3092\u5b66\u7fd2\u3057\u307e\u3059\u3002 \u5b66\u751f\u306f\u3001\u5b9f\u969b\u306e\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067MLflow\u3092\u67d4\u8edf\u306b\u4f7f\u7528\u3057\u3066\u3001\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u306e\u52b9\u7387\u3092\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u518d\u73fe\u6027\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u5411\u4e0a\u3055\u305b\u308b: MLflow \u3092\u4f7f\u7528\u3059\u308b\u3053\u3068\u3067\u3001\u5b66\u751f\u306f\u5b9f\u9a13\u30d7\u30ed\u30bb\u30b9\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u7d50\u679c\u3092\u3088\u308a\u9069\u5207\u306b\u6587\u66f8\u5316\u3057\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u518d\u73fe\u6027\u3092\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u540c\u6642\u306b\u3001\u30e2\u30c7\u30eb\u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u901a\u3058\u3066\u3001\u30c1\u30fc\u30e0\u30e1\u30f3\u30d0\u30fc\u306f\u30e2\u30c7\u30eb\u3092\u3088\u308a\u7c21\u5358\u306b\u5171\u6709\u304a\u3088\u3073\u5171\u540c\u4f5c\u696d\u3067\u304d\u3001\u5168\u4f53\u7684\u306a\u751f\u7523\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u4ed6\u306e\u30c4\u30fc\u30eb\u3084\u30af\u30e9\u30a6\u30c9\u30b5\u30fc\u30d3\u30b9\u3068\u306e\u7d71\u5408\u65b9\u6cd5\u3092\u5b66\u3076: \u3053\u306e\u30b3\u30fc\u30b9\u3067\u306f\u3001MLflow \u3092 TensorFlow \u3084 PyTorch \u306a\u3069\u306e\u4ed6\u306e\u4e00\u822c\u7684\u306a\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0 \u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3068\u7d71\u5408\u3059\u308b\u65b9\u6cd5\u3068\u3001\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u3067 MLflow \u3092\u67d4\u8edf\u306b\u4f7f\u7528\u3059\u308b\u305f\u3081\u306b Jupyter Notebook \u3084 AWS SageMaker \u3084 Azure ML \u306a\u3069\u306e\u30af\u30e9\u30a6\u30c9 \u30b5\u30fc\u30d3\u30b9\u3068\u7d71\u5408\u3059\u308b\u65b9\u6cd5\u3092\u53d7\u8b1b\u8005\u306b\u8aac\u660e\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>MLflow \u306e\u904b\u7528\u74b0\u5883\u306e\u30c7\u30d7\u30ed\u30a4\u3068\u30d9\u30b9\u30c8 \u30d7\u30e9\u30af\u30c6\u30a3\u30b9\u3092\u30de\u30b9\u30bf\u30fc\u3059\u308b: \u3053\u306e\u30b3\u30fc\u30b9\u3067\u306f\u3001MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u65b9\u6cd5\u3092\u5b66\u751f\u306b\u6559\u3048\u3001\u904b\u7528\u74b0\u5883\u3067 MLflow \u3092\u4f7f\u7528\u3059\u308b\u305f\u3081\u306e\u30d9\u30b9\u30c8 \u30d7\u30e9\u30af\u30c6\u30a3\u30b9\u3001\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3001\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u6700\u9069\u5316\u306e\u65b9\u6cd5\u3092\u7d39\u4ecb\u3057\u3066\u3001\u5b66\u751f\u304c\u5b9f\u969b\u306e\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b MLflow \u3092\u3088\u308a\u9069\u5207\u306b\u9069\u7528\u3067\u304d\u308b\u3088\u3046\u306b\u3057\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u5168\u4f53\u3068\u3057\u3066\u3001\u3053\u306e\u30b3\u30fc\u30b9\u306f\u3001\u5b66\u751f\u304cMLflow\u306e\u6a5f\u80fd\u3068\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u3092\u5b8c\u5168\u306b\u7406\u89e3\u3057\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u958b\u767a\u52b9\u7387\u3068\u54c1\u8cea\u3092\u5411\u4e0a\u3055\u305b\u3001\u30c1\u30fc\u30e0\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4fdd\u5b88\u6027\u3092\u4fc3\u9032\u3059\u308b\u306e\u306b\u5f79\u7acb\u3064\u3088\u3046\u306b\u8a2d\u8a08\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u3092\u9078\u3093\u3060\u7406\u7531\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u3092\u9078\u629e\u3059\u308b\u4e3b\u306a\u7406\u7531\u306f\u6b21\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u3067\u4f7f\u3044\u3084\u3059\u3044: MLflow \u306f\u3001\u6d3b\u767a\u306a\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u3068\u8c4a\u5bcc\u306a\u30ea\u30bd\u30fc\u30b9\u3092\u5099\u3048\u305f\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u306e\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u3067\u3059\u3002 \u6a5f\u68b0\u5b66\u7fd2\u30a8\u30f3\u30b8\u30cb\u30a2\u3084\u30c7\u30fc\u30bf \u30b5\u30a4\u30a8\u30f3\u30c6\u30a3\u30b9\u30c8\u304c\u8907\u96d1\u306a\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u7c21\u5358\u306b\u7ba1\u7406\u3067\u304d\u308b\u3088\u3046\u306b\u8a2d\u8a08\u3055\u308c\u3066\u3044\u307e\u3059\u3002 MLflow \u306e API \u8a2d\u8a08\u306f\u30b7\u30f3\u30d7\u30eb\u3067\u76f4\u611f\u7684\u3067\u4f7f\u3044\u3084\u3059\u3044\u3067\u3059\u3002<\/li>\n\n\n\n<li>\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u5316\u3068\u30e2\u30b8\u30e5\u30fc\u30eb\u6027: MLflow \u306f\u3001\u8ffd\u8de1\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3001\u30e2\u30c7\u30eb\u3001\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306e 4 \u3064\u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u69cb\u6210\u3055\u308c\u3066\u304a\u308a\u3001\u500b\u5225\u306b\u4f7f\u7528\u3059\u308b\u3053\u3068\u3082\u3001\u7d44\u307f\u5408\u308f\u305b\u3066\u4f7f\u7528\u3057\u3066\u3055\u307e\u3056\u307e\u306a\u30cb\u30fc\u30ba\u3092\u6e80\u305f\u3059\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u30e2\u30b8\u30e5\u30e9\u30fc\u8a2d\u8a08\u306b\u3088\u308a\u3001MLflow \u306f\u975e\u5e38\u306b\u67d4\u8edf\u306b\u306a\u308a\u3001\u5b9f\u969b\u306e\u30b7\u30ca\u30ea\u30aa\u306b\u5408\u308f\u305b\u3066\u30ab\u30b9\u30bf\u30de\u30a4\u30ba\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u8907\u6570\u306e\u6a5f\u68b0\u5b66\u7fd2\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306b\u9069\u5fdc: MLflow \u306f\u3001TensorFlow\u3001PyTorch\u3001Scikit-learn \u306a\u3069\u3001\u3055\u307e\u3056\u307e\u306a\u4e00\u822c\u7684\u306a\u6a5f\u68b0\u5b66\u7fd2\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u307e\u3059\u3002 \u3064\u307e\u308a\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u4f7f\u7528\u3059\u308b\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306b\u95a2\u4fc2\u306a\u304f\u3001MLflow \u3092\u4f7f\u7528\u3057\u3066\u7ba1\u7406\u304a\u3088\u3073\u8ffd\u8de1\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30c1\u30fc\u30e0\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4fdd\u5b88\u6027\u3092\u4fc3\u9032\u3059\u308b: MLflow \u306e\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u306f\u30e2\u30c7\u30eb\u306e\u5171\u6709\u3068\u5171\u540c\u4f5c\u696d\u304c\u5bb9\u6613\u306b\u306a\u308a\u307e\u3059\u3002 \u3055\u3089\u306b\u3001MLflow \u3067\u306f\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u6210\u679c\u7269\u306a\u3069\u306e\u5b9f\u9a13\u306e\u8a73\u7d30\u3092\u8a18\u9332\u3067\u304d\u308b\u305f\u3081\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u518d\u73fe\u6027\u3068\u4fdd\u5b88\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30c7\u30d7\u30ed\u30a4\u3068\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u304c\u7c21\u5358: MLflow \u30e2\u30c7\u30eb\u306f\u3055\u307e\u3056\u307e\u306a\u5f62\u5f0f\u3067\u4fdd\u5b58\u3067\u304d\u3001\u30aa\u30f3\u30d7\u30ec\u30df\u30b9 \u30b5\u30fc\u30d0\u30fc\u3001\u30af\u30e9\u30a6\u30c9 \u30b5\u30fc\u30d3\u30b9\u3001\u30b3\u30f3\u30c6\u30ca\u30fc \u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u306a\u3069\u306e\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u3078\u306e\u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u307e\u3059\u3002 \u3055\u3089\u306b\u3001MLflow \u306f\u3001AWS SageMaker\u3001Azure ML \u306a\u3069\u306e\u3055\u307e\u3056\u307e\u306a\u30af\u30e9\u30a6\u30c9\u30b5\u30fc\u30d3\u30b9\u3068\u306e\u7d71\u5408\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u304a\u308a\u3001\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u9593\u3067\u7c21\u5358\u306b\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3\u3092\u5b9f\u73fe\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30a2\u30af\u30c6\u30a3\u30d6\u306a\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u3068\u5e38\u306b\u66f4\u65b0\u3055\u308c\u308b\u6a5f\u80fd: MLflow \u306b\u306f\u3001\u65b0\u6a5f\u80fd\u3068\u6700\u9069\u5316\u3092\u5b9a\u671f\u7684\u306b\u30ea\u30ea\u30fc\u30b9\u3059\u308b\u30a2\u30af\u30c6\u30a3\u30d6\u306a\u958b\u767a\u8005\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u304c\u3042\u308a\u307e\u3059\u3002 \u3053\u308c\u306f\u3001MLflow \u304c\u6a5f\u68b0\u5b66\u7fd2\u306e\u5206\u91ce\u306b\u304a\u3051\u308b\u6700\u65b0\u306e\u30c6\u30af\u30ce\u30ed\u30b8\u3068\u30cb\u30fc\u30ba\u306b\u5e38\u306b\u9069\u5fdc\u3059\u308b\u3053\u3068\u3092\u610f\u5473\u3057\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow\u306f\u3001\u305d\u306e\u4f7f\u3044\u3084\u3059\u3055\u3001\u67d4\u8edf\u6027\u3001\u4e92\u63db\u6027\u3001\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3001\u62e1\u5f35\u6027\u306b\u3088\u308a\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u512a\u308c\u305f\u30c4\u30fc\u30eb\u3067\u3059\u3002 MLflow \u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u958b\u767a\u52b9\u7387\u3001\u54c1\u8cea\u3001\u4fdd\u5b88\u6027\u3092\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u306e\u5229\u70b9\u3068\u6a5f\u80fd\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u306e\u5229\u70b9\u3068\u6a5f\u80fd\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u4f7f\u3044\u3084\u3059\u3055: MLflow \u306e API \u8a2d\u8a08\u306f\u30b7\u30f3\u30d7\u30eb\u3067\u76f4\u611f\u7684\u3067\u3042\u308a\u3001Python\u3001R\u3001Java \u306e\u3044\u305a\u308c\u3067\u3042\u3063\u3066\u3082\u3001MLflow \u3068\u7c21\u5358\u306b\u7d71\u5408\u3067\u304d\u307e\u3059\u3002 \u540c\u6642\u306b\u3001MLflow \u306f\u3001\u5b9f\u9a13\u7d50\u679c\u3068\u30e2\u30c7\u30eb\u60c5\u5831\u3092\u8996\u899a\u7684\u306b\u8868\u793a\u3059\u308b\u305f\u3081\u306e\u30e6\u30fc\u30b6\u30fc \u30d5\u30ec\u30f3\u30c9\u30ea\u306a\u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30e2\u30b8\u30e5\u30e9\u30fc\u8a2d\u8a08: MLflow \u306f\u3001\u8ffd\u8de1\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3001\u30e2\u30c7\u30eb\u3001\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306e 4 \u3064\u306e\u4e3b\u8981\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u69cb\u6210\u3055\u308c\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u306f\u3001\u5358\u72ec\u3067\u307e\u305f\u306f\u7d44\u307f\u5408\u308f\u305b\u3066\u4f7f\u7528\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u3001\u5b9f\u969b\u306e\u8981\u4ef6\u306b\u5fdc\u3058\u3066\u67d4\u8edf\u306b\u69cb\u6210\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30de\u30eb\u30c1\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306e\u30b5\u30dd\u30fc\u30c8: MLflow \u306f\u3001TensorFlow\u3001PyTorch\u3001Scikit-learn \u306a\u3069\u306e\u3055\u307e\u3056\u307e\u306a\u4e00\u822c\u7684\u306a\u6a5f\u68b0\u5b66\u7fd2\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u304a\u308a\u3001\u3055\u307e\u3056\u307e\u306a\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3067\u30e2\u30c7\u30eb\u3092\u7c21\u5358\u306b\u7ba1\u7406\u304a\u3088\u3073\u8ffd\u8de1\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u7d71\u5408\u30e2\u30c7\u30eb\u5f62\u5f0f: MLflow \u306f\u3001\u3055\u307e\u3056\u307e\u306a\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306e\u30e2\u30c7\u30eb\u3092\u6a19\u6e96\u5316\u3055\u308c\u305f\u5f62\u5f0f\u3067\u4fdd\u5b58\u3057\u3066\u3001\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u3067\u30c7\u30d7\u30ed\u30a4\u304a\u3088\u3073\u4f7f\u7528\u3067\u304d\u308b\u7d71\u5408\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u304c\u53ef\u80fd\u306b\u306a\u308a\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u9593\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u304c\u5bb9\u6613\u306b\u306a\u308a\u307e\u3059\u3002 \u540c\u6642\u306b\u3001\u627f\u8a8d\u3001\u30a2\u30c3\u30d7\u30e9\u30a4\u30f3\u3001\u30c0\u30a6\u30f3\u30e9\u30a4\u30f3\u306a\u3069\u306e\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u3088\u308a\u9069\u5207\u306b\u7ba1\u7406\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30c7\u30d7\u30ed\u30a4\u304c\u7c21\u5358: MLflow \u3067\u306f\u3001\u30ed\u30fc\u30ab\u30eb \u30b5\u30fc\u30d0\u30fc\u3001\u30af\u30e9\u30a6\u30c9 \u30b5\u30fc\u30d3\u30b9\u3001\u30b3\u30f3\u30c6\u30ca\u30fc \u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u306a\u3069\u306e\u8907\u6570\u306e\u74b0\u5883\u3078\u306e\u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u308b\u305f\u3081\u3001\u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u7c21\u5358\u306b\u9069\u7528\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3: MLflow \u306f\u3001AWS SageMaker\u3001Azure ML \u306a\u3069\u306e\u3055\u307e\u3056\u307e\u306a\u30af\u30e9\u30a6\u30c9\u30b5\u30fc\u30d3\u30b9\u3068\u7d71\u5408\u3057\u3066\u3001\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u9593\u3067\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3067\u304d\u307e\u3059\u3002 \u540c\u6642\u306b\u3001MLflow\u306e\u8a2d\u8a08\u306b\u3088\u308a\u3001\u30e6\u30fc\u30b6\u30fc\u306f\u7279\u5b9a\u306e\u30cb\u30fc\u30ba\u306b\u5408\u308f\u305b\u3066\u30d7\u30e9\u30b0\u30a4\u30f3\u3092\u30ab\u30b9\u30bf\u30de\u30a4\u30ba\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u3067\u6d3b\u767a\u306a\u30b3\u30df\u30e5\u30cb\u30c6\u30a3: MLflow \u306f\u3001\u6d3b\u767a\u306a\u958b\u767a\u8005\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u3092\u6301\u3064\u30aa\u30fc\u30d7\u30f3 \u30bd\u30fc\u30b9 \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u3059\u3002 \u3053\u308c\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u306e\u5206\u91ce\u306b\u304a\u3051\u308b\u6700\u65b0\u306e\u30c6\u30af\u30ce\u30ed\u30b8\u3068\u30cb\u30fc\u30ba\u306b\u9069\u5fdc\u3059\u308b\u305f\u3081\u306b\u3001MLflow \u304c\u7d76\u3048\u305a\u66f4\u65b0\u304a\u3088\u3073\u6539\u5584\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u610f\u5473\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u518d\u73fe\u6027\u306e\u5411\u4e0a: MLflow \u3067\u306f\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u6210\u679c\u7269\u306a\u3069\u306e\u5b9f\u9a13\u306e\u8a73\u7d30\u3092\u8a18\u9332\u3067\u304d\u308b\u305f\u3081\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u518d\u73fe\u6027\u3068\u4fdd\u5b88\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u306e\u9577\u6240\u3068\u6a5f\u80fd\u306b\u306f\u3001\u4f7f\u3044\u3084\u3059\u3055\u3001\u67d4\u8edf\u6027\u3001\u4e92\u63db\u6027\u3001\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3001\u30c7\u30d7\u30ed\u30a4\u306e\u5bb9\u6613\u3055\u3001\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3\u3001\u518d\u73fe\u6027\u304c\u542b\u307e\u308c\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u7406\u60f3\u7684\u306a\u30c4\u30fc\u30eb\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow \u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u3092\u8a73\u3057\u304f\u8aac\u660e\u3059\u308b<\/p>\n\n\n\n<p>MLflow \u306f\u3001\u8ffd\u8de1\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3001\u30e2\u30c7\u30eb\u3001\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306e 4 \u3064\u306e\u4e3b\u8981\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u69cb\u6210\u3055\u308c\u308b\u30e2\u30b8\u30e5\u30fc\u30eb\u5f0f\u306e\u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u7ba1\u7406\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u3067\u3059\u3002 4 \u3064\u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3092\u8a73\u3057\u304f\u898b\u3066\u307f\u307e\u3057\u3087\u3046\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>ML\u30d5\u30ed\u30fc\u8ffd\u8de1:<\/li>\n<\/ol>\n\n\n\n<p>MLflow \u8ffd\u8de1\u306f\u3001\u5b9f\u9a13\u3092\u8a18\u9332\u304a\u3088\u3073\u30af\u30a8\u30ea\u3059\u308b\u305f\u3081\u306e API \u304a\u3088\u3073 UI \u3067\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u5b9f\u9a13\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8(\u30e2\u30c7\u30eb\u30d5\u30a1\u30a4\u30eb\u3001\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306a\u3069)\u3092\u8a18\u9332\u3057\u3066\u3001\u5f8c\u306e\u6bb5\u968e\u3067\u5206\u6790\u3068\u6bd4\u8f03\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 MLflow Tracking \u306f\u3001\u30ed\u30fc\u30ab\u30eb \u30d5\u30a1\u30a4\u30eb \u30b7\u30b9\u30c6\u30e0\u3001\u30c7\u30fc\u30bf\u30d9\u30fc\u30b9\u3001\u30ea\u30e2\u30fc\u30c8 \u30b5\u30fc\u30d0\u30fc\u306a\u3069\u306e\u8907\u6570\u306e\u30b9\u30c8\u30ec\u30fc\u30b8\u65b9\u6cd5\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u308b\u305f\u3081\u3001\u30c1\u30fc\u30e0\u9593\u304a\u3088\u3073\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u9593\u306e\u5b9f\u9a13\u7684\u306a\u30c7\u30fc\u30bf\u5171\u6709\u304c\u5bb9\u6613\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"2\">\n<li>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8:<\/li>\n<\/ol>\n\n\n\n<p>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u6574\u7406\u3057\u3066\u5b9f\u884c\u3059\u308b\u305f\u3081\u306e\u6a19\u6e96\u5f62\u5f0f\u3067\u3059\u3002 MLproject \u3068\u3044\u3046\u540d\u524d\u306e YAML \u30d5\u30a1\u30a4\u30eb\u3092\u5b9a\u7fa9\u3059\u308b\u3053\u3068\u3067\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u74b0\u5883\u3001\u4f9d\u5b58\u95a2\u4fc2\u3001\u304a\u3088\u3073\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u3092\u6307\u5b9a\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u6a19\u6e96\u5316\u3055\u308c\u305f\u5f62\u5f0f\u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u518d\u73fe\u6027\u3068\u79fb\u690d\u6027\u3092\u5411\u4e0a\u3055\u305b\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u4ed6\u306e\u958b\u767a\u8005\u306f\u3001\u57fa\u306b\u306a\u308b\u5b9f\u88c5\u306e\u8a73\u7d30\u3092\u6c17\u306b\u3059\u308b\u3053\u3068\u306a\u304f\u3001MLflow \u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3 \u30c4\u30fc\u30eb\u307e\u305f\u306f API \u3092\u4f7f\u7528\u3057\u3066\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u7c21\u5358\u306b\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"3\">\n<li>MLflow \u30e2\u30c7\u30eb:<\/li>\n<\/ol>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3001\u30c7\u30d7\u30ed\u30a4\u3001\u5171\u6709\u3059\u308b\u305f\u3081\u306e\u6a19\u6e96\u5f62\u5f0f\u3067\u3059\u3002 MLflow \u30e2\u30c7\u30eb\u306f\u3001TensorFlow\u3001PyTorch\u3001Scikit-learn \u306a\u3069\u306e\u4e00\u822c\u7684\u306a\u6a5f\u68b0\u5b66\u7fd2\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u3092 MLflow Model \u5f62\u5f0f\u3067\u4fdd\u5b58\u3059\u308b\u3053\u3068\u3067\u3001\u30ed\u30fc\u30ab\u30eb \u30b5\u30fc\u30d0\u30fc\u3001\u30af\u30e9\u30a6\u30c9 \u30b5\u30fc\u30d3\u30b9\u3001\u30b3\u30f3\u30c6\u30ca\u30fc \u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u306a\u3069\u306e\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u306b\u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3057\u3066\u3001\u30af\u30ed\u30b9\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u306e\u30e2\u30c7\u30eb\u306e\u5171\u6709\u3068\u30c7\u30d7\u30ed\u30a4\u3092\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"4\">\n<li>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea:<\/li>\n<\/ol>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u4e00\u5143\u5316\u3055\u308c\u305f\u30ea\u30dd\u30b8\u30c8\u30ea\u3067\u3059\u3002 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3001\u72b6\u614b\u3001\u30b3\u30e1\u30f3\u30c8\u306a\u3069\u306e\u30e1\u30bf\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3067\u304d\u307e\u3059\u3002 \u3055\u3089\u306b\u3001MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb\u306e\u627f\u8a8d\u3001\u30a2\u30c3\u30d7\u30e9\u30a4\u30f3\u3001\u30c0\u30a6\u30f3\u30e9\u30a4\u30f3\u306a\u3069\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u7ba1\u7406\u64cd\u4f5c\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u9593\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30e2\u30c7\u30eb\u306e\u5171\u6709\u3092\u5bb9\u6613\u306b\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u5168\u4f53\u3068\u3057\u3066\u3001MLflow \u306e\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u306f\u4e3b\u306b\u4e0a\u8a18\u306e 4 \u3064\u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u69cb\u6210\u3055\u308c\u3066\u304a\u308a\u3001\u3055\u307e\u3056\u307e\u306a\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30cb\u30fc\u30ba\u3092\u6e80\u305f\u3059\u305f\u3081\u306b\u3001\u5358\u72ec\u3067\u4f7f\u7528\u3059\u308b\u3053\u3068\u3082\u7d44\u307f\u5408\u308f\u305b\u3066\u4f7f\u7528\u3059\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002 MLflow \u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u5b9f\u9a13\u30c7\u30fc\u30bf\u306e\u7ba1\u7406\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u6574\u7406\u3001\u30e2\u30c7\u30eb\u306e\u4fdd\u5b58\u3068\u30c7\u30d7\u30ed\u30a4\u3001\u5171\u540c\u4f5c\u696d\u3092\u3088\u308a\u9069\u5207\u306b\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u30c8\u30e9\u30c3\u30ad\u30f3\u30b0\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u8ffd\u8de1\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u306e\u5b9f\u9a13\u4e2d\u306b\u751f\u6210\u3055\u308c\u305f\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3001\u30af\u30a8\u30ea\u3001\u8996\u899a\u5316\u3059\u308b MLflow \u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u3059\u3002 \u4e3b\u306b\u3001\u5b9f\u9a13\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30e9\u30d9\u30eb\u3001\u6210\u679c\u7269 (\u30e2\u30c7\u30eb \u30d5\u30a1\u30a4\u30eb\u3001\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306a\u3069) \u306e\u8ffd\u8de1\u306b\u5f79\u7acb\u3064 API \u3068\u30d3\u30b8\u30e5\u30a2\u30eb \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u304c\u542b\u307e\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p>MLflow \u8ffd\u8de1\u306e\u4e3b\u306a\u6a5f\u80fd\u306e\u4e00\u90e8\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u5b9f\u9a13\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3068\u6307\u6a19\u3092\u8ffd\u8de1\u3059\u308b: \u5b9f\u9a13\u4e2d\u306b\u3001\u3055\u307e\u3056\u307e\u306a\u30cf\u30a4\u30d1\u30fc\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30c7\u30fc\u30bf\u306e\u524d\u51e6\u7406\u65b9\u6cd5\u3001\u30e2\u30c7\u30eb\u69cb\u9020\u3092\u8a66\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 MLflow Tracking API \u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u3053\u308c\u3089\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3068\u5b9f\u9a13\u7d50\u679c (\u7cbe\u5ea6\u3001\u640d\u5931\u5024\u306a\u3069) \u3092\u7c21\u5358\u306b\u8a18\u9332\u3057\u3066\u3001\u5f8c\u306e\u6bb5\u968e\u3067\u6bd4\u8f03\u304a\u3088\u3073\u5206\u6790\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u6210\u679c\u7269\u3092\u8a18\u9332\u3059\u308b: \u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3068\u30e1\u30c8\u30ea\u30c3\u30af\u306b\u52a0\u3048\u3066\u3001MLflow \u8ffd\u8de1 API \u3092\u4f7f\u7528\u3057\u3066\u3001\u5b9f\u9a13\u4e2d\u306b\u751f\u6210\u3055\u308c\u305f\u6210\u679c\u7269 (\u30e2\u30c7\u30eb \u30d5\u30a1\u30a4\u30eb\u3001\u524d\u51e6\u7406\u30b9\u30af\u30ea\u30d7\u30c8\u3001\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306a\u3069) \u3092\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u5b9f\u9a13\u306e\u518d\u73fe\u6027\u304c\u5411\u4e0a\u3057\u3001\u3053\u308c\u3089\u306e\u30ea\u30bd\u30fc\u30b9\u3092\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u3067\u518d\u5229\u7528\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u5b9f\u9a13\u7d50\u679c\u3092\u8996\u899a\u5316\u3059\u308b: MLflow Tracking \u306f\u3001\u5b9f\u9a13\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u6210\u679c\u7269\u3092\u8996\u899a\u7684\u306b\u8868\u793a\u3059\u308b\u305f\u3081\u306e\u30e6\u30fc\u30b6\u30fc \u30d5\u30ec\u30f3\u30c9\u30ea\u306a\u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 \u3053\u306e\u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u4f7f\u7528\u3057\u3066\u3001\u5b9f\u9a13\u7d50\u679c\u306e\u4e26\u3079\u66ff\u3048\u3001\u30d5\u30a3\u30eb\u30bf\u30fc\u51e6\u7406\u3001\u6bd4\u8f03\u3092\u884c\u3044\u3001\u5b9f\u9a13\u7d50\u679c\u3092\u3088\u308a\u9069\u5207\u306b\u5206\u6790\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u8907\u6570\u306e\u30b9\u30c8\u30ec\u30fc\u30b8 \u30d0\u30c3\u30af\u30a8\u30f3\u30c9\u306e\u30b5\u30dd\u30fc\u30c8: MLflow Tracking \u306f\u3001\u30ed\u30fc\u30ab\u30eb \u30d5\u30a1\u30a4\u30eb \u30b7\u30b9\u30c6\u30e0\u3001\u30c7\u30fc\u30bf\u30d9\u30fc\u30b9\u3001\u30ea\u30e2\u30fc\u30c8 \u30b5\u30fc\u30d0\u30fc\u306a\u3069\u306e\u8907\u6570\u306e\u30b9\u30c8\u30ec\u30fc\u30b8 \u30d0\u30c3\u30af\u30a8\u30f3\u30c9\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u307e\u3059\u3002 \u3053\u308c\u306f\u3001\u5b9f\u969b\u306e\u30cb\u30fc\u30ba\u306b\u5fdc\u3058\u3066\u9069\u5207\u306a\u4fdd\u5b58\u65b9\u6cd5\u3092\u9078\u629e\u3057\u3001\u30c1\u30fc\u30e0\u9593\u304a\u3088\u3073\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u9593\u306e\u5b9f\u9a13\u30c7\u30fc\u30bf\u5171\u6709\u3092\u5b9f\u73fe\u3067\u304d\u308b\u3053\u3068\u3092\u610f\u5473\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u4ed6\u306e MLflow \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u306e\u7d71\u5408: MLflow \u8ffd\u8de1\u306f\u3001\u30a8\u30f3\u30c9\u30c4\u30fc\u30a8\u30f3\u30c9\u306e\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u306e\u305f\u3081\u306b\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u304a\u3088\u3073 MLflow \u30e2\u30c7\u30eb \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u30b7\u30fc\u30e0\u30ec\u30b9\u306b\u7d71\u5408\u3055\u308c\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u81ea\u52d5\u7684\u306b\u8a18\u9332\u3057\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6e08\u307f\u306e\u30e2\u30c7\u30eb\u3092 MLflow \u30e2\u30c7\u30eb\u5f62\u5f0f\u3067\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow Tracking \u306f\u5f37\u529b\u3067\u67d4\u8edf\u306a\u5b9f\u9a13\u7ba1\u7406\u30c4\u30fc\u30eb\u3067\u3042\u308a\u3001\u6a5f\u68b0\u5b66\u7fd2\u5b9f\u9a13\u306e\u30d7\u30ed\u30bb\u30b9\u3092\u3088\u308a\u9069\u5207\u306b\u8a18\u9332\u304a\u3088\u3073\u5206\u6790\u3057\u3001\u5b9f\u9a13\u306e\u518d\u73fe\u6027\u3068\u52b9\u7387\u3092\u5411\u4e0a\u3055\u305b\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u6574\u7406\u304a\u3088\u3073\u8a18\u8ff0\u3059\u308b\u305f\u3081\u306e\u6a19\u6e96\u5316\u3055\u308c\u305f\u65b9\u6cd5\u3092\u63d0\u4f9b\u3059\u308b MLflow \u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u3059\u3002 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u6a5f\u68b0\u5b66\u7fd2\u30b3\u30fc\u30c9\u3092\u3088\u308a\u7c21\u5358\u306b\u5171\u6709\u3001\u518d\u5229\u7528\u3001\u30ec\u30d7\u30ea\u30b1\u30fc\u30c8\u3067\u304d\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u79fb\u690d\u6027\u3068\u4fdd\u5b88\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4e3b\u306a\u6a5f\u80fd\u306f\u6b21\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u8a18\u8ff0\u30d5\u30a1\u30a4\u30eb: MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306f\u3001MLproject \u3068\u3044\u3046 YAML \u30d5\u30a1\u30a4\u30eb\u3092\u4f7f\u7528\u3057\u3066\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u540d\u3001\u30d0\u30fc\u30b8\u30e7\u30f3\u3001\u8aac\u660e\u306a\u3069\u306e\u9805\u76ee\u306e\u30e1\u30bf\u30c7\u30fc\u30bf\u3092\u8a18\u8ff0\u3057\u307e\u3059\u3002 \u3055\u3089\u306b\u3001MLproject \u30d5\u30a1\u30a4\u30eb\u3067\u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2 (Python \u30e9\u30a4\u30d6\u30e9\u30ea\u3001\u30b7\u30b9\u30c6\u30e0 \u30e9\u30a4\u30d6\u30e9\u30ea\u306a\u3069) \u3068\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8 (\u3064\u307e\u308a\u3001\u5b9f\u884c\u3067\u304d\u308b\u30b9\u30af\u30ea\u30d7\u30c8\u307e\u305f\u306f\u30b3\u30de\u30f3\u30c9) \u3092\u6307\u5b9a\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u74b0\u5883\u69cb\u6210: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u3067\u6b63\u3057\u304f\u5b9f\u884c\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u305f\u3081\u306b\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306f\u3001Conda\u3001Docker\u3001\u4eee\u60f3\u74b0\u5883\u306a\u3069\u3001\u8907\u6570\u306e\u74b0\u5883\u69cb\u6210\u65b9\u6cd5\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 MLproject \u30d5\u30a1\u30a4\u30eb\u3067\u76ee\u7684\u306e\u74b0\u5883\u69cb\u6210\u3092\u6307\u5b9a\u3059\u308b\u3068\u3001MLflow \u306b\u3088\u3063\u3066\u4f9d\u5b58\u95a2\u4fc2\u3068\u8a2d\u5b9a\u304c\u81ea\u52d5\u7684\u306b\u51e6\u7406\u3055\u308c\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u5316: MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3092\u5b9a\u7fa9\u3057\u3066\u3001\u4ed6\u306e\u30e6\u30fc\u30b6\u30fc\u304c\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u5b9f\u884c\u6642\u306b\u5fc5\u8981\u306b\u5fdc\u3058\u3066\u7570\u306a\u308b\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u5024\u3092\u6e21\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u30d1\u30e9\u30e1\u30c8\u30ea\u30c3\u30af\u8a2d\u8a08\u306b\u3088\u308a\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u67d4\u8edf\u6027\u304c\u5411\u4e0a\u3057\u3001\u3055\u307e\u3056\u307e\u306a\u5b9f\u9a13\u30b7\u30ca\u30ea\u30aa\u306b\u9069\u5fdc\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u7c21\u7565\u5316\u3055\u308c\u305f\u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3 \u30c4\u30fc\u30eb: MLflow \u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3 \u30c4\u30fc\u30eb\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u57fa\u306b\u306a\u308b\u5b9f\u88c5\u306e\u8a73\u7d30\u3092\u6c17\u306b\u305b\u305a\u306b MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u7c21\u5358\u306b\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002 \u3055\u3089\u306b\u3001MLflow \u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3 \u30c4\u30fc\u30eb\u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u60c5\u5831\u306e\u8868\u793a\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u306e\u4e00\u89a7\u8868\u793a\u306a\u3069\u3001\u305d\u306e\u4ed6\u306e\u4fbf\u5229\u306a\u6a5f\u80fd\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u307e\u3059\u3002<code>mlflow run &lt;project_uri&gt; -P &lt;parameter_name&gt;=&lt;parameter_value&gt;<\/code><\/li>\n\n\n\n<li>\u4ed6\u306e MLflow \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u306e\u7d71\u5408: MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306f\u3001\u30a8\u30f3\u30c9\u30c4\u30fc\u30a8\u30f3\u30c9\u306e\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u306e\u305f\u3081\u306b\u3001\u8ffd\u8de1\u3084\u30e2\u30c7\u30eb\u306a\u3069\u306e\u4ed6\u306e MLflow \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u30b7\u30fc\u30e0\u30ec\u30b9\u306b\u7d71\u5408\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306f\u3001MLflow \u8ffd\u8de1 API \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3057\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6e08\u307f\u306e\u30e2\u30c7\u30eb\u3092 MLflow \u30e2\u30c7\u30eb\u5f62\u5f0f\u3067\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>Git \u7d71\u5408: MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306f\u3001GitHub\u3001GitLab \u306a\u3069\u306e Git \u30ea\u30dd\u30b8\u30c8\u30ea\u304b\u3089\u76f4\u63a5\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002 \u3064\u307e\u308a\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u30c1\u30fc\u30e0\u30e1\u30f3\u30d0\u30fc\u3068\u7c21\u5358\u306b\u5171\u6709\u3057\u3001\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u3092\u7dad\u6301\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u7528\u306b\u6a19\u6e96\u5316\u3055\u308c\u3001\u79fb\u690d\u53ef\u80fd\u3067\u3001\u7c21\u5358\u306b\u5171\u6709\u3067\u304d\u308b\u5f62\u5f0f\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u6574\u7406\u3068\u7ba1\u7406\u304c\u5411\u4e0a\u3057\u3001\u518d\u73fe\u6027\u3068\u4fdd\u5b88\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u30e2\u30c7\u30eb\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u30e2\u30c7\u30eb\u3092\u7ba1\u7406\u3001\u4fdd\u5b58\u3001\u30c7\u30d7\u30ed\u30a4\u3059\u308b MLflow \u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u3059\u3002 \u3053\u308c\u306f\u3001\u3055\u307e\u3056\u307e\u306a\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af(TensorFlow\u3001PyTorch\u3001Scikit-learn\u306a\u3069)\u306e\u30e2\u30c7\u30eb\u3092\u6a19\u6e96\u5316\u3055\u308c\u305f\u5f62\u5f0f\u306b\u4fdd\u5b58\u3057\u3066\u3001\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u3067\u30c7\u30d7\u30ed\u30a4\u3057\u3066\u4f7f\u7528\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u7d71\u4e00\u3055\u308c\u305f\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u306e\u4e3b\u306a\u6a5f\u80fd\u306f\u6b21\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u7d71\u5408\u30e2\u30c7\u30eb\u5f62\u5f0f: MLflow \u30e2\u30c7\u30eb\u3067\u306f\u3001MLmodel \u3068\u3044\u3046 YAML \u30d5\u30a1\u30a4\u30eb\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb\u540d\u3001\u30d0\u30fc\u30b8\u30e7\u30f3\u3001\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306a\u3069\u306e\u30e2\u30c7\u30eb\u306e\u30e1\u30bf\u30c7\u30fc\u30bf\u3092\u8a18\u8ff0\u3057\u307e\u3059\u3002 \u3055\u3089\u306b\u3001MLmodel \u30d5\u30a1\u30a4\u30eb\u306b\u306f\u3001\u30e2\u30c7\u30eb \u30d5\u30a1\u30a4\u30eb\u3001\u524d\u51e6\u7406\u30b9\u30af\u30ea\u30d7\u30c8\u3001\u4f9d\u5b58\u95a2\u4fc2\u306a\u3069\u306e\u60c5\u5831\u304c\u542b\u307e\u308c\u3066\u3044\u307e\u3059\u3002 \u3053\u306e\u7d71\u4e00\u3055\u308c\u305f\u5f62\u5f0f\u306b\u3088\u308a\u3001\u3055\u307e\u3056\u307e\u306a\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3084\u74b0\u5883\u9593\u3067\u30e2\u30c7\u30eb\u3092\u7c21\u5358\u306b\u5171\u6709\u304a\u3088\u3073\u30c7\u30d7\u30ed\u30a4\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30de\u30eb\u30c1\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306e\u30b5\u30dd\u30fc\u30c8: MLflow \u30e2\u30c7\u30eb\u306f\u3001TensorFlow\u3001PyTorch\u3001Scikit-learn \u306a\u3069\u306e\u4e00\u822c\u7684\u306a\u6a5f\u68b0\u5b66\u7fd2\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u307e\u3059\u3002 MLflow \u306b\u3088\u3063\u3066\u63d0\u4f9b\u3055\u308c\u308b Python API \u307e\u305f\u306f\u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3 \u30c4\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u3001MLflow \u30e2\u30c7\u30eb\u5f62\u5f0f\u3067\u3055\u307e\u3056\u307e\u306a\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3067\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u67d4\u8edf\u306a\u30c7\u30d7\u30ed\u30a4 \u30aa\u30d7\u30b7\u30e7\u30f3: MLflow \u30e2\u30c7\u30eb\u306f\u3001\u30aa\u30f3\u30d7\u30ec\u30df\u30b9 \u30b5\u30fc\u30d0\u30fc\u3001\u30af\u30e9\u30a6\u30c9 \u30b5\u30fc\u30d3\u30b9 (AWS SageMaker\u3001Azure ML \u306a\u3069)\u3001\u30b3\u30f3\u30c6\u30ca\u30fc \u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0 (Docker\u3001Kubernetes \u306a\u3069) \u306a\u3069\u3001\u8907\u6570\u306e\u74b0\u5883\u3078\u306e\u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u307e\u3059\u3002 \u3064\u307e\u308a\u3001\u5b9f\u969b\u306e\u30cb\u30fc\u30ba\u306b\u5fdc\u3058\u3066\u9069\u5207\u306a\u30c7\u30d7\u30ed\u30a4\u65b9\u6cd5\u3092\u9078\u629e\u3057\u3001\u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u9069\u7528\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u4ed6\u306e MLflow \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u306e\u7d71\u5408: MLflow \u30e2\u30c7\u30eb\u306f\u3001\u30a8\u30f3\u30c9\u30c4\u30fc\u30a8\u30f3\u30c9\u306e\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u306e\u305f\u3081\u306b\u3001\u8ffd\u8de1\u3084\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306a\u3069\u306e\u4ed6\u306e MLflow \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u30b7\u30fc\u30e0\u30ec\u30b9\u306b\u7d71\u5408\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306f\u3001MLflow \u8ffd\u8de1 API \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3057\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6e08\u307f\u306e\u30e2\u30c7\u30eb\u3092 MLflow \u30e2\u30c7\u30eb\u5f62\u5f0f\u3067\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30e2\u30c7\u30eb\u4e88\u6e2c\u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9: MLflow \u30e2\u30c7\u30eb\u306f\u3001\u4fdd\u5b58\u3055\u308c\u305f\u30e2\u30c7\u30eb\u3092\u7c21\u5358\u306b\u8aad\u307f\u8fbc\u3093\u3067\u4f7f\u7528\u3059\u308b\u305f\u3081\u306e\u6a19\u6e96\u5316\u3055\u308c\u305f\u30e2\u30c7\u30eb\u4e88\u6e2c\u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 \u3053\u306e\u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30aa\u30f3\u30e9\u30a4\u30f3\u63a8\u8ad6\u3001\u30d0\u30c3\u30c1\u51e6\u7406\u306a\u3069\u306e\u3055\u307e\u3056\u307e\u306a\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u30b7\u30ca\u30ea\u30aa\u306b\u30e2\u30c7\u30eb\u3092\u7c21\u5358\u306b\u9069\u7528\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u30e2\u30c7\u30eb\u306f\u3001\u7d71\u5408\u3055\u308c\u3001\u79fb\u690d\u6027\u304c\u3042\u308a\u3001\u30c7\u30d7\u30ed\u30a4\u304c\u5bb9\u6613\u306a\u6a5f\u68b0\u5b66\u7fd2\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 MLflow \u30e2\u30c7\u30eb\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u7ba1\u7406\u3068\u5171\u6709\u304c\u5411\u4e0a\u3057\u3001\u30e2\u30c7\u30eb\u306e\u4fdd\u5b88\u6027\u3068\u751f\u7523\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u3042\u308b MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u5168\u4f53\u3092\u7ba1\u7406\u3059\u308b\u4e00\u5143\u5316\u3055\u308c\u305f\u30e2\u30c7\u30eb \u30ea\u30dd\u30b8\u30c8\u30ea\u3067\u3059\u3002 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3001\u72b6\u614b\u3001\u30b3\u30e1\u30f3\u30c8\u306a\u3069\u306e\u30e1\u30bf\u30c7\u30fc\u30bf\u3092\u3088\u308a\u9069\u5207\u306b\u6574\u7406\u304a\u3088\u3073\u8ffd\u8de1\u3067\u304d\u308b\u305f\u3081\u3001\u30c1\u30fc\u30e0\u9593\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30e2\u30c7\u30eb\u306e\u5171\u6709\u304c\u53ef\u80fd\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306e\u4e3b\u306a\u6a5f\u80fd\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3067\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u30e2\u30c7\u30eb \u30e9\u30a4\u30d6\u30e9\u30ea\u306b\u65b0\u3057\u3044\u30e2\u30c7\u30eb\u3092\u8ffd\u52a0\u3057\u305f\u308a\u3001\u65e2\u5b58\u306e\u30e2\u30c7\u30eb\u3092\u66f4\u65b0\u3057\u305f\u308a\u3059\u308b\u305f\u3073\u306b\u3001\u65b0\u3057\u3044\u30d0\u30fc\u30b8\u30e7\u30f3\u304c\u81ea\u52d5\u7684\u306b\u751f\u6210\u3055\u308c\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb\u306e\u5c65\u6b74\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u7c21\u5358\u306b\u3055\u304b\u306e\u307c\u3063\u3066\u6bd4\u8f03\u3057\u3001\u6700\u9069\u306a\u30e2\u30c7\u30eb\u69cb\u6210\u3092\u898b\u3064\u3051\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30e2\u30c7\u30eb\u306e\u72b6\u614b\u306e\u8ffd\u8de1: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001&#8221;\u30b9\u30c6\u30fc\u30b8\u30f3\u30b0&#8221; (\u30c6\u30b9\u30c8\u4e2d)\u3001&#8221;\u904b\u7528\u4e2d&#8221; (\u904b\u7528\u4e2d)\u3001&#8221;\u30a2\u30fc\u30ab\u30a4\u30d6\u6e08\u307f&#8221; (\u30a2\u30fc\u30ab\u30a4\u30d6\u6e08\u307f) \u306a\u3069\u3001\u30e2\u30c7\u30eb\u306e\u3055\u307e\u3056\u307e\u306a\u72b6\u614b\u3092\u5b9a\u7fa9\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u72b6\u614b\u8ffd\u8de1\u30e1\u30ab\u30cb\u30ba\u30e0\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u3088\u308a\u9069\u5207\u306b\u7ba1\u7406\u3057\u3001\u904b\u7528\u74b0\u5883\u3067\u4f7f\u7528\u3055\u308c\u308b\u30e2\u30c7\u30eb\u304c\u5e38\u306b\u6700\u65b0\u304b\u3064\u6700\u3082\u52b9\u7387\u7684\u306a\u30d0\u30fc\u30b8\u30e7\u30f3\u3067\u3042\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u6ce8\u91c8\u3068\u30e9\u30d9\u30eb: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3067\u306f\u3001\u30e2\u30c7\u30eb\u306e\u6574\u7406\u3068\u8a18\u8ff0\u3092\u6539\u5584\u3059\u308b\u305f\u3081\u306e\u30e2\u30c7\u30eb\u306e\u6ce8\u91c8\u3068\u30bf\u30b0\u4ed8\u3051\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u30e2\u30c7\u30eb\u306b\u8aac\u660e\u30e9\u30d9\u30eb (&#8220;\u30ec\u30b3\u30e1\u30f3\u30c0\u30fc \u30b7\u30b9\u30c6\u30e0&#8221;\u3001&#8221;\u753b\u50cf\u5206\u985e&#8221; \u306a\u3069) \u3092\u8ffd\u52a0\u3057\u3001\u30e2\u30c7\u30eb\u306e\u4f5c\u6210\u8005\u3001\u4f5c\u6210\u65e5\u306a\u3069\u306e\u60c5\u5831\u3092\u8a18\u9332\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30e2\u30c7\u30eb\u627f\u8a8d\u30d7\u30ed\u30bb\u30b9: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3067\u306f\u3001\u30e2\u30c7\u30eb\u306e\u627f\u8a8d\u30d7\u30ed\u30bb\u30b9\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u306b\u3055\u307e\u3056\u307e\u306a\u627f\u8a8d\u30b9\u30c6\u30fc\u30b8 (&#8220;\u4fdd\u7559\u4e2d&#8221;\u3001&#8221;\u627f\u8a8d\u6e08\u307f&#8221; \u306a\u3069) \u3092\u5b9a\u7fa9\u3057\u3001\u30e2\u30c7\u30eb\u304c\u627f\u8a8d\u3055\u308c\u305f\u5f8c\u306b\u305d\u306e\u72b6\u614b\u3092 &#8220;\u904b\u7528&#8221; \u306b\u5909\u66f4\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30c1\u30fc\u30e0\u9593\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30e2\u30c7\u30eb\u306e\u54c1\u8cea\u3092\u78ba\u4fdd\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u4ed6\u306e MLflow \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u306e\u7d71\u5408: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30a8\u30f3\u30c9\u30c4\u30fc\u30a8\u30f3\u30c9\u306e\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u306e\u305f\u3081\u306b\u3001\u8ffd\u8de1\u3084\u30e2\u30c7\u30eb\u306a\u3069\u306e\u4ed6\u306e MLflow \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u30b7\u30fc\u30e0\u30ec\u30b9\u306b\u7d71\u5408\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306f\u3001MLflow \u8ffd\u8de1 API \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3057\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6e08\u307f\u306e\u30e2\u30c7\u30eb\u3092 MLflow \u30e2\u30c7\u30eb\u5f62\u5f0f\u3067\u4fdd\u5b58\u3057\u3066\u3001MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306b\u767b\u9332\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30d3\u30b8\u30e5\u30a2\u30eb \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb \u30e9\u30a4\u30d6\u30e9\u30ea\u5185\u306e\u30e2\u30c7\u30eb\u3092\u8996\u899a\u7684\u306b\u8868\u793a\u304a\u3088\u3073\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u30e6\u30fc\u30b6\u30fc \u30d5\u30ec\u30f3\u30c9\u30ea\u306a\u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 \u3053\u306e\u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb\u3092\u691c\u7d22\u3001\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u3001\u6bd4\u8f03\u3057\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3068\u30b9\u30c6\u30fc\u30bf\u30b9\u3092\u3088\u308a\u3088\u304f\u7406\u89e3\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u5168\u4f53\u3092\u3088\u308a\u9069\u5207\u306b\u6574\u7406\u304a\u3088\u3073\u8ffd\u8de1\u3059\u308b\u306e\u306b\u5f79\u7acb\u3064\u5f37\u529b\u3067\u67d4\u8edf\u306a\u30e2\u30c7\u30eb\u7ba1\u7406\u30c4\u30fc\u30eb\u3067\u3059\u3002 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30c1\u30fc\u30e0\u9593\u3067\u5171\u540c\u4f5c\u696d\u3092\u884c\u3044\u3001\u30e2\u30c7\u30eb\u3092\u5171\u6709\u3067\u304d\u308b\u305f\u3081\u3001\u30e2\u30c7\u30eb\u306e\u4fdd\u5b88\u6027\u3068\u751f\u7523\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u306e\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u306e\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u306f\u6bd4\u8f03\u7684\u7c21\u5358\u3067\u3001pip \u306a\u3069\u306e Python \u306e\u30d1\u30c3\u30b1\u30fc\u30b8\u7ba1\u7406\u30c4\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3067\u304d\u307e\u3059\u3002 \u8a73\u7d30\u306a\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u624b\u9806\u306f\u6b21\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u307e\u305a\u3001\u30b3\u30f3\u30d4\u30e5\u30fc\u30bf\u30fc\u306bPython\u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059(Python 3.6\u4ee5\u964d\u3092\u63a8\u5968)\u3002 Python \u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u78ba\u8a8d\u3059\u308b\u306b\u306f\u3001\u30b3\u30de\u30f3\u30c9\u30e9\u30a4\u30f3\u3067 or \u3092\u5165\u529b\u3057\u307e\u3059\u3002<code>python --version<\/code><code>python3 --version<\/code><\/li>\n\n\n\n<li>\u5fc5\u8981\u306b\u5fdc\u3058\u3066\u3001\u4ed6\u306e Python \u30e9\u30a4\u30d6\u30e9\u30ea\u3068\u306e\u7af6\u5408\u3092\u907f\u3051\u308b\u305f\u3081\u306b\u3001\u4eee\u60f3\u74b0\u5883\u306b MLflow \u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u3053\u3068\u3092\u304a\u52e7\u3081\u3057\u307e\u3059\u3002 \u307e\u305f\u306f\u306a\u3069\u306e\u30c4\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u3001\u4eee\u60f3\u74b0\u5883\u3092\u4f5c\u6210\u3067\u304d\u307e\u3059\u3002 \u4ee5\u4e0b\u306f\u3001\u3092\u4f7f\u7528\u3057\u3066\u4eee\u60f3\u74b0\u5883\u3092\u4f5c\u6210\u3059\u308b\u4f8b\u3067\u3059\u3002 Windows \u30e6\u30fc\u30b6\u30fc\u7528\u306e\u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3057\u307e\u3059\u3002 macOS \u304a\u3088\u3073 Linux \u30e6\u30fc\u30b6\u30fc\u5411\u3051\u306e\u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3057\u307e\u3059\u3002 bashCopy code \u3092\u4f7f\u7528\u3057\u3066\u4eee\u60f3\u74b0\u5883\u3092\u4f5c\u6210\u3059\u308b\u5834\u5408\u306f\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u53c2\u7167\u3067\u304d\u307e\u3059\u3002 lua\u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc<code>virtualenv<\/code><code>conda<\/code><code>virtualenv<\/code><code>pip install virtualenv virtualenv mlflow_env&nbsp;<\/code><code>mlflow_env\\Scripts\\activate&nbsp;<\/code><code>source mlflow_env\/bin\/activate&nbsp;<\/code><code>conda<\/code><code>conda create -n mlflow_env python=3.8 conda activate mlflow_env<\/code><\/li>\n\n\n\n<li>\u3092\u4f7f\u7528\u3057\u3066 MLflow \u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u307e\u3059\u3002 \u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3\u3067\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u5165\u529b\u3057\u307e\u3059: \u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3059\u308b\u3068\u3001MLflow \u3068\u305d\u306e\u4f9d\u5b58\u30e9\u30a4\u30d6\u30e9\u30ea\u304c\u81ea\u52d5\u7684\u306b\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u307e\u3059\u3002<code>pip<\/code><code>pip install mlflow<\/code><\/li>\n\n\n\n<li>MLflow \u304c\u6b63\u5e38\u306b\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u304b\u3069\u3046\u304b\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002 \u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3\u306b\u5165\u529b\u3057\u307e\u3059\u3002 MLflow \u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u756a\u53f7\u304c\u8868\u793a\u3055\u308c\u3066\u3044\u308b\u5834\u5408\u3001\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u306f\u6210\u529f\u3057\u3066\u3044\u307e\u3059\u3002<code>mlflow --version<\/code><\/li>\n<\/ol>\n\n\n\n<p>\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u3068\u3001\u8ffd\u8de1\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3001\u30e2\u30c7\u30eb\u3001\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306a\u3069\u306e MLflow \u306e\u6a5f\u80fd\u306e\u4f7f\u7528\u3092\u958b\u59cb\u3067\u304d\u307e\u3059\u3002 \u5177\u4f53\u7684\u306a\u4f7f\u7528\u65b9\u6cd5\u306b\u3064\u3044\u3066\u306f\u3001MLflow (https:\/\/mlflow.org\/docs\/latest\/index.html)\u306e\u516c\u5f0f\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u3092\u53c2\u7167\u3057\u3066\u304f\u3060\u3055\u3044<a href=\"https:\/\/mlflow.org\/docs\/latest\/index.html%EF%BC%89%E3%80%82\">\u3002<\/a><\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u306e\u74b0\u5883\u3068\u30d0\u30c3\u30af\u30a8\u30f3\u30c9\u30b9\u30c8\u30ec\u30fc\u30b8\u306e\u8a2d\u5b9a\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u306f\u3001\u3055\u307e\u3056\u307e\u306a\u30b7\u30ca\u30ea\u30aa\u306e\u30cb\u30fc\u30ba\u3092\u6e80\u305f\u3059\u305f\u3081\u306b\u3001\u8907\u6570\u306e\u74b0\u5883\u3068\u30d0\u30c3\u30af\u30a8\u30f3\u30c9 \u30b9\u30c8\u30ec\u30fc\u30b8\u69cb\u6210\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u307e\u3059\u3002 \u8a73\u7d30\u306a\u69cb\u6210\u65b9\u6cd5\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>MLflow \u304c\u30ed\u30fc\u30ab\u30eb \u30b5\u30fc\u30d0\u30fc\u307e\u305f\u306f\u30ea\u30e2\u30fc\u30c8 \u30b5\u30fc\u30d0\u30fc\u3067\u5b9f\u884c\u3067\u304d\u308b\u74b0\u5883\u3092\u69cb\u6210\u3057\u307e\u3059\u3002 \u65e2\u5b9a\u3067\u306f\u3001MLflow \u306f Web \u30b5\u30fc\u30d0\u30fc\u3092\u30ed\u30fc\u30ab\u30eb\u3067\u8d77\u52d5\u3057\u307e\u3059\u3002 MLflow \u30b5\u30fc\u30d0\u30fc\u3092\u8d77\u52d5\u3059\u308b\u306b\u306f\u3001\u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3\u3067\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u5165\u529b\u3057\u307e\u3059\u3002 \u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3059\u308b\u3053\u308c\u306b\u3088\u308a\u3001\u65e2\u5b9a\u306e\u30dd\u30fc\u30c8 (5000) \u3067 MLflow \u30b5\u30fc\u30d0\u30fc\u304c\u30ed\u30fc\u30ab\u30eb\u306b\u8d77\u52d5\u3057\u307e\u3059\u3002 MLflow \u306e\u30e6\u30fc\u30b6\u30fc \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u8868\u793a\u3059\u308b\u306b\u306f\u3001\u30d6\u30e9\u30a6\u30b6\u30fc\u304b\u3089\u30a2\u30af\u30bb\u30b9\u3057\u307e\u3059\u3002 \u7279\u5b9a\u306e\u30dd\u30fc\u30c8\u307e\u305f\u306f IP \u30a2\u30c9\u30ec\u30b9\u3067 MLflow \u30b5\u30fc\u30d0\u30fc\u3092\u5b9f\u884c\u3059\u308b\u5834\u5408\u306f\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3068\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001cssCopy \u30b3\u30fc\u30c9\u306e\u5834\u5408\u3001IP \u30a2\u30c9\u30ec\u30b9\u3068\u30dd\u30fc\u30c8\u3067 MLflow \u30b5\u30fc\u30d0\u30fc\u304c\u8d77\u52d5\u3057\u307e\u3059\u3002<code>mlflow server&nbsp;<\/code><code>http:\/\/localhost:5000<\/code><code>--host<\/code><code>--port<\/code><code>mlflow server --host 0.0.0.0 --port 6000&nbsp;<\/code><code>0.0.0.0<\/code><code>6000<\/code><\/li>\n\n\n\n<li>\u30d0\u30c3\u30af\u30a8\u30f3\u30c9 \u30b9\u30c8\u30ec\u30fc\u30b8\u306e\u69cb\u6210 MLflow \u3067\u306f\u3001\u30ed\u30fc\u30ab\u30eb \u30d5\u30a1\u30a4\u30eb \u30b7\u30b9\u30c6\u30e0\u3001\u30ea\u30e2\u30fc\u30c8 \u30b5\u30fc\u30d0\u30fc (Amazon S3\u3001Azure BLOB \u30b9\u30c8\u30ec\u30fc\u30b8\u306a\u3069)\u3001\u30c7\u30fc\u30bf\u30d9\u30fc\u30b9 (MySQL\u3001PostgreSQL \u306a\u3069) \u306a\u3069\u3001\u8907\u6570\u306e\u30d0\u30c3\u30af\u30a8\u30f3\u30c9 \u30b9\u30c8\u30ec\u30fc\u30b8 \u30aa\u30d7\u30b7\u30e7\u30f3\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u65e2\u5b9a\u3067\u306f\u3001MLflow \u306f\u30ed\u30fc\u30ab\u30eb \u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u306b\u30c7\u30fc\u30bf\u3092\u683c\u7d0d\u3057\u307e\u3059\u3002 MLflow \u306e\u30d0\u30c3\u30af\u30a8\u30f3\u30c9 \u30b9\u30c8\u30ec\u30fc\u30b8\u3092\u69cb\u6210\u3059\u308b\u306b\u306f\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3092\u4f7f\u7528\u3057\u307e\u3059\u3002 \u6b21\u306b\u4f8b\u3092\u3044\u304f\u3064\u304b\u793a\u3057\u307e\u3059\u3002<code>.\/mlruns<\/code><code>--backend-store-uri<\/code><ul><li>\u30ed\u30fc\u30ab\u30eb\u30d5\u30a1\u30a4\u30eb\u30b7\u30b9\u30c6\u30e0\u3092\u4f7f\u7528\u3059\u308b: javascript\u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3059\u308b<code>mlflow server --backend-store-uri file:\/path\/to\/local\/directory<\/code><\/li><li>Amazon S3: rubyCopy \u30b3\u30fc\u30c9\u3067\u306f\u3001\u30e9\u30a4\u30d6\u30e9\u30ea () \u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3001\u304a\u3088\u3073 AWS \u8a8d\u8a3c\u60c5\u5831\u304c\u6b63\u3057\u304f\u8a2d\u5b9a\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059 (\u74b0\u5883\u5909\u6570\u3092\u8a2d\u5b9a\u3059\u308b\u304b\u3001AWS CLI \u3092\u4f7f\u7528\u3057\u3066\u8a2d\u5b9a\u3067\u304d\u307e\u3059)\u3002<code>mlflow server --backend-store-uri s3:\/\/your-bucket-name\/path\/to\/mlruns&nbsp;<\/code><code>boto3<\/code><code>pip install boto3<\/code><code>AWS_ACCESS_KEY_ID<\/code><code>AWS_SECRET_ACCESS_KEY<\/code><\/li><li>Azure Blob Storage \u3092\u4f7f\u7528\u3059\u308b:rubyCopy \u30b3\u30fc\u30c9 Library() \u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3001\u304a\u3088\u3073 Azure Storage \u306e\u8cc7\u683c\u60c5\u5831\u304c\u6b63\u3057\u304f\u69cb\u6210\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059 (\u3053\u308c\u3092\u884c\u3046\u306b\u306f\u3001\u74b0\u5883\u5909\u6570\u3092\u8a2d\u5b9a\u3059\u308b\u304b\u3001\u30b3\u30de\u30f3\u30c9\u306b\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3092\u8ffd\u52a0\u3057\u307e\u3059)\u3002<code>mlflow server --backend-store-uri azure:\/\/your-container-name\/path\/to\/mlruns&nbsp;<\/code><code>azure-storage-blob<\/code><code>pip install azure-storage-blob<\/code><code>AZURE_STORAGE_CONNECTION_STRING<\/code><code>mlflow server<\/code><code>--azure-storage-connection-string<\/code><\/li><li>MySQL:bashCopy \u30b3\u30fc\u30c9\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e9\u30a4\u30d6\u30e9\u30ea () \u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002<code>mlflow server --backend-store-uri mysql:\/\/username:password@host:port\/database&nbsp;<\/code><code>mysql-connector-python<\/code><code>pip install mysql-connector-python<\/code><\/li><li>PostgreSQL:bashCopy code \u30e9\u30a4\u30d6\u30e9\u30ea\u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u3066\u304f\u3060\u3055\u3044()\u3002<code>mlflow server --backend-store-uri postgresql:\/\/username:password@host:port\/database&nbsp;<\/code><code>psycopg2<\/code><code>pip install psycopg2<\/code><\/li><\/ul>\u3053\u306e\u69cb\u6210\u306f\u3001MLflow \u30b5\u30fc\u30d0\u30fc\u3092\u8d77\u52d5\u3059\u308b\u3068\u304d\u306b\u306e\u307f\u884c\u3046\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u5b9f\u969b\u306b\u306f\u3001\u3053\u308c\u3089\u306e\u69cb\u6210\u3092\u6b21\u306e\u3088\u3046\u306a\u4ed6\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3068\u7d44\u307f\u5408\u308f\u305b\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u306a\u3069) \u3092\u4f7f\u7528\u3057\u3066\u3001\u5b9f\u969b\u306e\u30cb\u30fc\u30ba\u306b\u5fdc\u3058\u3066 MLflow \u30e9\u30f3\u30bf\u30a4\u30e0\u74b0\u5883\u3092\u30ab\u30b9\u30bf\u30de\u30a4\u30ba\u3057\u307e\u3059\u3002<code>--default-artifact-root<\/code><code>--host<\/code><\/p>\n\n\n\n<p>Amazon S3 \u3092\u30d0\u30c3\u30af\u30a8\u30f3\u30c9\u30b9\u30c8\u30a2\u304a\u3088\u3073\u30c7\u30d5\u30a9\u30eb\u30c8\u306e\u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8\u30b9\u30c8\u30a2\u3068\u3057\u3066\u4f7f\u7528\u3057\u3066\u3001\u6307\u5b9a\u3055\u308c\u305f IP \u30a2\u30c9\u30ec\u30b9\u3068\u30dd\u30fc\u30c8\u3067 MLflow \u30b5\u30fc\u30d0\u30fc\u3092\u5b9f\u884c\u3059\u308b\u65b9\u6cd5\u3092\u793a\u3059\u5b8c\u5168\u306a\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">arduinoCopy code<code>mlflow server \\\n  --backend-store-uri s3:\/\/your-bucket-name\/path\/to\/mlruns \\\n  --default-artifact-root s3:\/\/your-bucket-name\/path\/to\/artifacts \\\n  --host 0.0.0.0 \\\n  --port 6000\n<\/code><\/pre>\n\n\n\n<p>MLflow \u306e\u74b0\u5883\u3068\u30d0\u30c3\u30af\u30a8\u30f3\u30c9 \u30b9\u30c8\u30ec\u30fc\u30b8\u3092\u69cb\u6210\u3057\u305f\u3089\u3001\u8ffd\u8de1\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3001\u30e2\u30c7\u30eb\u3001\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306a\u3069\u306e MLflow \u306e\u6a5f\u80fd\u306e\u4f7f\u7528\u3092\u958b\u59cb\u3067\u304d\u307e\u3059\u3002 MLflow API \u3092\u4f7f\u7528\u3059\u308b\u5834\u5408\u306f\u3001MLflow \u30b5\u30fc\u30d0\u30fc\u306e URI \u3092\u6307\u5b9a\u3057\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001Python \u30b9\u30af\u30ea\u30d7\u30c8\u3067\u306f\u3001\u6b21\u306e\u30b3\u30fc\u30c9\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow\n\nmlflow.set_tracking_uri(\"http:\/\/your-mlflow-server-ip:your-mlflow-server-port\")\n<\/code><\/pre>\n\n\n\n<p>\u5b9f\u969b\u306e\u72b6\u6cc1\u306b\u5408\u308f\u305b\u3066\u4ea4\u63db\u3057\u3066\u304f\u3060\u3055\u3044\u3002 \u30c8\u30e9\u30c3\u30ad\u30f3\u30b0 URI \u3092\u69cb\u6210\u3057\u305f\u3089\u3001MLflow API \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u306e\u8a18\u9332\u3001\u30e2\u30c7\u30eb\u306e\u4fdd\u5b58\u306a\u3069\u3092\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<code>your-mlflow-server-ip<\/code><code>your-mlflow-server-port<\/code><\/p>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u306e\u74b0\u5883\u3068\u30d0\u30c3\u30af\u30a8\u30f3\u30c9 \u30b9\u30c8\u30ec\u30fc\u30b8\u306e\u69cb\u6210\u306f\u6bd4\u8f03\u7684\u7c21\u5358\u3067\u3001\u3044\u304f\u3064\u304b\u306e\u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3\u5f15\u6570\u3060\u3051\u3067\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002 MLflow \u3092\u9069\u5207\u306b\u69cb\u6210\u3059\u308b\u3053\u3068\u3067\u3001\u3088\u308a\u52b9\u7387\u7684\u306a\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u3092\u5b9f\u73fe\u3057\u3001\u30c1\u30fc\u30e0\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30c7\u30fc\u30bf\u5171\u6709\u306e\u8981\u6c42\u3092\u3088\u308a\u9069\u5207\u306b\u6e80\u305f\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow \u8ffd\u8de1\u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3059\u308b\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3059\u308b<\/p>\n\n\n\n<p>MLflow \u8ffd\u8de1\u306f\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u6210\u679c\u7269\u306a\u3069\u3001\u6a5f\u68b0\u5b66\u7fd2\u5b9f\u9a13\u304b\u3089\u306e\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u304a\u3088\u3073\u8868\u793a\u3059\u308b MLflow \u306e\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u3059\u3002 MLflow \u8ffd\u8de1\u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3059\u308b\u306b\u306f\u3001\u6b21\u306e\u624b\u9806\u306b\u5f93\u3046\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u307e\u305a\u3001\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u30a4\u30f3\u30dd\u30fc\u30c8\u3057\u3001MLflow \u30b5\u30fc\u30d0\u30fc\u306e URI \u3092\u8a2d\u5b9a\u3057\u307e\u3059\u3002 Python \u30b9\u30af\u30ea\u30d7\u30c8\u3067\u306f\u3001\u6b21\u306e\u30b3\u30fc\u30c9\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059: python \u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3057\u3001\u5fc5\u8981\u306b\u5fdc\u3058\u3066\u7f6e\u304d\u63db\u3048\u3066\u304f\u3060\u3055\u3044\u3002<code>mlflow<\/code><code>import mlflow mlflow.set_tracking_uri(\"http:\/\/your-mlflow-server-ip:your-mlflow-server-port\")&nbsp;<\/code><code>your-mlflow-server-ip<\/code><code>your-mlflow-server-port<\/code><\/li>\n\n\n\n<li>\u65b0\u3057\u3044\u5b9f\u9a13\u3092\u307e\u3060\u4f5c\u6210\u3057\u3066\u3044\u306a\u3044\u5834\u5408\u306f\u4f5c\u6210\u3057\u307e\u3059\u3002 \u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u3092\u4f5c\u6210\u3057\u3001\u540d\u524d\u3092\u4ed8\u3051\u307e\u3059: python\u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3059\u308b\u5b9f\u9a13\u304c\u65e2\u306b\u5b58\u5728\u3059\u308b\u5834\u5408\u306f\u3001\u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066\u305d\u306e ID \u3092\u53d6\u5f97\u3067\u304d\u307e\u3059\u3002<code>mlflow.create_experiment<\/code><code>experiment_id = mlflow.create_experiment(\"your_experiment_name\")&nbsp;<\/code><code>mlflow.get_experiment_by_name<\/code><code>experiment = mlflow.get_experiment_by_name(\"your_experiment_name\") experiment_id = experiment.experiment_id<\/code><\/li>\n\n\n\n<li>\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u30de\u30cd\u30fc\u30b8\u30e3\u3092\u4f7f\u7528\u3057\u3066\u3001\u65b0\u3057\u3044\u5b9f\u884c\u3092\u958b\u59cb\u3057\u307e\u3059\u3002 \u30b9\u30c6\u30fc\u30c8\u30e1\u30f3\u30c8 \u30d6\u30ed\u30c3\u30af\u3067\u306f\u3001MLflow API \u3092\u4f7f\u7528\u3057\u3066\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u6210\u679c\u7269\u306a\u3069\u306e\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3067\u304d\u307e\u3059\u3002 pythonCopy \u30b3\u30fc\u30c9 \u30b9\u30c6\u30fc\u30c8\u30e1\u30f3\u30c8 \u30d6\u30ed\u30c3\u30af\u3067\u306f\u3001\u95a2\u6570\u3092\u8907\u6570\u56de\u547c\u3073\u51fa\u3057\u3066\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u306e\u7570\u306a\u308b\u5024\u3092\u8a18\u9332\u3067\u304d\u307e\u3059 (\u305f\u3068\u3048\u3070\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0 \u30d7\u30ed\u30bb\u30b9\u306e\u5404\u30b9\u30c6\u30c3\u30d7\u3067)\u3002 MLflow \u3067\u306f\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u306e\u5c65\u6b74\u30c7\u30fc\u30bf\u304c\u81ea\u52d5\u7684\u306b\u8a18\u9332\u3055\u308c\u308b\u305f\u3081\u3001\u30e6\u30fc\u30b6\u30fc \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3067\u30e1\u30c8\u30ea\u30c3\u30af\u306e\u50be\u5411\u3092\u8868\u793a\u304a\u3088\u3073\u6bd4\u8f03\u3067\u304d\u307e\u3059\u3002<code>mlflow.start_run<\/code><code>with<\/code><code>with mlflow.start_run(experiment_id=experiment_id): # \u8bb0\u5f55\u53c2\u6570 mlflow.log_param(\"param_name\", param_value) # \u8bb0\u5f55\u6307\u6807 mlflow.log_metric(\"metric_name\", metric_value) # \u8bb0\u5f55\u5de5\u4ef6\uff08\u5982\u6a21\u578b\u6587\u4ef6\u3001\u56fe\u7247\u7b49\uff09 mlflow.log_artifact(\"path\/to\/local\/file\", \"artifact_subfolder\") # ...&nbsp;<\/code><code>with<\/code><code>mlflow.log_metric<\/code><\/li>\n\n\n\n<li>MLflow \u30e6\u30fc\u30b6\u30fc \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3067\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8868\u793a\u3057\u307e\u3059\u3002 \u4f5c\u6210\u3057\u305f\u5b9f\u9a13\u306b\u30a2\u30af\u30bb\u30b9\u3057\u3066\u9078\u629e\u3057\u307e\u3059\u3002 \u8a18\u9332\u3055\u308c\u305f\u5b9f\u884c\u306e\u30ea\u30b9\u30c8\u3068\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8\u306a\u3069\u306e\u95a2\u9023\u30c7\u30fc\u30bf\u3092\u8868\u793a\u3067\u304d\u307e\u3059\u3002 \u5b9f\u884c\u306e ID \u3092\u30af\u30ea\u30c3\u30af\u3059\u308b\u3068\u3001\u8a73\u7d30\u60c5\u5831\u304c\u8868\u793a\u3055\u308c\u307e\u3059\u3002<code>http:\/\/your-mlflow-server-ip:your-mlflow-server-port<\/code><\/li>\n<\/ol>\n\n\n\n<p>\u3053\u308c\u3089\u306e\u624b\u9806\u3067\u306f\u3001MLflow \u8ffd\u8de1\u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u304a\u3088\u3073\u8868\u793a\u3057\u3001\u6a5f\u68b0\u5b66\u7fd2\u5b9f\u9a13\u3092\u3088\u308a\u9069\u5207\u306b\u6574\u7406\u304a\u3088\u3073\u7ba1\u7406\u3067\u304d\u307e\u3059\u3002 MLflow \u306b\u306f\u8c4a\u5bcc\u306a API \u3068\u30d3\u30b8\u30e5\u30a2\u30eb \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u304c\u7528\u610f\u3055\u308c\u3066\u304a\u308a\u3001\u6700\u9069\u306a\u30e2\u30c7\u30eb\u69cb\u6210\u3092\u3059\u3070\u3084\u304f\u898b\u3064\u3051\u3066\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3068\u4fdd\u5b88\u6027\u3092\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow UI\u3067\u5b9f\u9a13\u7d50\u679c\u3092\u898b\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u306f\u3001\u5b9f\u9a13\u7d50\u679c\u3092\u7c21\u5358\u306b\u8868\u793a\u304a\u3088\u3073\u6bd4\u8f03\u3067\u304d\u308b\u30e6\u30fc\u30b6\u30fc \u30d5\u30ec\u30f3\u30c9\u30ea\u306a Web \u30e6\u30fc\u30b6\u30fc \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9 (UI) \u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 MLflow UI \u3092\u4f7f\u7528\u3057\u3066\u30e9\u30dc\u306e\u7d50\u679c\u3092\u8868\u793a\u3059\u308b\u65b9\u6cd5\u306e\u8a73\u7d30\u306a\u8aac\u660e\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u307e\u305a\u3001MLflow \u30b5\u30fc\u30d0\u30fc\u304c\u5b9f\u884c\u4e2d\u3067\u3042\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002 \u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3\u3067\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u5165\u529b\u3057\u3066 MLflow \u30b5\u30fc\u30d0\u30fc\u3092\u8d77\u52d5\u3057\u307e\u3059 (\u5fc5\u8981\u306b\u5fdc\u3058\u3066\u30d0\u30c3\u30af\u30a8\u30f3\u30c9 \u30b9\u30c8\u30ec\u30fc\u30b8\u3068\u30dd\u30fc\u30c8\u756a\u53f7\u3092\u5909\u66f4\u3057\u307e\u3059)\u3002 javascript\u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3059\u308b<code>mlflow server --backend-store-uri file:\/path\/to\/local\/directory --host 0.0.0.0 --port 5000<\/code><\/li>\n\n\n\n<li>\u30d6\u30e9\u30a6\u30b6\u30fc\u3092\u4f7f\u7528\u3057\u3066 MLflow UI \u306b\u30a2\u30af\u30bb\u30b9\u3057\u307e\u3059\u3002 MLflow \u30b5\u30fc\u30d0\u30fc\u306e\u30a2\u30c9\u30ec\u30b9\u3092 : \u306e\u5f62\u5f0f\u3067\u5165\u529b\u3057\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001MLflow \u30b5\u30fc\u30d0\u30fc\u304c\u30ed\u30fc\u30ab\u30eb\u3067\u5b9f\u884c\u3055\u308c\u3066\u3044\u308b\u5834\u5408\u3001\u65e2\u5b9a\u306e\u30a2\u30c9\u30ec\u30b9\u306f \u3067\u3059\u3002<code>http:\/\/your-mlflow-server-ip:your-mlflow-server-port<\/code><code>http:\/\/localhost:5000<\/code><\/li>\n\n\n\n<li>MLflow UI \u3067\u306f\u3001\u5b9f\u9a13\u306e\u4e00\u89a7\u3092\u8868\u793a\u3067\u304d\u307e\u3059\u3002 \u5b9f\u9a13\u540d\u3092\u30af\u30ea\u30c3\u30af\u3057\u3066\u3001\u5b9f\u9a13\u306e\u8a73\u7d30\u30da\u30fc\u30b8\u306b\u79fb\u52d5\u3057\u307e\u3059\u3002 \u3053\u306e\u30da\u30fc\u30b8\u306b\u306f\u3001\u5b9f\u9a13\u306e\u3059\u3079\u3066\u306e\u5b9f\u884c\u3068\u3001\u5404\u5b9f\u884c\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30e6\u30fc\u30b6\u30fc\u3001\u5b9f\u884c\u6642\u9593\u306a\u3069\u306e\u60c5\u5831\u3092\u793a\u3059\u30c6\u30fc\u30d6\u30eb\u304c\u8868\u793a\u3055\u308c\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30c6\u30fc\u30d6\u30eb \u30d8\u30c3\u30c0\u30fc\u306e\u5217\u540d\u3092\u30af\u30ea\u30c3\u30af\u3059\u308b\u3068\u3001\u5b9f\u884c\u3092\u4e26\u3079\u66ff\u3048\u3066\u3001\u6700\u9069\u306a\u30e1\u30c8\u30ea\u30c3\u30af\u5024\u3092\u6301\u3064\u5b9f\u884c\u3092\u3059\u3070\u3084\u304f\u898b\u3064\u3051\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u30c6\u30fc\u30d6\u30eb\u306e\u53f3\u4e0a\u9685\u306b\u3042\u308b\u691c\u7d22\u30dc\u30c3\u30af\u30b9\u3092\u4f7f\u7528\u3057\u3066\u5b9f\u884c\u3092\u30d5\u30a3\u30eb\u30bf\u30fc\u51e6\u7406\u3057\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30e9\u30d9\u30eb\u306a\u3069\u306b\u57fa\u3065\u3044\u3066\u7279\u5b9a\u306e\u5b9f\u884c\u3092\u691c\u7d22\u3059\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u5b9f\u884c\u3092\u9078\u629e\u3057\u3001\u305d\u306e ID \u3092\u30af\u30ea\u30c3\u30af\u3059\u308b\u3068\u3001\u5b9f\u884c\u306e\u8a73\u7d30\u30da\u30fc\u30b8\u304c\u8868\u793a\u3055\u308c\u307e\u3059\u3002 \u3053\u306e\u30da\u30fc\u30b8\u3067\u306f\u3001\u5b9f\u884c\u306e\u3059\u3079\u3066\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30e9\u30d9\u30eb\u306e\u307b\u304b\u3001MLflow \u306b\u3088\u3063\u3066\u81ea\u52d5\u7684\u306b\u8a18\u9332\u3055\u308c\u308b\u30b7\u30b9\u30c6\u30e0\u60c5\u5831 (\u5b9f\u884c\u306e\u958b\u59cb\u6642\u523b\u3068\u7d42\u4e86\u6642\u523b\u3001\u30bd\u30fc\u30b9 \u30b3\u30fc\u30c9 \u30d0\u30fc\u30b8\u30e7\u30f3\u306a\u3069) \u3092\u8868\u793a\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u5b9f\u884c\u306e\u8a73\u7d30\u30da\u30fc\u30b8\u306e\u4e0b\u306b\u3001\u5b9f\u884c\u4e2d\u306e\u6210\u679c\u7269\u306e\u30ea\u30b9\u30c8\u3092\u8868\u793a\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8\u306b\u306f\u3001\u30e2\u30c7\u30eb \u30d5\u30a1\u30a4\u30eb\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0 \u30c7\u30fc\u30bf\u3001\u751f\u6210\u3055\u308c\u305f\u30a4\u30e1\u30fc\u30b8\u306a\u3069\u304c\u542b\u307e\u308c\u307e\u3059\u3002 \u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8\u540d\u3092\u30af\u30ea\u30c3\u30af\u3059\u308b\u3068\u3001\u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8\u3092\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u307e\u305f\u306f\u30d7\u30ec\u30d3\u30e5\u30fc\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30e1\u30c8\u30ea\u30c3\u30af\u306e\u5c65\u6b74\u30c7\u30fc\u30bf (\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0 \u30d7\u30ed\u30bb\u30b9\u306e\u5404\u30b9\u30c6\u30c3\u30d7\u306a\u3069) \u3092\u8a18\u9332\u3059\u308b\u5834\u5408\u3001[\u5b9f\u884c\u306e\u8a73\u7d30] \u30da\u30fc\u30b8\u306e [\u30e1\u30c8\u30ea\u30c3\u30af] \u30bb\u30af\u30b7\u30e7\u30f3\u3067\u30e1\u30c8\u30ea\u30c3\u30af\u306e\u50be\u5411\u3092\u8868\u793a\u3067\u304d\u307e\u3059\u3002 \u30a4\u30f3\u30b8\u30b1\u30fc\u30bf\u30fc\u30c1\u30e3\u30fc\u30c8\u306b\u30ab\u30fc\u30bd\u30eb\u3092\u5408\u308f\u305b\u308b\u3068\u3001\u7279\u5b9a\u306e\u5024\u3068\u30b9\u30c6\u30c3\u30d7\u304c\u8868\u793a\u3055\u308c\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>MLflow UI \u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u5b9f\u9a13\u7d50\u679c\u3092\u7c21\u5358\u306b\u8868\u793a\u3001\u6bd4\u8f03\u3001\u5206\u6790\u3057\u3066\u3001\u6700\u9069\u306a\u30e2\u30c7\u30eb\u69cb\u6210\u3092\u898b\u3064\u3051\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3055\u3089\u306b\u3001MLflow UI \u306b\u306f\u3001\u5b9f\u9a13\u3068\u5b9f\u884c\u306e\u30a4\u30f3\u30dd\u30fc\u30c8\u6a5f\u80fd\u3068\u30a8\u30af\u30b9\u30dd\u30fc\u30c8\u6a5f\u80fd\u304c\u7528\u610f\u3055\u308c\u3066\u3044\u308b\u305f\u3081\u3001\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3068\u5171\u6709\u3057\u305f\u308a\u3001\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u4ed6\u306e MLflow \u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u306b\u79fb\u884c\u3057\u305f\u308a\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u5b9f\u9a13\u7d50\u679c\u306e\u30af\u30a8\u30ea\u3068\u6bd4\u8f03\u3059\u308b\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u3067\u306f\u3001Python API \u3068 Web \u30e6\u30fc\u30b6\u30fc \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9 (UI) \u3092\u4f7f\u7528\u3057\u3066\u3001\u5b9f\u9a13\u7d50\u679c\u306e\u30af\u30a8\u30ea\u3068\u6bd4\u8f03\u3092\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u3092\u884c\u3046\u65b9\u6cd5\u306e\u8a73\u7d30\u306a\u8aac\u660e\u306f\u6b21\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<p><strong>Python API \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u7d50\u679c\u3092\u30af\u30a8\u30ea\u3057\u3066\u6bd4\u8f03\u3057\u307e\u3059\u3002<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u307e\u305a\u3001\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u30a4\u30f3\u30dd\u30fc\u30c8\u3057\u3001MLflow \u30b5\u30fc\u30d0\u30fc\u306e URI \u3092\u8a2d\u5b9a\u3057\u307e\u3059\u3002 Python \u30b9\u30af\u30ea\u30d7\u30c8\u3067\u306f\u3001\u6b21\u306e\u30b3\u30fc\u30c9\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059: python \u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3057\u3001\u5fc5\u8981\u306b\u5fdc\u3058\u3066\u7f6e\u304d\u63db\u3048\u3066\u304f\u3060\u3055\u3044\u3002<code>mlflow<\/code><code>import mlflow mlflow.set_tracking_uri(\"http:\/\/your-mlflow-server-ip:your-mlflow-server-port\")&nbsp;<\/code><code>your-mlflow-server-ip<\/code><code>your-mlflow-server-port<\/code><\/li>\n\n\n\n<li>\u3053\u306e\u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066\u3001\u5b9f\u9a13\u7d50\u679c\u3092\u7167\u4f1a\u3057\u307e\u3059\u3002 \u3053\u306e\u95a2\u6570\u306f\u3001\u5b9f\u9a13 ID \u306e\u30ea\u30b9\u30c8\u3068\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30e9\u30d9\u30eb\u306a\u3069\u306e\u6761\u4ef6\u306b\u57fa\u3065\u3044\u3066\u5b9f\u884c\u3092\u30d5\u30a3\u30eb\u30bf\u30fc\u51e6\u7406\u3067\u304d\u308b\u30aa\u30d7\u30b7\u30e7\u30f3\u306e\u30af\u30a8\u30ea\u5f0f\u3092\u53d7\u3051\u5165\u308c\u307e\u3059: pythonCopy \u30b3\u30fc\u30c9\u95a2\u6570\u306f\u3001\u6761\u4ef6\u3092\u6e80\u305f\u3059\u5b9f\u884c\u60c5\u5831\u3092\u542b\u3080 DataFrame \u3092\u8fd4\u3057\u307e\u3059\u3002 Pandas\u306e\u6a5f\u80fd\u3092\u4f7f\u7528\u3057\u3066\u3001\u305f\u3068\u3048\u3070\u3001\u6700\u9069\u306a\u30e1\u30c8\u30ea\u30c3\u30af\u5024\u3092\u6301\u3064\u5b9f\u884c\u3092\u898b\u3064\u3051\u308b\u305f\u3081\u306b\u3001\u30c7\u30fc\u30bf\u30d5\u30ec\u30fc\u30e0\u3092\u3055\u3089\u306b\u51e6\u7406\u304a\u3088\u3073\u5206\u6790\u3067\u304d\u307e\u3059\u3002<code>mlflow.search_runs<\/code><code>experiment = mlflow.get_experiment_by_name(\"your_experiment_name\") runs = mlflow.search_runs( experiment_ids=[experiment.experiment_id], filter_string=\"params.param_name &gt; 0.5 and metrics.metric_name &lt; 0.1\" )&nbsp;<\/code><code>mlflow.search_runs<\/code><code>best_run = runs.loc[runs[\"metrics.metric_name\"].idxmin()]<\/code><\/li>\n<\/ol>\n\n\n\n<p><strong>Web \u30e6\u30fc\u30b6\u30fc \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30fc\u30b9\u3092\u4f7f\u7528\u3057\u3066\u30c6\u30b9\u30c8\u7d50\u679c\u3092\u30af\u30a8\u30ea\u3057\u3066\u6bd4\u8f03\u3059\u308b\u306b\u306f:<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u30d6\u30e9\u30a6\u30b6\u30fc\u3092\u4f7f\u7528\u3057\u3066 MLflow UI \u306b\u30a2\u30af\u30bb\u30b9\u3057\u307e\u3059\u3002 MLflow \u30b5\u30fc\u30d0\u30fc\u306e\u30a2\u30c9\u30ec\u30b9\u3092 : \u306e\u5f62\u5f0f\u3067\u5165\u529b\u3057\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001MLflow \u30b5\u30fc\u30d0\u30fc\u304c\u30ed\u30fc\u30ab\u30eb\u3067\u5b9f\u884c\u3055\u308c\u3066\u3044\u308b\u5834\u5408\u3001\u65e2\u5b9a\u306e\u30a2\u30c9\u30ec\u30b9\u306f \u3067\u3059\u3002<code>http:\/\/your-mlflow-server-ip:your-mlflow-server-port<\/code><code>http:\/\/localhost:5000<\/code><\/li>\n\n\n\n<li>MLflow UI \u3067\u3001\u5b9f\u9a13\u540d\u3092\u30af\u30ea\u30c3\u30af\u3057\u3066\u5b9f\u9a13\u306e\u8a73\u7d30\u30da\u30fc\u30b8\u306b\u79fb\u52d5\u3057\u307e\u3059\u3002 \u3053\u306e\u30da\u30fc\u30b8\u306b\u306f\u3001\u5b9f\u9a13\u306e\u3059\u3079\u3066\u306e\u5b9f\u884c\u3068\u3001\u5404\u5b9f\u884c\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30e6\u30fc\u30b6\u30fc\u3001\u5b9f\u884c\u6642\u9593\u306a\u3069\u306e\u60c5\u5831\u3092\u793a\u3059\u30c6\u30fc\u30d6\u30eb\u304c\u8868\u793a\u3055\u308c\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30c6\u30fc\u30d6\u30eb\u306e\u53f3\u4e0a\u9685\u306b\u3042\u308b\u691c\u7d22\u30dc\u30c3\u30af\u30b9\u3092\u4f7f\u7528\u3057\u3066\u5b9f\u884c\u3092\u30d5\u30a3\u30eb\u30bf\u30fc\u51e6\u7406\u3057\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30e9\u30d9\u30eb\u306a\u3069\u306b\u57fa\u3065\u3044\u3066\u7279\u5b9a\u306e\u5b9f\u884c\u3092\u691c\u7d22\u3057\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u304c 0.5 \u3088\u308a\u5927\u304d\u304f\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u304c 0.1 \u672a\u6e80\u306e\u5b9f\u884c\u3092\u691c\u7d22\u3059\u308b\u306b\u306f\u3001\u6b21\u306e\u30af\u30a8\u30ea\u5f0f\u3092\u5165\u529b\u3067\u304d\u307e\u3059\u3002<code>param_name<\/code><code>metric_name<\/code><code>params.param_name &gt; 0.5 and metrics.metric_name &lt; 0.1<\/code><\/li>\n\n\n\n<li>\u30c6\u30fc\u30d6\u30eb \u30d8\u30c3\u30c0\u30fc\u306e\u5217\u540d\u3092\u30af\u30ea\u30c3\u30af\u3059\u308b\u3068\u3001\u5b9f\u884c\u3092\u4e26\u3079\u66ff\u3048\u3066\u3001\u6700\u9069\u306a\u30e1\u30c8\u30ea\u30c3\u30af\u5024\u3092\u6301\u3064\u5b9f\u884c\u3092\u3059\u3070\u3084\u304f\u898b\u3064\u3051\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u8907\u6570\u306e\u5b9f\u884c\u3092\u6bd4\u8f03\u3059\u308b\u306b\u306f\u3001\u8868\u306e\u5de6\u5074\u306b\u3042\u308b\u30c1\u30a7\u30c3\u30af\u30dc\u30c3\u30af\u30b9\u3092\u30aa\u30f3\u306b\u3057\u3066\u3001\u8868\u306e\u4e0a\u306b\u3042\u308b[\u6bd4\u8f03]\u30dc\u30bf\u30f3\u3092\u30af\u30ea\u30c3\u30af\u3057\u307e\u3059\u3002 \u6bd4\u8f03\u30da\u30fc\u30b8\u3067\u306f\u3001\u9078\u629e\u3057\u305f\u5b9f\u884c\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30e9\u30d9\u30eb\u306e\u307b\u304b\u3001\u5b9f\u884c\u306e\u958b\u59cb\u6642\u523b\u3068\u7d42\u4e86\u6642\u523b\u3001\u30bd\u30fc\u30b9 \u30b3\u30fc\u30c9 \u30d0\u30fc\u30b8\u30e7\u30f3\u306a\u3069\u3001MLflow \u306b\u3088\u3063\u3066\u81ea\u52d5\u7684\u306b\u8a18\u9332\u3055\u308c\u308b\u30b7\u30b9\u30c6\u30e0\u60c5\u5831\u3092\u8868\u793a\u3067\u304d\u307e\u3059\u3002 \u30e1\u30c8\u30ea\u30c3\u30af\u3092\u8868\u793a\u3059\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059<\/li>\n<\/ol>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u7570\u306a\u308b\u5b9f\u884c\u9593\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u9055\u3044\u3092\u5206\u6790\u3059\u308b\u305f\u3081\u306e\u30c8\u30ec\u30f3\u30c9\u30d7\u30ed\u30c3\u30c8\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"6\">\n<li>\u6bd4\u8f03\u30da\u30fc\u30b8\u306e\u4e0a\u90e8\u306b\u3042\u308b\u300c\u6563\u5e03\u56f3\u300d\u30dc\u30bf\u30f3\u3068\u300c\u5e73\u884c\u5ea7\u6a19\u30d7\u30ed\u30c3\u30c8\u300d\u30dc\u30bf\u30f3\u3092\u30af\u30ea\u30c3\u30af\u3057\u3066\u3001\u9078\u629e\u3057\u305f\u5b9f\u884c\u306e\u6563\u5e03\u56f3\u3068\u5e73\u884c\u5ea7\u6a19\u30d7\u30ed\u30c3\u30c8\u3092\u8868\u793a\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u30b0\u30e9\u30d5\u306f\u3001\u305f\u3068\u3048\u3070\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3068\u30e1\u30c8\u30ea\u30c3\u30af\u306e\u9593\u306e\u76f8\u95a2\u95a2\u4fc2\u3092\u898b\u3064\u3051\u308b\u305f\u3081\u306b\u3001\u8907\u6570\u306e\u5b9f\u884c\u9593\u306e\u95a2\u4fc2\u3092\u3088\u308a\u8996\u899a\u7684\u306b\u5206\u6790\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>MLflow Python API \u3068 Web \u30e6\u30fc\u30b6\u30fc \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u5b9f\u9a13\u7d50\u679c\u3092\u7c21\u5358\u306b\u30af\u30a8\u30ea\u3057\u3066\u6bd4\u8f03\u3057\u3001\u6700\u9069\u306a\u30e2\u30c7\u30eb\u69cb\u6210\u3092\u898b\u3064\u3051\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u6a5f\u80fd\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3068\u4fdd\u5b88\u6027\u3092\u5411\u4e0a\u3055\u305b\u308b\u3068\u540c\u6642\u306b\u3001\u30c1\u30fc\u30e0\u306b\u3088\u308a\u826f\u3044\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30c7\u30fc\u30bf\u5171\u6709\u306e\u30b5\u30dd\u30fc\u30c8\u3092\u63d0\u4f9b\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u30d1\u30e9\u30e1\u30fc\u30bf\u3001\u6307\u6a19\u3001\u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8\u306e\u8a18\u9332\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u3067\u306f\u3001Tracking API \u3092\u4f7f\u7528\u3057\u3066\u3001\u5b9f\u9a13\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc (\u30cf\u30a4\u30d1\u30fc\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u306a\u3069)\u3001\u30e1\u30c8\u30ea\u30c3\u30af (\u7cbe\u5ea6\u3001\u640d\u5931\u306a\u3069)\u3001\u6210\u679c\u7269 (\u30e2\u30c7\u30eb \u30d5\u30a1\u30a4\u30eb\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0 \u30c7\u30fc\u30bf\u3001\u751f\u6210\u3055\u308c\u305f\u753b\u50cf\u306a\u3069) \u3092\u8a18\u9332\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u60c5\u5831\u3092\u8a18\u9332\u3059\u308b\u65b9\u6cd5\u306e\u8a73\u7d30\u306a\u8aac\u660e\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u307e\u305a\u3001\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u30a4\u30f3\u30dd\u30fc\u30c8\u3057\u3001MLflow \u30b5\u30fc\u30d0\u30fc\u306e URI \u3092\u8a2d\u5b9a\u3057\u307e\u3059\u3002 Python \u30b9\u30af\u30ea\u30d7\u30c8\u3067\u306f\u3001\u6b21\u306e\u30b3\u30fc\u30c9\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059: python \u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3057\u3001\u5fc5\u8981\u306b\u5fdc\u3058\u3066\u7f6e\u304d\u63db\u3048\u3066\u304f\u3060\u3055\u3044\u3002<code>mlflow<\/code><code>import mlflow mlflow.set_tracking_uri(\"http:\/\/your-mlflow-server-ip:your-mlflow-server-port\")&nbsp;<\/code><code>your-mlflow-server-ip<\/code><code>your-mlflow-server-port<\/code><\/li>\n\n\n\n<li>\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u30de\u30cd\u30fc\u30b8\u30e3\u3092\u4f7f\u7528\u3057\u3066\u3001\u65b0\u3057\u3044\u5b9f\u884c\u3092\u958b\u59cb\u3057\u307e\u3059\u3002 \u30b9\u30c6\u30fc\u30c8\u30e1\u30f3\u30c8 \u30d6\u30ed\u30c3\u30af\u3067\u306f\u3001MLflow API \u3092\u4f7f\u7528\u3057\u3066\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u6210\u679c\u7269\u306a\u3069\u306e\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u8a18\u9332\u3067\u304d\u307e\u3059\u3002<code>mlflow.start_run<\/code><code>with<\/code><code>with mlflow.start_run(): # \u8bb0\u5f55\u53c2\u6570 mlflow.log_param(\"param_name\", param_value) # \u8bb0\u5f55\u6307\u6807 mlflow.log_metric(\"metric_name\", metric_value) # \u8bb0\u5f55\u5de5\u4ef6\uff08\u5982\u6a21\u578b\u6587\u4ef6\u3001\u56fe\u7247\u7b49\uff09 mlflow.log_artifact(\"path\/to\/local\/file\", \"artifact_subfolder\")<\/code><\/li>\n<\/ol>\n\n\n\n<p><strong>\u30ec\u30b3\u30fc\u30c9\u30d1\u30e9\u30e1\u30fc\u30bf:<\/strong><\/p>\n\n\n\n<p>\u3053\u306e\u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u8a18\u9332\u3057\u307e\u3059\u3002 \u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u540d\u3001\u306f\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u5024\u3067\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u5b66\u7fd2\u7387\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3092\u8a18\u9332\u3057\u307e\u3059\u3002<code>mlflow.log_param(key, value)<\/code><code>key<\/code><code>value<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>learning_rate = 0.01\nmlflow.log_param(\"learning_rate\", learning_rate)\n<\/code><\/pre>\n\n\n\n<p><strong>\u30e1\u30c8\u30ea\u30c3\u30af\u306e\u8a18\u9332:<\/strong><\/p>\n\n\n\n<p>\u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066\u30e1\u30c8\u30ea\u30c3\u30af\u3092\u8a18\u9332\u3057\u307e\u3059\u3002 \u306f\u6a19\u8b58\u540d\u3001 \u306f\u6a19\u8b58\u5024\u3001 \u306f\u30aa\u30d7\u30b7\u30e7\u30f3\u30fb\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001 \u306f\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u307e\u305f\u306f\u6700\u9069\u5316\u30d7\u30ed\u30bb\u30b9\u306e\u30b9\u30c6\u30c3\u30d7\u3092\u793a\u3057\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u4e2d\u306b\u5404\u30a8\u30dd\u30c3\u30af\u306e\u640d\u5931\u3092\u8a18\u9332\u3057\u307e\u3059\u3002<code>mlflow.log_metric(key, value, step)<\/code><code>key<\/code><code>value<\/code><code>step<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>for epoch in range(epochs):\n    # \u8bad\u7ec3\u6a21\u578b\u5e76\u8ba1\u7b97\u635f\u5931\n    loss = train_model()\n\n    # \u8bb0\u5f55\u635f\u5931\n    mlflow.log_metric(\"loss\", loss, step=epoch)\n<\/code><\/pre>\n\n\n\n<p><strong>\u30ef\u30fc\u30af\u30d4\u30fc\u30b9\u306e\u6587\u66f8\u5316:<\/strong><\/p>\n\n\n\n<p>\u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066\u6210\u679c\u7269\u3092\u8a18\u9332\u3057\u307e\u3059\u3002 \u306f\u30ed\u30fc\u30ab\u30eb \u30d5\u30a1\u30a4\u30eb\u307e\u305f\u306f\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3078\u306e\u30d1\u30b9\u3067\u3042\u308a\u3001 \u306f MLflow \u30b5\u30fc\u30d0\u30fc\u4e0a\u306e\u6210\u679c\u7269\u306e\u76f8\u5bfe\u30d1\u30b9\u3092\u8868\u3059\u7701\u7565\u53ef\u80fd\u306a\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3067\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u751f\u6210\u3055\u308c\u305f\u753b\u50cf\u30d5\u30a1\u30a4\u30eb\u3092\u8a18\u9332\u3057\u307e\u3059\u3002<code>mlflow.log_artifact(local_path, artifact_path)<\/code><code>local_path<\/code><code>artifact_path<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import matplotlib.pyplot as plt\n\n# \u521b\u5efa\u4e00\u5f20\u56fe\u7247\nplt.plot([0, 1, 2, 3, 4], [0, 1, 4, 9, 16])\nplt.savefig(\"plot.png\")\n\n# \u8bb0\u5f55\u56fe\u7247\u6587\u4ef6\nmlflow.log_artifact(\"plot.png\", \"plots\")\n<\/code><\/pre>\n\n\n\n<p>\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8\u3092\u8a18\u9332\u3059\u308b\u3053\u3068\u3067\u3001\u5b9f\u9a13\u30c7\u30fc\u30bf\u3092\u7c21\u5358\u306b\u8ffd\u8de1\u304a\u3088\u3073\u7ba1\u7406\u3067\u304d\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3068\u4fdd\u5b88\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002 MLflow \u306b\u306f\u8c4a\u5bcc\u306a API \u3068\u30d3\u30b8\u30e5\u30a2\u30eb \u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u304c\u7528\u610f\u3055\u308c\u3066\u304a\u308a\u3001\u3088\u308a\u52b9\u7387\u7684\u306a\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u306b\u6700\u9069\u306a\u30e2\u30c7\u30eb\u69cb\u6210\u3092\u3059\u3070\u3084\u304f\u898b\u3064\u3051\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f5c\u6210\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f5c\u6210\u306b\u306f\u3001\u4e3b\u306b\u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30d5\u30a1\u30a4\u30eb\u306e\u6e96\u5099\u3001\u74b0\u5883\u69cb\u6210\u306e\u5b9a\u7fa9\u3001\u304a\u3088\u3073\u5b9f\u884c\u30b2\u30fc\u30c8\u306e\u8a18\u8ff0\u304c\u542b\u307e\u308c\u307e\u3059\u3002 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f5c\u6210\u3059\u308b\u65b9\u6cd5\u306e\u8a73\u7d30\u306a\u8aac\u660e\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1. \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u6587\u66f8\u306e\u6e96\u5099<\/strong><\/p>\n\n\n\n<p>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3067\u3001\u6b21\u306e\u30d5\u30a1\u30a4\u30eb\u3068\u30d5\u30a9\u30eb\u30c0\u30fc\u3092\u542b\u3080\u69cb\u9020\u3092\u4f5c\u6210\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">luaCopy code<code>your_project\/\n|-- MLproject\n|-- conda.yaml\n|-- main.py\n|-- other_files\/\n<\/code><\/pre>\n\n\n\n<ul class=\"wp-block-list\">\n<li><code>MLproject<\/code>: \u9805\u76ee\u540d\u3001\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u306a\u3069\u3001\u9805\u76ee\u306e\u30e1\u30bf\u30c7\u30fc\u30bf\u3092\u8a18\u8ff0\u3059\u308b YAML \u30d5\u30a1\u30a4\u30eb\u3002<\/li>\n\n\n\n<li><code>conda.yaml<\/code>: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u3068\u74b0\u5883\u69cb\u6210\u3092\u8a18\u8ff0\u3059\u308b Conda YAML \u30d5\u30a1\u30a4\u30eb\u3002<\/li>\n\n\n\n<li><code>main.py<\/code>: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30e1\u30a4\u30f3 \u30ed\u30b8\u30c3\u30af\u3068 MLflow API \u547c\u3073\u51fa\u3057\u3092\u542b\u3080\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30e1\u30a4\u30f3\u5b9f\u884c\u30a8\u30f3\u30c8\u30ea\u3002<\/li>\n\n\n\n<li><code>other_files\/<\/code>: \u305d\u306e\u4ed6\u306e\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u95a2\u9023\u30d5\u30a1\u30a4\u30eb (\u30c7\u30fc\u30bf \u30d5\u30a1\u30a4\u30eb\u3001\u30e2\u30c7\u30eb \u30d5\u30a1\u30a4\u30eb\u306a\u3069)\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>2. ML\u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30d5\u30a1\u30a4\u30eb\u3092\u5b9a\u7fa9\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u30d5\u30a1\u30a4\u30eb\u3067\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u540d\u3001\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u306a\u3069\u306e\u30e1\u30bf\u30c7\u30fc\u30bf\u3092\u5b9a\u7fa9\u3057\u307e\u3059\u3002 \u4f8b\u3048\u3070\uff1a<code>MLproject<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">yamlCopy code<code>name: your_project_name\n\nentry_points:\n  main:\n    parameters:\n      param1: {type: float, default: 0.1, description: \"Example parameter 1\"}\n      param2: {type: int, default: 10, description: \"Example parameter 2\"}\n    command: \"python main.py --param1 {param1} --param2 {param2}\"\n<\/code><\/pre>\n\n\n\n<ul class=\"wp-block-list\">\n<li><code>name<\/code>: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u540d\u3002<\/li>\n\n\n\n<li><code>entry_points<\/code>: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u3092\u5b9a\u7fa9\u3057\u307e\u3059\u3002 \u3053\u306e\u5834\u5408\u3001\u3068\u3044\u3046\u540d\u524d\u306e\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u304c\u3042\u308a\u307e\u3059\u3002<code>main<\/code><\/li>\n\n\n\n<li><code>parameters<\/code>: \u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u306b\u5fc5\u8981\u306a\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3068\u305d\u306e\u578b\u3001\u65e2\u5b9a\u5024\u3001\u304a\u3088\u3073\u8aac\u660e\u3092\u5b9a\u7fa9\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><code>command<\/code>: \u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u3092\u5b9f\u884c\u3059\u308b\u30b3\u30de\u30f3\u30c9\u3092\u5b9a\u7fa9\u3057\u307e\u3059\u3002 \u3053\u306e\u5834\u5408\u3001\u30b3\u30de\u30f3\u30c9\u306f and \u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3092\u6e21\u3057\u3066\u30d5\u30a1\u30a4\u30eb\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<code>main.py<\/code><code>param1<\/code><code>param2<\/code><\/li>\n<\/ul>\n\n\n\n<p><strong>3. conda.yaml \u30d5\u30a1\u30a4\u30eb\u3092\u5b9a\u7fa9\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u30d5\u30a1\u30a4\u30eb\u3067\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u3068\u74b0\u5883\u69cb\u6210\u3092\u5b9a\u7fa9\u3057\u307e\u3059\u3002 \u4f8b\u3048\u3070\uff1a<code>conda.yaml<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">yamlCopy code<code>name: your_project_env\nchannels:\n  - defaults\n  - conda-forge\ndependencies:\n  - python=3.8\n  - scikit-learn\n  - pandas\n  - mlflow\n  - pip:\n    - -r requirements.txt\n<\/code><\/pre>\n\n\n\n<ul class=\"wp-block-list\">\n<li><code>name<\/code>: \u74b0\u5883\u540d\u3002<\/li>\n\n\n\n<li><code>channels<\/code>: \u30d1\u30c3\u30b1\u30fc\u30b8\u306e\u691c\u7d22\u306b\u4f7f\u7528\u3055\u308c\u308b Conda \u30c1\u30e3\u30f3\u30cd\u30eb \u30ea\u30b9\u30c8\u3002<\/li>\n\n\n\n<li><code>dependencies<\/code>: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u5fc5\u8981\u306a\u30d1\u30c3\u30b1\u30fc\u30b8\u3068\u30d0\u30fc\u30b8\u30e7\u30f3\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>4.main.py \u30d5\u30a1\u30a4\u30eb\u3092\u66f8\u304d\u8fbc\u307f\u307e\u3059<\/strong><\/p>\n\n\n\n<p>\u30d5\u30a1\u30a4\u30eb\u306b\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30e1\u30a4\u30f3 \u30ed\u30b8\u30c3\u30af\u3068 MLflow API \u547c\u3073\u51fa\u3057\u3092\u8a18\u8ff0\u3057\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u3092\u4f7f\u7528\u3057\u3066\u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3\u5f15\u6570\u3092\u51e6\u7406\u3057\u3001MLflow API \u3092\u4f7f\u7528\u3057\u3066\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u6210\u679c\u7269\u306a\u3069\u3092\u30ed\u30b0\u306b\u8a18\u9332\u3067\u304d\u307e\u3059\u3002<code>main.py<\/code><code>argparse<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import argparse\nimport mlflow\n\ndef main(param1, param2):\n    # \u9879\u76ee\u903b\u8f91\uff0c\u5982\u8bad\u7ec3\u6a21\u578b\u3001\u8ba1\u7b97\u6307\u6807\u7b49\n    ...\n    with mlflow.start_run():\n        mlflow.log_param(\"param1\", param1)\n        mlflow.log_param(\"param2\", param2)\n        mlflow.log_metric(\"example_metric\", example_metric_value)\n        ...\n        mlflow.log_artifact(\"path\/to\/local\/file\", \"artifact_subfolder\")\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--\n<\/code><\/pre>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>param1&#8243;, type=float, help=&#8221;\u30d1\u30e9\u30e1\u30fc\u30bf 1 \u306e\u4f8b&#8221;) parser.add_argument(&#8220;\u2013param2&#8243;, type=int, help=&#8221;\u30d1\u30e9\u30e1\u30fc\u30bf 2 \u306e\u4f8b&#8221;) args = parser.parse_args()<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">cssCopy code<code>main(args.param1, args.param2)\n<\/code><\/pre>\n\n\n\n<pre class=\"wp-block-preformatted\">markdownCopy code<code>\n\u5728 `main.py` \u6587\u4ef6\u4e2d\uff1a\n\n- \u4f7f\u7528 `argparse.ArgumentParser()` \u5b9a\u4e49\u547d\u4ee4\u884c\u53c2\u6570\u3002\n- \u7f16\u5199 `main()` \u51fd\u6570\uff0c\u5305\u542b\u9879\u76ee\u7684\u4e3b\u8981\u903b\u8f91\u548c MLflow API \u8c03\u7528\u3002\n- \u5728 `if __name__ == \"__main__\":` \u4ee3\u7801\u5757\u4e2d\uff0c\u89e3\u6790\u547d\u4ee4\u884c\u53c2\u6570\uff0c\u5e76\u8c03\u7528 `main()` \u51fd\u6570\u3002\n\n**5. \u8fd0\u884c MLflow \u9879\u76ee**\n\n\u5728\u5b8c\u6210\u9879\u76ee\u6587\u4ef6\u51c6\u5907\u540e\uff0c\u60a8\u53ef\u4ee5\u4f7f\u7528 `mlflow run` \u547d\u4ee4\u8fd0\u884c MLflow \u9879\u76ee\u3002\u5728\u9879\u76ee\u76ee\u5f55\u6240\u5728\u7684\u7ec8\u7aef\u4e2d\uff0c\u6267\u884c\u4ee5\u4e0b\u547d\u4ee4\uff1a\n\n<\/code><\/pre>\n\n\n\n<p>MLFLOW \u3092\u5b9f\u884c\u3057\u307e\u3059\u3002-P \u30d1\u30e9\u30e1\u30fc\u30bf 1=0.5 -P \u30d1\u30e9\u30e1\u30fc\u30bf 2=20<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">markdownCopy code<code>\n- `mlflow run`: MLflow \u547d\u4ee4\uff0c\u7528\u4e8e\u8fd0\u884c\u9879\u76ee\u3002\n- `.`: \u9879\u76ee\u76ee\u5f55\u7684\u8def\u5f84\u3002\u5728\u672c\u4f8b\u4e2d\uff0c\u9879\u76ee\u4f4d\u4e8e\u5f53\u524d\u76ee\u5f55\u3002\n- `-P param1=0.5 -P param2=20`: \u4f20\u9012\u7ed9\u9879\u76ee\u5165\u53e3\u70b9\u7684\u53c2\u6570\u3002\u5728\u672c\u4f8b\u4e2d\uff0c\u4f20\u9012 `param1` \u548c `param2` \u53c2\u6570\u3002\n\n\u6b64\u547d\u4ee4\u5c06\u8fd0\u884c MLflow \u9879\u76ee\uff0c\u5e76\u4f7f\u7528\u6307\u5b9a\u7684\u53c2\u6570\u3002\u5728\u8fd0\u884c\u8fc7\u7a0b\u4e2d\uff0cMLflow \u5c06\u81ea\u52a8\u521b\u5efa\u548c\u6fc0\u6d3b Conda \u73af\u5883\uff0c\u5e76\u6267\u884c\u5b9a\u4e49\u5728 `MLproject` \u6587\u4ef6\u4e2d\u7684\u547d\u4ee4\u3002\n\n\u901a\u8fc7\u521b\u5efa\u4e00\u4e2a MLflow \u9879\u76ee\uff0c\u60a8\u53ef\u4ee5\u66f4\u597d\u5730\u7ec4\u7ec7\u548c\u7ba1\u7406\u673a\u5668\u5b66\u4e60\u5de5\u4f5c\u6d41\uff0c\u5b9e\u73b0\u8de8\u56e2\u961f\u7684\u534f\u4f5c\u548c\u5171\u4eab\u3002MLflow \u63d0\u4f9b\u4e86\u4e00\u79cd\u7b80\u5355\u7684\u65b9\u5f0f\u6765\u6253\u5305\u3001\u8fd0\u884c\u548c\u5206\u4eab\u673a\u5668\u5b66\u4e60\u9879\u76ee\uff0c\u4f7f\u5f97\u60a8\u53ef\u4ee5\u8f7b\u677e\u5730\u5728\u4e0d\u540c\u73af\u5883\u4e2d\u91cd\u590d\u5b9e\u9a8c\u548c\u90e8\u7f72\u6a21\u578b\u3002\n<\/code><\/pre>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p><strong>6.\u5171\u6709\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3<\/strong><\/p>\n\n\n\n<p>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u6e96\u5099\u304c\u3067\u304d\u305f\u3089\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3068\u5171\u6709\u3057\u3066\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u3092\u5b9f\u73fe\u3059\u308b\u306b\u306f\u3001\u6b21\u306e\u30a2\u30d7\u30ed\u30fc\u30c1\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u30d0\u30fc\u30b8\u30e7\u30f3<\/strong>\u7ba1\u7406: Git \u306a\u3069\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u30b7\u30b9\u30c6\u30e0\u3067\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u30db\u30b9\u30c8\u3057\u3066\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u5c65\u6b74\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u8ffd\u8de1\u3057\u307e\u3059\u3002 \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u306f\u3001\u30ea\u30dd\u30b8\u30c8\u30ea\u3092\u8907\u88fd\u3057\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u30ed\u30fc\u30ab\u30eb\u3067\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30ea\u30e2\u30fc\u30c8 MLflow \u30b5\u30fc\u30d0\u30fc\u3092\u69cb\u6210\u3059\u308b: \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u5b9f\u9a13\u7d50\u679c\u3092\u8868\u793a\u304a\u3088\u3073\u5171\u6709\u3067\u304d\u308b\u3088\u3046\u306b<strong>\u3001\u30ea\u30e2\u30fc\u30c8 MLflow \u30b5\u30fc\u30d0\u30fc\u3092\u69cb\u6210\u3057\u307e\u3059<\/strong>\u3002 Python \u30b9\u30af\u30ea\u30d7\u30c8\u3067\u306f\u3001\u3053\u306e\u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066\u30ea\u30e2\u30fc\u30c8 \u30b5\u30fc\u30d0\u30fc\u306e URI \u3092\u8a2d\u5b9a\u3067\u304d\u307e\u3059\u3002<code>mlflow.set_tracking_uri()<\/code><\/li>\n\n\n\n<li>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b:&nbsp;<strong>MLflow \u30e2\u30c7\u30eb<\/strong>&nbsp;\u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u7ba1\u7406\u304a\u3088\u3073\u5171\u6709\u3057\u307e\u3059\u3002 \u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6e08\u307f\u306e\u30e2\u30c7\u30eb\u3092\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306b\u767b\u9332\u3057\u3001\u305d\u306e\u30e2\u30c7\u30eb\u306e\u7570\u306a\u308b\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u4f5c\u6210\u3067\u304d\u307e\u3059\u3002 \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u306f\u3001MLflow UI \u3067\u30e2\u30c7\u30eb\u306e\u3055\u307e\u3056\u307e\u306a\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u8868\u793a\u3001\u6bd4\u8f03\u3001\u30c7\u30d7\u30ed\u30a4\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ul>\n\n\n\n<p>\u3053\u308c\u3089\u306e\u30a2\u30d7\u30ed\u30fc\u30c1\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3068\u5354\u529b\u3057\u3066\u5b9f\u9a13\u30c7\u30fc\u30bf\u3001\u30e2\u30c7\u30eb\u3001\u7d50\u679c\u3092\u5171\u6709\u3067\u304d\u308b\u305f\u3081\u3001\u30c1\u30fc\u30e0\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u304c\u6539\u5584\u3055\u308c\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u958b\u767a\u3068\u30c7\u30d7\u30ed\u30a4\u304c\u52a0\u901f\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f5c\u6210\u3059\u308b\u3068\u3001\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u3092\u3088\u308a\u9069\u5207\u306b\u6574\u7406\u304a\u3088\u3073\u7ba1\u7406\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30d5\u30a1\u30a4\u30eb\u3001\u74b0\u5883\u69cb\u6210\u3001\u5b9f\u884c\u30b2\u30fc\u30c8\u3092\u5b9a\u7fa9\u3059\u308b\u3053\u3068\u3067\u3001\u5b9f\u9a13\u3092\u7c21\u5358\u306b\u7e70\u308a\u8fd4\u3057\u3001\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u306b\u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3067\u304d\u307e\u3059\u3002 \u3055\u3089\u306b\u3001MLflow \u306b\u306f\u3001\u30c1\u30fc\u30e0\u9593\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u5171\u6709\u3092\u30b5\u30dd\u30fc\u30c8\u3059\u308b\u8c4a\u5bcc\u306a\u6a5f\u80fd\u3068\u30c4\u30fc\u30eb\u304c\u7528\u610f\u3055\u308c\u3066\u304a\u308a\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u958b\u767a\u52b9\u7387\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u7ba1\u7406\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u4f9d\u5b58\u95a2\u4fc2\u306e\u7ba1\u7406\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u3068\u30c7\u30fc\u30bf \u30b5\u30a4\u30a8\u30f3\u30b9\u306e\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u975e\u5e38\u306b\u91cd\u8981\u3067\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u304c\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u3067\u5b89\u5b9a\u3057\u3066\u5b9f\u884c\u3055\u308c\u308b\u3068\u540c\u6642\u306b\u3001\u30c1\u30fc\u30e0\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4\u304c\u7c21\u7d20\u5316\u3055\u308c\u307e\u3059\u3002 \u6b21\u306b\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u7ba1\u7406\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1.\u4eee\u60f3\u74b0\u5883\u3092\u4f7f\u7528\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u5206\u96e2\u3068\u79fb\u690d\u6027\u3092\u78ba\u4fdd\u3059\u308b\u305f\u3081\u306b\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3054\u3068\u306b\u4eee\u60f3\u74b0\u5883\u3092\u4f5c\u6210\u3059\u308b\u3053\u3068\u3092\u304a\u52e7\u3081\u3057\u307e\u3059\u3002 \u4eee\u60f3\u74b0\u5883\u306f\u3001\u7570\u306a\u308b\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u9593\u306e\u4f9d\u5b58\u95a2\u4fc2\u306e\u7af6\u5408\u3092\u9632\u3050\u3001\u72ec\u7acb\u3057\u305f Python \u30a4\u30f3\u30bf\u30fc\u30d7\u30ea\u30bf\u30fc\u3068\u30d1\u30c3\u30b1\u30fc\u30b8\u306e\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u5834\u6240\u3067\u3059\u3002 \u4e00\u822c\u7684\u306a\u4eee\u60f3\u74b0\u5883\u30c4\u30fc\u30eb\u306f\u3001(Python\u72ec\u81ea\u306e\u4eee\u60f3\u74b0\u5883\u30c4\u30fc\u30eb)\u3068(Anaconda\u306e\u4eee\u60f3\u74b0\u5883\u304a\u3088\u3073\u30d1\u30c3\u30b1\u30fc\u30b8\u7ba1\u7406\u30c4\u30fc\u30eb)\u3067\u3059\u3002<code>venv<\/code><code>conda<\/code><\/p>\n\n\n\n<p><strong>2.\u4f9d\u5b58\u30d5\u30a1\u30a4\u30eb\u3092\u4f7f\u7528\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u3092\u7c21\u5358\u306b\u7ba1\u7406\u304a\u3088\u3073\u5171\u6709\u3059\u308b\u306b\u306f\u3001\u4f9d\u5b58\u95a2\u4fc2\u30d5\u30a1\u30a4\u30eb\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u30d5\u30a1\u30a4\u30eb\u306b\u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u5fc5\u8981\u306a\u3059\u3079\u3066\u306e\u30d1\u30c3\u30b1\u30fc\u30b8\u3068\u305d\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u60c5\u5831\u304c\u542b\u307e\u308c\u3066\u3044\u307e\u3059\u3002 \u4e00\u822c\u7684\u306a\u4f9d\u5b58\u30d5\u30a1\u30a4\u30eb\u306f\u6b21\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><code>requirements.txt<\/code>: \u306e\u4f9d\u5b58\u30d5\u30a1\u30a4\u30eb \u30d3\u30eb\u30c9\u3092\u4f7f\u7528\u3059\u308b\u3053\u3068\u3082\u3001\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3092\u4f7f\u7528\u3059\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002<code>pip<\/code><code>pip freeze &gt; requirements.txt<\/code><code>pip install -r requirements.txt<\/code><\/li>\n\n\n\n<li><code>conda.yaml<\/code>&nbsp;\u307e\u305f\u306f : \u306e\u4f9d\u5b58\u30d5\u30a1\u30a4\u30eb \u30d3\u30eb\u30c9\u3092\u4f7f\u7528\u3059\u308b\u3053\u3068\u3082\u3001\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3092\u4f7f\u7528\u3059\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002<code>environment.yaml<\/code><code>conda<\/code><code>conda env export &gt; environment.yaml<\/code><code>conda env create -f environment.yaml<\/code><\/li>\n<\/ul>\n\n\n\n<p>\u4f9d\u5b58\u95a2\u4fc2\u30d5\u30a1\u30a4\u30eb\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u540c\u3058\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u30d1\u30c3\u30b1\u30fc\u30b8\u3092\u7570\u306a\u308b\u74b0\u5883\u3067\u4f7f\u7528\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u3001\u6f5c\u5728\u7684\u306a\u4e92\u63db\u6027\u306e\u554f\u984c\u3092\u56de\u907f\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>3. MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u3092\u7ba1\u7406\u3059\u308b<\/strong><\/p>\n\n\n\n<p>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306f\u3001\u4f9d\u5b58\u95a2\u4fc2\u306e\u7ba1\u7406\u306f\u30d5\u30a1\u30a4\u30eb\u3092\u901a\u3058\u3066\u5b9f\u88c5\u3055\u308c\u307e\u3059\u3002 \u3053\u306e\u30d5\u30a1\u30a4\u30eb\u306b\u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u5fc5\u8981\u306a\u30d1\u30c3\u30b1\u30fc\u30b8\u3068\u305d\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u60c5\u5831\u304c\u8a18\u8ff0\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3059\u308b\u3068\u3001\u30d5\u30a1\u30a4\u30eb\u306b\u5f93\u3063\u3066\u5fc5\u8981\u306a\u4f9d\u5b58\u95a2\u4fc2\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b Conda \u74b0\u5883\u304c\u81ea\u52d5\u7684\u306b\u4f5c\u6210\u3055\u308c\u3001\u30a2\u30af\u30c6\u30a3\u30d6\u5316\u3055\u308c\u307e\u3059\u3002<code>conda.yaml<\/code><code>mlflow run<\/code><code>conda.yaml<\/code><\/p>\n\n\n\n<p>\u305f\u3068\u3048\u3070\u3001\u30d5\u30a1\u30a4\u30eb\u3067\u6b21\u306e\u4f9d\u5b58\u95a2\u4fc2\u3092\u5b9a\u7fa9\u3057\u307e\u3059\u3002<code>conda.yaml<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">yamlCopy code<code>name: your_project_env\nchannels:\n  - defaults\n  - conda-forge\ndependencies:\n  - python=3.8\n  - scikit-learn\n  - pandas\n  - mlflow\n  - pip:\n    - -r requirements.txt\n<\/code><\/pre>\n\n\n\n<p>\u3053\u306e\u3088\u3046\u306b\u3057\u3066\u3001MLflow \u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u5b9f\u884c\u306b\u5fc5\u8981\u306a\u4f9d\u5b58\u95a2\u4fc2\u304c\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u9593\u3067\u4e00\u8cab\u3057\u3066\u3044\u308b\u3053\u3068\u3092\u4fdd\u8a3c\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001\u6a5f\u68b0\u5b66\u7fd2\u304a\u3088\u3073\u30c7\u30fc\u30bf \u30b5\u30a4\u30a8\u30f3\u30b9 \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306f\u3001\u4f9d\u5b58\u95a2\u4fc2\u7ba1\u7406\u306b\u3088\u308a\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u304c\u3055\u307e\u3056\u307e\u306a\u74b0\u5883\u3067\u5b89\u5b9a\u3057\u3066\u5b9f\u884c\u3055\u308c\u3001\u30c1\u30fc\u30e0\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4\u304c\u7c21\u7d20\u5316\u3055\u308c\u307e\u3059\u3002 \u4eee\u60f3\u74b0\u5883\u3001\u4f9d\u5b58\u95a2\u4fc2\u30d5\u30a1\u30a4\u30eb\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f7f\u7528\u3059\u308b\u3053\u3068\u3067\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u3092\u52b9\u7387\u7684\u306b\u7ba1\u7406\u3057\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4fdd\u5b88\u6027\u3068\u79fb\u690d\u6027\u3092\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u3092\u4f7f\u7528\u3057\u3066\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3059\u308b\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u3092\u4f7f\u7528\u3057\u305f\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u5b9f\u884c\u306f\u7c21\u5358\u3067\u3059\u3002 \u30d5\u30a1\u30a4\u30eb\u3068\u95a2\u9023\u3059\u308b\u4f9d\u5b58\u30d5\u30a1\u30a4\u30eb\u3092\u542b\u3080 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f5c\u6210\u3057\u305f\u3089\u3001\u3053\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002 MLflow \u3092\u4f7f\u7528\u3057\u3066\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3059\u308b\u8a73\u7d30\u306a\u624b\u9806\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<code>MLproject<\/code><code>mlflow run<\/code><\/p>\n\n\n\n<p><strong>1. \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u306b\u79fb\u52d5\u3057\u307e\u3059\u3002<\/strong><\/p>\n\n\n\n<p>\u30bf\u30fc\u30df\u30ca\u30eb\u3067\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30eb\u30fc\u30c8\u306b\u79fb\u52d5\u3057\u307e\u3059\u3002 \u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u306b\u30d5\u30a1\u30a4\u30eb (\u307e\u305f\u306f\u305d\u306e\u4ed6\u306e\u4f9d\u5b58\u30d5\u30a1\u30a4\u30eb) \u304c\u542b\u307e\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002<code>MLproject<\/code><code>conda.yaml<\/code><\/p>\n\n\n\n<p><strong>2. MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3068\u540c\u3058\u30bf\u30fc\u30df\u30ca\u30eb\u3067\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">cssCopy code<code>mlflow run . -P param1=value1 -P param2=value2\n<\/code><\/pre>\n\n\n\n<ul class=\"wp-block-list\">\n<li><code>mlflow run<\/code>: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u5b9f\u884c\u306b\u4f7f\u7528\u3055\u308c\u308b MLflow \u30b3\u30de\u30f3\u30c9\u3002<\/li>\n\n\n\n<li><code>.<\/code>: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3078\u306e\u30d1\u30b9\u3002 \u3053\u306e\u4f8b\u3067\u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306f\u73fe\u5728\u306e\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u306b\u3042\u308a\u307e\u3059\u3002<\/li>\n\n\n\n<li><code>-P param1=value1 -P param2=value2<\/code>: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u306b\u6e21\u3055\u308c\u308b\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3002 \u3053\u306e\u5834\u5408\u3001\u304a\u3088\u3073\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u304c\u6e21\u3055\u308c\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u306f\u3001\u30d5\u30a1\u30a4\u30eb\u3067\u5b9a\u7fa9\u3055\u308c\u3066\u3044\u308b\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u306b\u5bfe\u5fdc\u3057\u3066\u3044\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002<code>param1<\/code><code>param2<\/code><code>MLproject<\/code><\/li>\n<\/ul>\n\n\n\n<p>\u3053\u306e\u30b3\u30de\u30f3\u30c9\u306f\u3001\u6307\u5b9a\u3055\u308c\u305f\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3092\u4f7f\u7528\u3057\u3066 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002 \u5b9f\u884c\u4e2d\u3001MLflow \u306f\u30d5\u30a1\u30a4\u30eb\u304b\u3089 Conda \u74b0\u5883\u3092\u81ea\u52d5\u7684\u306b\u4f5c\u6210\u3057\u3066\u30a2\u30af\u30c6\u30a3\u30d6\u5316\u3057\u3001\u30d5\u30a1\u30a4\u30eb\u3067\u5b9a\u7fa9\u3055\u308c\u3066\u3044\u308b\u30b3\u30de\u30f3\u30c9\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<code>conda.yaml<\/code><code>MLproject<\/code><\/p>\n\n\n\n<p><strong>3. \u30ea\u30e2\u30fc\u30c8\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u30ed\u30fc\u30ab\u30eb \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u5b9f\u884c\u306b\u52a0\u3048\u3066\u3001MLflow \u3092\u4f7f\u7528\u3057\u3066\u30ea\u30e2\u30fc\u30c8 \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3059\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u304c Git \u30ea\u30dd\u30b8\u30c8\u30ea\u3067\u30db\u30b9\u30c8\u3055\u308c\u3066\u3044\u308b\u5834\u5408\u306f\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">arduinoCopy code<code>mlflow run https:\/\/github.com\/your-username\/your-repo.git -P param1=value1 -P param2=value2\n<\/code><\/pre>\n\n\n\n<ul class=\"wp-block-list\">\n<li><code>https:\/\/github.com\/your-username\/your-repo.git<\/code>: Git \u30ea\u30dd\u30b8\u30c8\u30ea\u306e URL\u3002<\/li>\n\n\n\n<li><code>-P param1=value1 -P param2=value2<\/code>: \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u306b\u6e21\u3055\u308c\u308b\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3002<\/li>\n<\/ul>\n\n\n\n<p>\u30ea\u30e2\u30fc\u30c8 \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3059\u308b\u3068\u3001MLflow \u306f\u81ea\u52d5\u7684\u306b Git \u30ea\u30dd\u30b8\u30c8\u30ea\u3092\u30ed\u30fc\u30ab\u30eb\u306b\u8907\u88fd\u3057\u3001\u540c\u3058\u5b9f\u884c\u30d7\u30ed\u30bb\u30b9\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u3053\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u30ed\u30fc\u30ab\u30eb\u307e\u305f\u306f\u30ea\u30e2\u30fc\u30c8\u3067\u7c21\u5358\u306b\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u30a2\u30d7\u30ed\u30fc\u30c1\u306b\u3088\u308a\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u306e\u4e00\u8cab\u6027\u3068\u79fb\u690d\u6027\u304c\u78ba\u4fdd\u3055\u308c\u3001\u5b9f\u9a13\u306e\u53cd\u5fa9\u3068\u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4\u304c\u7c21\u7d20\u5316\u3055\u308c\u307e\u3059\u3002<code>mlflow run<\/code><\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u30e2\u30c7\u30eb\u306e\u4f5c\u6210\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u3067\u306f\u3001\u30e2\u30c7\u30eb\u306f\u4e88\u6e2c\u7528\u306e\u30c7\u30d7\u30ed\u30a4\u53ef\u80fd\u306a\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3067\u3059\u3002 MLflow \u30e2\u30c7\u30eb\u306f\u3001\u30e2\u30c7\u30eb \u30d5\u30a1\u30a4\u30eb\u3001\u30e2\u30c7\u30eb \u30e1\u30bf\u30c7\u30fc\u30bf\u3001\u304a\u3088\u3073\u30e2\u30c7\u30eb\u5f62\u5f0f\u3067\u69cb\u6210\u3055\u308c\u307e\u3059\u3002 MLflow \u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4f5c\u6210\u3059\u308b\u8a73\u7d30\u306a\u624b\u9806\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1. \u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u307e\u305a\u3001\u6a5f\u68b0\u5b66\u7fd2\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u3053\u306e\u30d7\u30ed\u30bb\u30b9\u306f\u3001\u4f7f\u7528\u3057\u3066\u3044\u308b\u7279\u5b9a\u306e\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3068\u30e9\u30a4\u30d6\u30e9\u30ea\u306b\u3088\u3063\u3066\u7570\u306a\u308a\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001scikit-learn \u3092\u4f7f\u7528\u3057\u3066\u7dda\u5f62\u56de\u5e30\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import numpy as np\nfrom sklearn.linear_model import LinearRegression\nfrom sklearn.model_selection import train_test_split\n\n# \u521b\u5efa\u6570\u636e\u96c6\nX, y = np.random.rand(100, 1), np.random.rand(100, 1)\nX_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)\n\n# \u8bad\u7ec3\u6a21\u578b\nmodel = LinearRegression()\nmodel.fit(X_train, y_train)\n<\/code><\/pre>\n\n\n\n<p><strong>2. MLflow \u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u30e2\u30c7\u30eb\u306e\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u304c\u5b8c\u4e86\u3057\u305f\u3089\u3001MLflow API \u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002 MLflow \u3067\u306f\u3001scikit-learn\u3001PyTorch\u3001TensorFlow \u306a\u3069\u306e\u8907\u6570\u306e\u30e2\u30c7\u30eb\u5f62\u5f0f\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u3053\u306e\u4f8b\u3067\u306f\u3001scikit-learn \u5f62\u5f0f\u3092\u4f7f\u7528\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow.sklearn\n\n# \u4fdd\u5b58\u6a21\u578b\nmlflow.sklearn.save_model(model, \"model\")\n<\/code><\/pre>\n\n\n\n<p>\u3053\u306e\u4f8b\u3067\u306f\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6e08\u307f\u306e scikit-learn \u30e2\u30c7\u30eb\u306f\u3001\u30e2\u30c7\u30eb\u304c\u4fdd\u5b58\u3055\u308c\u308b\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3067\u3059\u3002 MLflow \u306f\u3001\u6307\u5b9a\u3055\u308c\u305f\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u306b\u3001\u30e2\u30c7\u30eb\u306e\u5f62\u5f0f\u3001\u30d0\u30fc\u30b8\u30e7\u30f3\u3001\u304a\u3088\u3073\u305d\u306e\u4ed6\u306e\u60c5\u5831\u3092\u8a18\u8ff0\u3059\u308b \u3068\u3044\u3046\u540d\u524d\u306e\u30e1\u30bf\u30c7\u30fc\u30bf \u30d5\u30a1\u30a4\u30eb\u3092\u4f5c\u6210\u3057\u307e\u3059\u3002 \u540c\u6642\u306b\u3001MLflow \u306f\u3001scikit-learn \u30e2\u30c7\u30eb\u306a\u3069\u306e\u30e2\u30c7\u30eb \u30d5\u30a1\u30a4\u30eb\u3092\u3053\u306e\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u306b\u4fdd\u5b58\u3057\u307e\u3059\u3002<code>model<\/code><code>\"model\"<\/code><code>MLmodel<\/code><code>.pkl<\/code><\/p>\n\n\n\n<p><strong>3. MLflow \u30e2\u30c7\u30eb\u3092\u8aad\u307f\u8fbc\u3093\u3067\u4f7f\u7528\u3059\u308b<\/strong><\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3057\u305f\u3089\u3001MLflow API \u3092\u4f7f\u7528\u3057\u3066\u3001\u4ed6\u306e Python \u30b9\u30af\u30ea\u30d7\u30c8\u307e\u305f\u306f\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u3067\u30e2\u30c7\u30eb\u3092\u8aad\u307f\u8fbc\u3093\u3067\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow.sklearn\n\n# \u52a0\u8f7d\u6a21\u578b\nloaded_model = mlflow.sklearn.load_model(\"model\")\n\n# \u4f7f\u7528\u6a21\u578b\u8fdb\u884c\u9884\u6d4b\npredictions = loaded_model.predict(X_test)\n<\/code><\/pre>\n\n\n\n<p>\u3053\u306e\u4f8b\u3067\u306f\u3001\u30e2\u30c7\u30eb\u304c\u4fdd\u5b58\u3055\u308c\u308b\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3067\u3059\u3002 \u95a2\u6570\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u6307\u5b9a\u3057\u305f\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u304b\u3089 MLflow \u30e2\u30c7\u30eb\u3092\u8aad\u307f\u8fbc\u3080\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u305d\u306e\u5f8c\u3001\u8aad\u307f\u8fbc\u307e\u308c\u305f\u30e2\u30c7\u30eb\u3092\u4f7f\u7528\u3057\u3066\u3001\u901a\u5e38\u306e scikit-learn \u30e2\u30c7\u30eb\u306e\u5834\u5408\u3068\u540c\u69d8\u306b\u4e88\u6e2c\u3092\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<code>\"model\"<\/code><code>mlflow.sklearn.load_model()<\/code><\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u3092\u4f5c\u6210\u3059\u308b\u30d7\u30ed\u30bb\u30b9\u306f\u7c21\u5358\u3067\u3059\u3002 MLflow API \u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u304a\u3088\u3073\u8aad\u307f\u8fbc\u3080\u3053\u3068\u3067\u3001\u30e2\u30c7\u30eb\u306e\u66f8\u5f0f\u8a2d\u5b9a\u3068\u30e1\u30bf\u30c7\u30fc\u30bf\u306e\u4e00\u8cab\u6027\u3092\u78ba\u4fdd\u3057\u3001\u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4\u3068\u5171\u6709\u3092\u7c21\u7565\u5316\u3067\u304d\u307e\u3059\u3002 \u3055\u3089\u306b\u3001MLflow \u306f\u8907\u6570\u306e\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u308b\u305f\u3081\u3001\u3055\u307e\u3056\u307e\u306a\u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d6\u30e9\u30ea\u3084\u74b0\u5883\u3067\u30e2\u30c7\u30eb\u3092\u7c21\u5358\u306b\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7570\u306a\u308b\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u3067\u306f\u3001scikit-learn\u3001TensorFlow\u3001PyTorch \u306a\u3069\u306e\u8907\u6570\u306e\u30e2\u30c7\u30eb\u5f62\u5f0f\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b\u3068\u304d\u306f\u3001\u30e2\u30c7\u30eb\u306b\u5bfe\u5fdc\u3059\u308b MLflow \u30b5\u30d6\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u6b21\u306b\u3001\u3055\u307e\u3056\u307e\u306a\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b\u4f8b\u3092\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1. \u30b7\u30ad\u30c3\u30c8\u5b66\u7fd2\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b<\/strong><\/p>\n\n\n\n<p>scikit-learn \u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u305f\u5834\u5408\u306f\u3001\u30b5\u30d6\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002<code>mlflow.sklearn<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow.sklearn\n\n# \u8bad\u7ec3\u4e00\u4e2a scikit-learn \u6a21\u578b\n# ...\n\n# \u4fdd\u5b58\u6a21\u578b\nmlflow.sklearn.save_model(model, \"model_sklearn\")\n<\/code><\/pre>\n\n\n\n<p><strong>2. \u30c6\u30f3\u30bd\u30eb\u30d5\u30ed\u30fc\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b<\/strong><\/p>\n\n\n\n<p>TensorFlow \u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u305f\u5834\u5408\u306f\u3001\u30b5\u30d6\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002<code>mlflow.tensorflow<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow.tensorflow\n\n# \u8bad\u7ec3\u4e00\u4e2a TensorFlow \u6a21\u578b\n# ...\n\n# \u4fdd\u5b58\u6a21\u578b\nmlflow.tensorflow.save_model(model, \"model_tensorflow\")\n<\/code><\/pre>\n\n\n\n<p>\u6ce8: \u4f7f\u7528\u3059\u308b\u524d\u306b\u3001TensorFlow \u30e9\u30a4\u30d6\u30e9\u30ea\u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u3066\u304f\u3060\u3055\u3044\u3002<code>mlflow.tensorflow<\/code><\/p>\n\n\n\n<p><strong>3. PyTorch \u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b<\/strong><\/p>\n\n\n\n<p>PyTorch \u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u305f\u5834\u5408\u306f\u3001\u30b5\u30d6\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002<code>mlflow.pytorch<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow.pytorch\n\n# \u8bad\u7ec3\u4e00\u4e2a PyTorch \u6a21\u578b\n# ...\n\n# \u4fdd\u5b58\u6a21\u578b\nmlflow.pytorch.save_model(model, \"model_pytorch\")\n<\/code><\/pre>\n\n\n\n<p>\u6ce8: \u4f7f\u7528\u3059\u308b\u524d\u306b\u3001PyTorch \u30e9\u30a4\u30d6\u30e9\u30ea\u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u3066\u304f\u3060\u3055\u3044\u3002<code>mlflow.pytorch<\/code><\/p>\n\n\n\n<p><strong>4.\u4ed6\u306e\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u4fdd\u5b58\u3057\u307e\u3059<\/strong><\/p>\n\n\n\n<p>MLflow \u306f\u3001XGBoost\u3001LightGBM \u306a\u3069\u3001\u4ed6\u306e\u591a\u304f\u306e\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u307e\u3059\u3002 \u3055\u307e\u3056\u307e\u306a\u30b5\u30d6\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u3053\u308c\u3089\u306e\u30e2\u30c7\u30eb\u5f62\u5f0f\u3092\u4fdd\u5b58\u3059\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u306f\u3001MLflow \u306e\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u3092\u53c2\u7167\u3057\u3066\u304f\u3060\u3055\u3044\u3002&nbsp;<a href=\"https:\/\/www.mlflow.org\/docs\/latest\/models.html\">MLflow \u30e2\u30c7\u30eb<\/a><\/p>\n\n\n\n<p>\u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b\u3068\u304d\u306f\u3001\u5fc5\u305a\u30e2\u30c7\u30eb\u306b\u5bfe\u5fdc\u3059\u308bMLflow\u30b5\u30d6\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u304f\u3060\u3055\u3044\u3002 \u3053\u308c\u306b\u3088\u308a\u3001MLflow \u306f\u30e2\u30c7\u30eb \u30d5\u30a1\u30a4\u30eb\u3001\u30e1\u30bf\u30c7\u30fc\u30bf\u3001\u304a\u3088\u3073\u5f62\u5f0f\u3092\u6b63\u3057\u304f\u4fdd\u5b58\u3067\u304d\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u307e\u305f\u306f\u5171\u6709\u3059\u308b\u5fc5\u8981\u304c\u3042\u308b\u5834\u5408\u3001MLflow \u306f\u3001\u3055\u307e\u3056\u307e\u306a\u5f62\u5f0f\u306e\u30e2\u30c7\u30eb\u3092\u8aad\u307f\u8fbc\u3093\u3067\u4f7f\u7528\u3059\u308b\u305f\u3081\u306e\u7d71\u5408 API \u3092\u63d0\u4f9b\u3059\u308b\u3053\u3068\u3067\u3001\u30e2\u30c7\u30eb\u306e\u7ba1\u7406\u3068\u30c7\u30d7\u30ed\u30a4\u3092\u7c21\u7d20\u5316\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u306f\u3001\u30aa\u30f3\u30d7\u30ec\u30df\u30b9\u3001\u30af\u30e9\u30a6\u30c9 \u30b5\u30fc\u30d3\u30b9\u3001\u30b3\u30f3\u30c6\u30ca\u30fc\u306a\u3069\u3001\u3055\u307e\u3056\u307e\u306a\u65b9\u6cd5\u3067\u30c7\u30d7\u30ed\u30a4\u3067\u304d\u307e\u3059\u3002 \u3053\u3053\u3067\u306f\u3001MLflow \u30e2\u30c7\u30eb\u3092\u30ed\u30fc\u30ab\u30eb\u306b\u30c7\u30d7\u30ed\u30a4\u3057\u3001\u305d\u306e\u30e2\u30c7\u30eb\u3092 REST API \u3068\u3057\u3066\u63d0\u4f9b\u3059\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>MLflow \u30e2\u30c7\u30eb\u3092 REST API \u3068\u3057\u3066\u30c7\u30d7\u30ed\u30a4\u3059\u308b<\/strong><\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u3092 REST API \u3068\u3057\u3066\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u306b\u306f\u3001MLflow \u72ec\u81ea\u306e\u30e2\u30c7\u30eb \u30b5\u30fc\u30d0\u30fc\u3092\u4f7f\u7528\u3057\u307e\u3059\u3002 \u5c55\u958b\u624b\u9806\u306f\u6b21\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u5fc5\u8981\u306a\u30e9\u30a4\u30d6\u30e9\u30ea<strong>\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u3066\u304f\u3060\u3055\u3044\u307e\u305a<\/strong>\u3001\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3068\u305d\u306e\u4ed6\u306e\u5fc5\u8981\u306a\u30e9\u30a4\u30d6\u30e9\u30ea\u304c\u3042\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002 \u4e00\u90e8\u306e\u30e2\u30c7\u30eb (TensorFlow\u3001PyTorch \u306a\u3069) \u3067\u306f\u3001\u9069\u5207\u306a\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u4f7f\u7528\u3059\u308b:bashCopy\u30b3\u30fc\u30c9<code>mlflow<\/code><code>pip<\/code><code>pip install mlflow<\/code><\/li>\n\n\n\n<li>MLflow \u30e2\u30c7\u30eb \u30b5\u30fc\u30d0\u30fc\u3092\u8d77\u52d5\u3059\u308b\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066&nbsp;<strong>MLflow \u30e2\u30c7\u30eb \u30b5\u30fc\u30d0\u30fc\u3092\u8d77\u52d5\u3057\u307e\u3059<\/strong>: bashCopy code\u3053\u3053\u3067:<code>mlflow models serve -m \/path\/to\/your\/model -h 0.0.0.0 -p 1234<\/code><ul><li><code>-m \/path\/to\/your\/model<\/code>: MLflow \u30e2\u30c7\u30eb\u304c\u914d\u7f6e\u3055\u308c\u3066\u3044\u308b\u30d1\u30b9\u3092\u6307\u5b9a\u3057\u307e\u3059\u3002<\/li><li><code>-h 0.0.0.0<\/code>: \u30b5\u30fc\u30d0\u30fc \u30d0\u30a4\u30f3\u30c9\u306e IP \u30a2\u30c9\u30ec\u30b9\u3092\u6307\u5b9a\u3057\u307e\u3059\u3002 \u3053\u3053\u3067\u306f\u3001\u4efb\u610f\u306eIP\u30a2\u30c9\u30ec\u30b9\u304b\u3089\u30b5\u30fc\u30d0\u30fc\u3078\u306e\u30a2\u30af\u30bb\u30b9\u3092\u8a31\u53ef\u3059\u308b\u305f\u3081\u306b\u4f7f\u7528\u3057\u307e\u3059\u3002<code>0.0.0.0<\/code><\/li><li><code>-p 1234<\/code>: \u30b5\u30fc\u30d0\u30fc\u304c\u30ea\u30c3\u30b9\u30f3\u3059\u308b\u30dd\u30fc\u30c8\u756a\u53f7\u3092\u6307\u5b9a\u3057\u307e\u3059\u3002 \u3053\u3053\u3067\u306f\u3001\u30dd\u30fc\u30c8\u3092\u4f7f\u7528\u3057\u307e\u3059\u3002<code>1234<\/code><\/li><\/ul>\u30b5\u30fc\u30d0\u30fc\u3092\u8d77\u52d5\u3059\u308b\u3068\u3001\u30b3\u30f3\u30bd\u30fc\u30eb\u306b\u540c\u69d8\u306e\u51fa\u529b\u304c\u8868\u793a\u3055\u308c\u307e\u3059\u3002 yamlCopy \u30b3\u30fc\u30c9<code>2023\/04\/16 12:34:56 INFO mlflow.models.cli: Selected backend for flavor 'python_function' 2023\/04\/16 12:34:56 INFO mlflow.pyfunc.backend: No conda environment specified. Using current python version. 2023\/04\/16 12:34:56 INFO mlflow.pyfunc.backend: No gunicorn_opts specified, defaulting gunicorn_opts to \"\". 2023\/04\/16 12:34:56 INFO mlflow.models.cli: Starting gunicorn server with command: gunicorn --timeout=60 -k gevent -b 0.0.0.0:1234 -w 1 wsgi:app 2023\/04\/16 12:34:56 [INFO] Starting gunicorn 20.1.0 2023\/04\/16 12:34:56 [INFO] Listening at: http:\/\/0.0.0.0:1234 (1) 2023\/04\/16 12:34:56 [INFO] Using worker: gevent 2023\/04\/16 12:34:56 [INFO] Booting worker with pid: 12345<\/code><\/li>\n\n\n\n<li><strong>REST API \u306b\u8981\u6c42\u3092\u9001\u4fe1\u3059\u308b<\/strong>&nbsp;MLflow \u30e2\u30c7\u30eb \u30b5\u30fc\u30d0\u30fc\u306f\u3001\u6307\u5b9a\u3055\u308c\u305f IP \u30a2\u30c9\u30ec\u30b9\u3068\u30dd\u30fc\u30c8\u3067\u8981\u6c42\u3092\u30ea\u30c3\u30b9\u30f3\u3057\u307e\u3059\u3002 \u30c7\u30d5\u30a9\u30eb\u30c8\u3067\u306f\u3001\u30b5\u30fc\u30d0\u30fc\u306f\u30d1\u30b9\u4e0a\u3067\u4e88\u6e2c\u30b5\u30fc\u30d3\u30b9\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 \u3001Postman\u3001\u307e\u305f\u306fPython\u306e\u30e9\u30a4\u30d6\u30e9\u30ea\u306a\u3069\u306e\u4efb\u610f\u306eHTTP\u30af\u30e9\u30a4\u30a2\u30f3\u30c8\u3092\u4f7f\u7528\u3057\u3066\u30ea\u30af\u30a8\u30b9\u30c8\u3092\u9001\u4fe1\u3067\u304d\u307e\u3059<code>\/invocations<\/code><code>curl<\/code><code>requests<\/code><\/li>\n<\/ol>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u5f85\u3064\u3002 Python \u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u4f7f\u7528\u3057\u3066\u30ea\u30af\u30a8\u30b9\u30c8\u3092\u9001\u4fe1\u3059\u308b\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<code>requests<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import json\nimport requests\n\n# \u51c6\u5907\u8f93\u5165\u6570\u636e\ninput_data = {\n    \"columns\": [\"feature1\", \"feature2\"],\n    \"data\": [[1.0, 2.0], [3.0, 4.0]]\n}\n\n# \u5c06\u8f93\u5165\u6570\u636e\u8f6c\u6362\u4e3a JSON \u683c\u5f0f\ninput_json = json.dumps(input_data)\n\n# \u53d1\u9001 POST \u8bf7\u6c42\u5230 MLflow Model \u670d\u52a1\u5668\nresponse = requests.post(\"http:\/\/0.0.0.0:1234\/invocations\", data=input_json, headers={\"Content-Type\": \"application\/json\"})\n\n# \u89e3\u6790\u670d\u52a1\u5668\u8fd4\u56de\u7684\u9884\u6d4b\u7ed3\u679c\npredictions = json.loads(response.text)\nprint(predictions)\n<\/code><\/pre>\n\n\n\n<p>\u3053\u306e\u4f8b\u3067\u306f\u3001\u6700\u521d\u306b\u5165\u529b\u30c7\u30fc\u30bf\u3092\u6e96\u5099\u3057\u3066\u304b\u3089\u3001JSON \u5f62\u5f0f\u306b\u5909\u63db\u3057\u307e\u3059\u3002 \u6b21\u306b\u3001\u3053\u306e\u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066 MLflow \u30e2\u30c7\u30eb \u30b5\u30fc\u30d0\u30fc\u306b POST \u8981\u6c42\u3092\u9001\u4fe1\u3057\u307e\u3059\u3002 \u6700\u5f8c\u306b\u3001\u30b5\u30fc\u30d0\u30fc\u304b\u3089\u8fd4\u3055\u308c\u305f\u4e88\u6e2c\u3092\u89e3\u6790\u3057\u307e\u3059\u3002<code>requests.post()<\/code><\/p>\n\n\n\n<p>\u3053\u306e\u30c7\u30d7\u30ed\u30a4\u306b\u3088\u308a\u3001MLflow \u30e2\u30c7\u30eb\u3092 REST API \u3068\u3057\u3066\u63d0\u4f9b\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u3001\u4ed6\u306e\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u3067\u30e2\u30c7\u30eb\u3092\u547c\u3073\u51fa\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3055\u3089\u306b\u3001\u30e2\u30c7\u30eb\u306e\u7ba1\u7406\u3068\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u306e\u305f\u3081\u306b\u3001MLflow \u30e2\u30c7\u30eb\u3092\u30af\u30e9\u30a6\u30c9 \u30b5\u30fc\u30d3\u30b9 (AWS Sagemaker\u3001Azure ML \u306a\u3069) \u307e\u305f\u306f\u30b3\u30f3\u30c6\u30ca\u30fc (Docker\u3001Kubernetes \u306a\u3069) \u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u3053\u3068\u3092\u691c\u8a0e\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3057\u305f\u30e2\u30c7\u30eb\u306e\u7ba1\u7406\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u4e00\u5143\u5316\u3055\u308c\u305f\u30e2\u30c7\u30eb \u30ea\u30dd\u30b8\u30c8\u30ea\u3067\u3059\u3002 \u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u8ffd\u8de1\u3057\u305f\u308a\u3001\u958b\u767a\u3084\u904b\u7528\u306a\u3069\u306e\u30e2\u30c7\u30eb\u306e\u30d5\u30a7\u30fc\u30ba\u3092\u7ba1\u7406\u3057\u305f\u308a\u3001\u30c1\u30fc\u30e0\u30e1\u30f3\u30d0\u30fc\u3068\u5171\u540c\u4f5c\u696d\u3057\u305f\u308a\u3067\u304d\u308b\u30a4\u30f3\u30bf\u30fc\u30d5\u30a7\u30a4\u30b9\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u7ba1\u7406\u3059\u308b\u65b9\u6cd5\u306e\u8a73\u7d30\u306a\u8aac\u660e\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1. \u30e2\u30c7\u30eb\u306e\u767b\u9332<\/strong><\/p>\n\n\n\n<p>\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306e\u4f7f\u7528\u3092\u958b\u59cb\u3059\u308b\u306b\u306f\u3001\u307e\u305a\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6e08\u307f\u306e MLflow \u30e2\u30c7\u30eb\u3092\u30ea\u30dd\u30b8\u30c8\u30ea\u306b\u767b\u9332\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb\u3092\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306b\u767b\u9332\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow\n\n# \u6ce8\u518c\u6a21\u578b\nmodel_uri = \"runs:\/&lt;run_id&gt;\/model\"  # \u5c06 &lt;run_id&gt; \u66ff\u6362\u4e3a\u5b9e\u9645\u7684\u8fd0\u884c ID\nregistered_model = mlflow.register_model(model_uri, \"MyModel\")\n<\/code><\/pre>\n\n\n\n<p>\u3053\u3053\u3067\u306f\u3001\u767b\u9332\u3059\u308b\u30e2\u30c7\u30eb\u306e\u5834\u6240\u3092 \u306e\u5f62\u5f0f\u3067\u6307\u5b9a\u3057\u307e\u3059\u3002 \u306f\u3001\u767b\u9332\u3055\u308c\u3066\u3044\u308b\u30e2\u30c7\u30eb\u306e\u540d\u524d\u3067\u3059\u3002 \u30e2\u30c7\u30eb\u304c\u767b\u9332\u3055\u308c\u308b\u3068\u3001MLflow \u306b\u3088\u3063\u3066\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u304c\u4f5c\u6210\u3055\u308c\u307e\u3059 (\u65e2\u5b9a\u3067\u306f\u30d0\u30fc\u30b8\u30e7\u30f3 1)\u3002<code>model_uri<\/code><code>runs:\/&lt;run_id&gt;\/model<\/code><code>\"MyModel\"<\/code><\/p>\n\n\n\n<p><strong>2. \u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u78ba\u8a8d\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3067\u30e2\u30c7\u30eb\u306e\u3059\u3079\u3066\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u8868\u793a\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code># \u67e5\u770b\u6a21\u578b\u7248\u672c\nmodel_versions = mlflow.search_model_versions(\"name='MyModel'\")\nfor mv in model_versions:\n    print(\"Version: {}\\nURI: {}\\n\".format(mv.version, mv.source))\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb\u300cMyModel\u300d\u306e\u3059\u3079\u3066\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3068\u305d\u308c\u3089\u304c\u4fdd\u5b58\u3055\u308c\u3066\u3044\u308b\u5834\u6240\u304c\u51fa\u529b\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>3. \u30e2\u30c7\u30eb\u306e\u66f4\u65b0\u30d5\u30a7\u30fc\u30ba<\/strong><\/p>\n\n\n\n<p>\u30e2\u30c7\u30eb\u30b9\u30c6\u30fc\u30b8\u306f\u3001\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u306b\u304a\u3051\u308b\u30e2\u30c7\u30eb\u306e\u73fe\u5728\u306e\u72b6\u614b\u3092\u8b58\u5225\u3059\u308b\u305f\u3081\u306b\u4f7f\u7528\u3055\u308c\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u30e2\u30c7\u30eb\u304c\u958b\u767a\u30d5\u30a7\u30fc\u30ba\u306b\u3042\u308b\u5834\u5408\u3001\u305d\u306e\u30b9\u30c6\u30fc\u30b8\u3092\u300c\u30b9\u30c6\u30fc\u30b8\u30f3\u30b0\u300d\u306b\u8a2d\u5b9a\u3067\u304d\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u6e96\u5099\u304c\u3067\u304d\u305f\u3089\u3001\u305d\u306e\u30b9\u30c6\u30fc\u30b8\u3092 [\u904b\u7528] \u306b\u8a2d\u5b9a\u3067\u304d\u307e\u3059\u3002 \u30e2\u30c7\u30eb \u30b9\u30c6\u30fc\u30b8\u3092\u66f4\u65b0\u3059\u308b\u65b9\u6cd5\u306e\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code># \u5c06\u6a21\u578b\u9636\u6bb5\u8bbe\u7f6e\u4e3a \"Staging\"\nmlflow.set_registered_model_tag(\"MyModel\", \"Staging\")\n\n# \u5c06\u6a21\u578b\u9636\u6bb5\u8bbe\u7f6e\u4e3a \"Production\"\nmlflow.set_registered_model_tag(\"MyModel\", \"Production\")\n<\/code><\/pre>\n\n\n\n<p>\u3053\u306e\u4f8b\u3067\u306f\u3001\u3053\u306e\u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb &#8220;MyModel&#8221; \u306e\u30b9\u30c6\u30fc\u30b8\u3092\u305d\u308c\u305e\u308c &#8220;\u30b9\u30c6\u30fc\u30b8\u30f3\u30b0&#8221; \u3068 &#8220;\u904b\u7528&#8221; \u306b\u8a2d\u5b9a\u3057\u307e\u3059\u3002<code>mlflow.set_registered_model_tag()<\/code><\/p>\n\n\n\n<p><strong>4. \u30e2\u30c7\u30eb\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u524a\u9664\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u524a\u9664\u3059\u308b\u5fc5\u8981\u304c\u3042\u308b\u5834\u5408\u306f\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code># \u5220\u9664\u6a21\u578b\u7248\u672c\nmlflow.delete_model_version(\"MyModel\", 1)\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb &#8220;MyModel&#8221; \u306e\u30d0\u30fc\u30b8\u30e7\u30f3 1 \u304c\u524a\u9664\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>5. \u5b9f\u7a3c\u50cd\u30e2\u30c7\u30eb\u3092\u5165\u624b\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u73fe\u5728\u300c\u672c\u756a\u300d\u6bb5\u968e\u306b\u3042\u308b\u30e2\u30c7\u30eb\u3092\u53d6\u5f97\u3059\u308b\u5fc5\u8981\u304c\u3042\u308b\u5834\u5408\u306f\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code># \u83b7\u53d6\u751f\u4ea7\u6a21\u578b\nproduction_model_version = mlflow.get_latest_versions(\"MyModel\", stages=[\"Production\"])[0]\nproduction_model_uri = production_model_version.source\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u73fe\u5728 &#8220;\u904b\u7528&#8221; \u6bb5\u968e\u306b\u3042\u308b\u30e2\u30c7\u30eb &#8220;MyModel&#8221; \u306e\u6700\u65b0\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u4fdd\u5b58\u5834\u6240\u304c\u8fd4\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u3092\u4f7f\u7528\u3059\u308b<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3068\u30b9\u30c6\u30fc\u30b8\u3092\u7c21\u5358\u306b\u7ba1\u7406\u3067\u304d\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u3092\u3088\u308a\u9069\u5207\u306b\u7ba1\u7406\u3059\u308b\u306e\u306b\u5f79\u7acb\u3064\u3044\u304f\u3064\u304b\u306e\u8ffd\u52a0\u6a5f\u80fd\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>6. \u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u6bd4\u8f03\u3059\u308b<\/strong><\/p>\n\n\n\n<p>MLflow UI \u306e\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3067\u3001\u3055\u307e\u3056\u307e\u306a\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u304a\u3088\u3073\u6210\u679c\u7269\u3092\u6bd4\u8f03\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u5909\u5316\u3092\u7406\u89e3\u3057\u3001\u30c7\u30d7\u30ed\u30a4\u306b\u6700\u9069\u306a\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u6c7a\u5b9a\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u6bd4\u8f03\u3059\u308b\u306b\u306f\u3001[\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea] \u30da\u30fc\u30b8\u306b\u79fb\u52d5\u3057\u3001\u6bd4\u8f03\u3059\u308b\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u9078\u629e\u3057\u3066\u3001[\u6bd4\u8f03] \u30dc\u30bf\u30f3\u3092\u30af\u30ea\u30c3\u30af\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>7.\u8aac\u660e\u3068\u30bf\u30b0\u3092\u8ffd\u52a0\u3057\u307e\u3059<\/strong><\/p>\n\n\n\n<p>\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u30e2\u30c7\u30eb\u306e\u76ee\u7684\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u7406\u89e3\u3057\u3084\u3059\u304f\u3059\u308b\u305f\u3081\u306b\u3001\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u306b\u8aac\u660e\u3068\u30bf\u30b0\u3092\u8ffd\u52a0\u3067\u304d\u307e\u3059\u3002 \u8aac\u660e\u306b\u306f\u3001\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3001\u6a5f\u80fd\u306e\u9078\u629e\u3001\u30cf\u30a4\u30d1\u30fc\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u306e\u8abf\u6574\u306a\u3069\u3001\u30e2\u30c7\u30eb\u306e\u8a73\u7d30\u3092\u542b\u3081\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u30e9\u30d9\u30eb\u306f\u3001\u56de\u5e30\u3001\u5206\u985e\u3001\u30ec\u30b3\u30e1\u30f3\u30c0\u30fc \u30b7\u30b9\u30c6\u30e0\u306a\u3069\u306e\u30e2\u30c7\u30eb\u3092\u5206\u985e\u3059\u308b\u305f\u3081\u306b\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002 \u8aac\u660e\u3068\u30bf\u30b0\u3092\u8ffd\u52a0\u3059\u308b\u306b\u306f\u3001\u30e2\u30c7\u30eb\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u8a73\u7d30\u30da\u30fc\u30b8\u306b\u79fb\u52d5\u3057\u3001[\u7de8\u96c6] \u30dc\u30bf\u30f3\u3092\u30af\u30ea\u30c3\u30af\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>8.\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u6a29\u9650\u7ba1\u7406<\/strong><\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3067\u306f\u3001\u30de\u30eb\u30c1\u30e6\u30fc\u30b6\u30fc \u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3092 MLflow \u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u306b\u62db\u5f85\u3057\u3001\u30ed\u30fc\u30eb\u3068\u30a2\u30af\u30bb\u30b9\u8a31\u53ef\u3092\u5272\u308a\u5f53\u3066\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u7279\u5b9a\u306e\u30e6\u30fc\u30b6\u30fc\u306b\u30e2\u30c7\u30eb\u306e\u8868\u793a\u3001\u65b0\u3057\u3044\u30e2\u30c7\u30eb\u306e\u767b\u9332\u3001\u30e2\u30c7\u30eb\u306e\u30b9\u30c6\u30fc\u30b8\u306e\u5909\u66f4\u3092\u8a31\u53ef\u3067\u304d\u307e\u3059\u3002 \u30a2\u30af\u30bb\u30b9\u8a31\u53ef\u7ba1\u7406\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u5fc5\u8981\u306a\u30e2\u30c7\u30eb\u306b\u306e\u307f\u30a2\u30af\u30bb\u30b9\u3057\u3066\u5909\u66f4\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u4e00\u5143\u7684\u306a\u65b9\u6cd5\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u8ffd\u8de1\u3001\u30e2\u30c7\u30eb\u306e\u30b9\u30c6\u30fc\u30b8\u306e\u7ba1\u7406\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3068\u306e\u5171\u540c\u4f5c\u696d\u3001\u30e2\u30c7\u30eb\u306e\u6574\u7406\u3068\u7ba1\u7406\u3092\u7c21\u5358\u306b\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306e\u767b\u9332\u30e2\u30c7\u30eb\u3068\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u6559\u3048\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb\u3068\u305d\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u4e00\u5143\u5316\u3055\u308c\u305f\u30e2\u30c7\u30eb \u30ea\u30dd\u30b8\u30c8\u30ea\u3067\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u8ffd\u8de1\u3001\u30e2\u30c7\u30eb\u306e\u30b9\u30c6\u30fc\u30b8\u306e\u7ba1\u7406\u3001\u30c1\u30fc\u30e0\u30e1\u30f3\u30d0\u30fc\u3068\u306e\u5171\u540c\u4f5c\u696d\u3001\u30e2\u30c7\u30eb\u306e\u6574\u7406\u3068\u7ba1\u7406\u306e\u6539\u5584\u3092\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u767b\u9332\u3057\u3001\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u7ba1\u7406\u3059\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>\u30e2\u30c7\u30eb\u3092\u767b\u9332\u3059\u308b<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u3066 MLflow \u306b\u4fdd\u5b58\u3057\u305f\u3089\u3001\u30e2\u30c7\u30eb\u3092\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306b\u767b\u9332\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u306f\u3001\u6b21\u306e\u65b9\u6cd5\u3067\u767b\u9332\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow\n\n# \u6ce8\u518c\u6a21\u578b\nmodel_uri = \"runs:\/&lt;run_id&gt;\/model\"  # \u5c06 &lt;run_id&gt; \u66ff\u6362\u4e3a\u5b9e\u9645\u7684\u8fd0\u884c ID\nregistered_model = mlflow.register_model(model_uri, \"MyModel\")\n<\/code><\/pre>\n\n\n\n<p>\u3053\u3053\u3067\u306f\u3001\u767b\u9332\u3059\u308b\u30e2\u30c7\u30eb\u306e\u5834\u6240\u3092 \u306e\u5f62\u5f0f\u3067\u6307\u5b9a\u3057\u307e\u3059\u3002 \u306f\u3001\u767b\u9332\u3055\u308c\u3066\u3044\u308b\u30e2\u30c7\u30eb\u306e\u540d\u524d\u3067\u3059\u3002 \u30e2\u30c7\u30eb\u304c\u767b\u9332\u3055\u308c\u308b\u3068\u3001MLflow \u306b\u3088\u3063\u3066\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u304c\u4f5c\u6210\u3055\u308c\u307e\u3059 (\u65e2\u5b9a\u3067\u306f\u30d0\u30fc\u30b8\u30e7\u30f3 1)\u3002<code>model_uri<\/code><code>runs:\/&lt;run_id&gt;\/model<\/code><code>\"MyModel\"<\/code><\/p>\n\n\n\n<p><strong>\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u8868\u793a\u3059\u308b<\/strong><\/li>\n<\/ol>\n\n\n\n<p>\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u5185\u306e\u30e2\u30c7\u30eb\u306e\u3059\u3079\u3066\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u8868\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code># \u67e5\u770b\u6a21\u578b\u7248\u672c\nmodel_versions = mlflow.search_model_versions(\"name='MyModel'\")\nfor mv in model_versions:\n    print(\"Version: {}\\nURI: {}\\n\".format(mv.version, mv.source))\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb\u300cMyModel\u300d\u306e\u3059\u3079\u3066\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3068\u305d\u308c\u3089\u304c\u4fdd\u5b58\u3055\u308c\u3066\u3044\u308b\u5834\u6240\u304c\u51fa\u529b\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"2\">\n<li><strong>\u65b0\u3057\u3044\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u4f5c\u6210\u3059\u308b<\/strong><\/li>\n<\/ol>\n\n\n\n<p>\u30e2\u30c7\u30eb\u3092\u6539\u5584\u3057\u3066\u518d\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u305f\u5834\u5408\u306f\u3001\u540c\u3058\u30e2\u30c7\u30eb\u540d\u3067\u65b0\u3057\u3044\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u4f5c\u6210\u3067\u304d\u307e\u3059\u3002 MLflow \u306f\u3001\u65b0\u3057\u3044\u30e2\u30c7\u30eb\u306b\u6b21\u306b\u4f7f\u7528\u53ef\u80fd\u306a\u30d0\u30fc\u30b8\u30e7\u30f3\u756a\u53f7\u3092\u81ea\u52d5\u7684\u306b\u5272\u308a\u5f53\u3066\u307e\u3059\u3002 \u65b0\u3057\u3044\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u4f5c\u6210\u3059\u308b\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code># \u521b\u5efa\u65b0\u6a21\u578b\u7248\u672c\nnew_model_uri = \"runs:\/&lt;new_run_id&gt;\/model\"  # \u5c06 &lt;new_run_id&gt; \u66ff\u6362\u4e3a\u65b0\u7684\u8fd0\u884c ID\nmlflow.register_model(new_model_uri, \"MyModel\")\n<\/code><\/pre>\n\n\n\n<ol class=\"wp-block-list\" start=\"3\">\n<li><strong>\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u524a\u9664\u3059\u308b<\/strong><\/li>\n<\/ol>\n\n\n\n<p>\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u524a\u9664\u3059\u308b\u5fc5\u8981\u304c\u3042\u308b\u5834\u5408\u306f\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code># \u5220\u9664\u6a21\u578b\u7248\u672c\nmlflow.delete_model_version(\"MyModel\", 1)\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb &#8220;MyModel&#8221; \u306e\u30d0\u30fc\u30b8\u30e7\u30f3 1 \u304c\u524a\u9664\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"4\">\n<li><strong>\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u6bd4\u8f03\u3059\u308b<\/strong><\/li>\n<\/ol>\n\n\n\n<p>MLflow UI \u3067\u306f\u3001\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u5185\u306e\u3055\u307e\u3056\u307e\u306a\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u304a\u3088\u3073\u6210\u679c\u7269\u3092\u6bd4\u8f03\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u304c\u3069\u306e\u3088\u3046\u306b\u5909\u5316\u3059\u308b\u304b\u3092\u7406\u89e3\u3057\u3001\u30c7\u30d7\u30ed\u30a4\u306b\u6700\u9069\u306a\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u6c7a\u5b9a\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u6bd4\u8f03\u3059\u308b\u306b\u306f\u3001[\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea] \u30da\u30fc\u30b8\u306b\u79fb\u52d5\u3057\u3001\u6bd4\u8f03\u3059\u308b\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u9078\u629e\u3057\u3066\u3001[\u6bd4\u8f03] \u30dc\u30bf\u30f3\u3092\u30af\u30ea\u30c3\u30af\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u767b\u9332\u304a\u3088\u3073\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u3059\u308b\u3053\u3068\u3067\u3001\u30e2\u30c7\u30eb\u306e\u6539\u5584\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u5909\u66f4\u3092\u3088\u308a\u9069\u5207\u306b\u8ffd\u8de1\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u5e38\u306b\u6700\u9069\u306a\u30e2\u30c7\u30eb\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u4f7f\u7528\u3057\u3001\u3044\u3064\u3067\u3082\u4ee5\u524d\u306e\u30e2\u30c7\u30eb\u30d0\u30fc\u30b8\u30e7\u30f3\u306b\u30ed\u30fc\u30eb\u30d0\u30c3\u30af\u3067\u304d\u307e\u3059\u3002 \u3055\u3089\u306b\u3001\u30e2\u30c7\u30eb\u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u30c1\u30fc\u30e0\u30e1\u30f3\u30d0\u30fc\u3068\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3001\u4e16\u4ee3\u9593\u3092\u53ef\u80fd\u306b\u3057\u307e\u3059<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u30c1\u30fc\u30e0\u306e\u30e2\u30c7\u30eb\u306e\u5171\u6709\u3068\u7ba1\u7406\u304c\u3055\u3089\u306b\u7c21\u5358\u306b\u306a\u308a\u307e\u3057\u305f\u3002<\/p>\n\n\n\n<p><strong>\u30e2\u30c7\u30eb\u30d5\u30a7\u30fc\u30ba\u7ba1\u7406<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u30e2\u30c7\u30eb\u30b9\u30c6\u30fc\u30b8\u306e\u66f4\u65b0<\/strong><\/li>\n<\/ol>\n\n\n\n<p>\u30e2\u30c7\u30eb\u30b9\u30c6\u30fc\u30b8\u306f\u3001\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u306b\u304a\u3051\u308b\u30e2\u30c7\u30eb\u306e\u73fe\u5728\u306e\u72b6\u614b\u3092\u8b58\u5225\u3059\u308b\u305f\u3081\u306b\u4f7f\u7528\u3055\u308c\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u30e2\u30c7\u30eb\u304c\u958b\u767a\u30d5\u30a7\u30fc\u30ba\u306b\u3042\u308b\u5834\u5408\u3001\u305d\u306e\u30b9\u30c6\u30fc\u30b8\u3092\u300c\u30b9\u30c6\u30fc\u30b8\u30f3\u30b0\u300d\u306b\u8a2d\u5b9a\u3067\u304d\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u6e96\u5099\u304c\u3067\u304d\u305f\u3089\u3001\u305d\u306e\u30b9\u30c6\u30fc\u30b8\u3092 [\u904b\u7528] \u306b\u8a2d\u5b9a\u3067\u304d\u307e\u3059\u3002 \u30e2\u30c7\u30eb \u30b9\u30c6\u30fc\u30b8\u3092\u66f4\u65b0\u3059\u308b\u65b9\u6cd5\u306e\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code># \u5c06\u6a21\u578b\u9636\u6bb5\u8bbe\u7f6e\u4e3a \"Staging\"\nmlflow.set_registered_model_tag(\"MyModel\", \"Staging\")\n\n# \u5c06\u6a21\u578b\u9636\u6bb5\u8bbe\u7f6e\u4e3a \"Production\"\nmlflow.set_registered_model_tag(\"MyModel\", \"Production\")\n<\/code><\/pre>\n\n\n\n<p>\u3053\u306e\u4f8b\u3067\u306f\u3001\u3053\u306e\u95a2\u6570\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb &#8220;MyModel&#8221; \u306e\u30b9\u30c6\u30fc\u30b8\u3092\u305d\u308c\u305e\u308c &#8220;\u30b9\u30c6\u30fc\u30b8\u30f3\u30b0&#8221; \u3068 &#8220;\u904b\u7528&#8221; \u306b\u8a2d\u5b9a\u3057\u307e\u3059\u3002<code>mlflow.set_registered_model_tag()<\/code><\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"2\">\n<li><strong>\u7279\u5b9a\u306e\u30b9\u30c6\u30fc\u30b8\u306e\u30e2\u30c7\u30eb\u3092\u53d6\u5f97\u3059\u308b<\/strong><\/li>\n<\/ol>\n\n\n\n<p>\u7279\u5b9a\u306e\u6bb5\u968e\u3067\u30e2\u30c7\u30eb\u3092\u53d6\u5f97\u3059\u308b\u5fc5\u8981\u304c\u3042\u308b\u5834\u5408\u306f\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code># \u83b7\u53d6\u751f\u4ea7\u6a21\u578b\nproduction_model_version = mlflow.get_latest_versions(\"MyModel\", stages=[\"Production\"])[0]\nproduction_model_uri = production_model_version.source\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u73fe\u5728 &#8220;\u904b\u7528&#8221; \u6bb5\u968e\u306b\u3042\u308b\u30e2\u30c7\u30eb &#8220;MyModel&#8221; \u306e\u6700\u65b0\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u4fdd\u5b58\u5834\u6240\u304c\u8fd4\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u6a29\u9650\u7ba1\u7406<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3067\u306f\u3001\u30de\u30eb\u30c1\u30e6\u30fc\u30b6\u30fc \u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3092 MLflow \u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u306b\u62db\u5f85\u3057\u3001\u30ed\u30fc\u30eb\u3068\u30a2\u30af\u30bb\u30b9\u8a31\u53ef\u3092\u5272\u308a\u5f53\u3066\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u7279\u5b9a\u306e\u30e6\u30fc\u30b6\u30fc\u306b\u30e2\u30c7\u30eb\u306e\u8868\u793a\u3001\u65b0\u3057\u3044\u30e2\u30c7\u30eb\u306e\u767b\u9332\u3001\u30e2\u30c7\u30eb\u306e\u30b9\u30c6\u30fc\u30b8\u306e\u5909\u66f4\u3092\u8a31\u53ef\u3067\u304d\u307e\u3059\u3002 \u30a2\u30af\u30bb\u30b9\u8a31\u53ef\u7ba1\u7406\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u5fc5\u8981\u306a\u30e2\u30c7\u30eb\u306b\u306e\u307f\u30a2\u30af\u30bb\u30b9\u3057\u3066\u5909\u66f4\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u307e\u305f\u3001\u30e2\u30c7\u30eb\u30d0\u30fc\u30b8\u30e7\u30f3\u306b\u8aac\u660e\u3068\u30bf\u30b0\u3092\u8ffd\u52a0\u3057\u3066\u3001\u30c1\u30fc\u30e0\u30e1\u30f3\u30d0\u30fc\u304c\u30e2\u30c7\u30eb\u306e\u76ee\u7684\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u7406\u89e3\u3059\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002 \u8aac\u660e\u306b\u306f\u3001\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3001\u6a5f\u80fd\u306e\u9078\u629e\u3001\u30cf\u30a4\u30d1\u30fc\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u306e\u8abf\u6574\u306a\u3069\u3001\u30e2\u30c7\u30eb\u306e\u8a73\u7d30\u3092\u542b\u3081\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u30e9\u30d9\u30eb\u306f\u3001\u56de\u5e30\u3001\u5206\u985e\u3001\u30ec\u30b3\u30e1\u30f3\u30c0\u30fc \u30b7\u30b9\u30c6\u30e0\u306a\u3069\u306e\u30e2\u30c7\u30eb\u3092\u5206\u985e\u3059\u308b\u305f\u3081\u306b\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002 \u8aac\u660e\u3068\u30bf\u30b0\u3092\u8ffd\u52a0\u3059\u308b\u306b\u306f\u3001\u30e2\u30c7\u30eb\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u8a73\u7d30\u30da\u30fc\u30b8\u306b\u79fb\u52d5\u3057\u3001[\u7de8\u96c6] \u30dc\u30bf\u30f3\u3092\u30af\u30ea\u30c3\u30af\u3057\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u4e00\u5143\u7684\u306a\u65b9\u6cd5\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u8ffd\u8de1\u3001\u30e2\u30c7\u30eb\u306e\u30b9\u30c6\u30fc\u30b8\u306e\u7ba1\u7406\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3068\u306e\u5171\u540c\u4f5c\u696d\u3001\u30e2\u30c7\u30eb\u306e\u6574\u7406\u3068\u7ba1\u7406\u3092\u7c21\u5358\u306b\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30c1\u30fc\u30e0\u304c\u30e2\u30c7\u30eb\u3092\u5171\u6709\u304a\u3088\u3073\u7ba1\u7406\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u3001\u5168\u4f53\u7684\u306a\u751f\u7523\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea \u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u7ba1\u7406\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u77e5\u308b<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u8ffd\u8de1\u3001\u30e2\u30c7\u30eb\u306e\u30d5\u30a7\u30fc\u30ba\u306e\u7ba1\u7406\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3068\u306e\u5171\u540c\u4f5c\u696d\u306a\u3069\u3001\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u4e00\u5143\u7684\u306b\u7ba1\u7406\u3059\u308b\u65b9\u6cd5\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea \u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u7ba1\u7406\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u898b\u3066\u3044\u304d\u307e\u3057\u3087\u3046\u3002<\/p>\n\n\n\n<p><strong>\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u7ba1\u7406<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u30e2\u30c7\u30eb&nbsp;<strong>\u30d0\u30fc\u30b8\u30e7\u30f3:<\/strong>&nbsp;\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u5404\u30e2\u30c7\u30eb\u306b\u30d0\u30fc\u30b8\u30e7\u30f3\u756a\u53f7\u3092\u81ea\u52d5\u7684\u306b\u5272\u308a\u5f53\u3066\u3001\u30e2\u30c7\u30eb\u3078\u306e\u5909\u66f4\u3092\u7c21\u5358\u306b\u8ffd\u8de1\u3067\u304d\u308b\u3088\u3046\u306b\u3057\u307e\u3059\u3002 \u65b0\u3057\u3044\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u4f5c\u6210\u3057\u305f\u308a\u3001\u4e0d\u8981\u306b\u306a\u3063\u305f\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u524a\u9664\u3057\u305f\u308a\u3001\u7570\u306a\u308b\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u6bd4\u8f03\u3057\u305f\u308a\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30e2\u30c7\u30eb \u30d5\u30a7\u30fc\u30ba: &#8220;\u958b\u767a&#8221;\u3001&#8221;\u30b9\u30c6\u30fc\u30b8\u30f3\u30b0&#8221;\u3001&#8221;\u5b9f\u7a3c\u50cd&#8221; \u306a\u3069\u306e\u30b9\u30c6\u30fc\u30b8 \u30e9\u30d9\u30eb\u3092<strong>\u30e2\u30c7\u30eb<\/strong>\u306b\u5272\u308a\u5f53\u3066\u3066\u3001\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u306b\u304a\u3051\u308b\u30e2\u30c7\u30eb\u306e\u73fe\u5728\u306e\u72b6\u614b\u3092\u8868\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306f\u3001\u3069\u306e\u30e2\u30c7\u30eb\u304c\u30c7\u30d7\u30ed\u30a4\u306e\u6e96\u5099\u304c\u3067\u304d\u3066\u3044\u3066\u3001\u3069\u306e\u30e2\u30c7\u30eb\u304c\u307e\u3060\u958b\u767a\u4e2d\u307e\u305f\u306f\u30c6\u30b9\u30c8\u4e2d\u3067\u3042\u308b\u304b\u3092\u7406\u89e3\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p><strong>\u30c1\u30fc\u30e0\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u30de\u30eb\u30c1\u30e6\u30fc\u30b6\u30fc \u30b5\u30dd\u30fc\u30c8<\/strong>: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3067\u306f\u3001\u30de\u30eb\u30c1\u30e6\u30fc\u30b6\u30fc \u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3092 MLflow \u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u306b\u62db\u5f85\u3057\u3001\u30ed\u30fc\u30eb\u3068\u30a2\u30af\u30bb\u30b9\u8a31\u53ef\u3092\u5272\u308a\u5f53\u3066\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u306f\u30e2\u30c7\u30eb\u306e\u5171\u6709\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u8868\u793a\u3001\u30e2\u30c7\u30eb\u306e\u5909\u66f4\u306e\u8ffd\u8de1\u3092\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u6a29\u9650\u7ba1\u7406<\/strong>: \u30e2\u30c7\u30eb\u306e\u8868\u793a\u3001\u65b0\u3057\u3044\u30e2\u30c7\u30eb\u306e\u767b\u9332\u3001\u30e2\u30c7\u30eb\u306e\u30b9\u30c6\u30fc\u30b8\u306e\u5909\u66f4\u306a\u3069\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u306b\u7570\u306a\u308b\u6a29\u9650\u3092\u5272\u308a\u5f53\u3066\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u30a2\u30af\u30bb\u30b9\u8a31\u53ef\u7ba1\u7406\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u5fc5\u8981\u306a\u30e2\u30c7\u30eb\u306b\u306e\u307f\u30a2\u30af\u30bb\u30b9\u3057\u3066\u5909\u66f4\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u3053\u3068\u3067\u3001\u30e2\u30c7\u30eb\u306e\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3068\u6574\u5408\u6027\u3092\u4fdd\u8b77\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u8aac\u660e\u3068\u30bf\u30b0: \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u30e2\u30c7\u30eb\u306e\u76ee\u7684\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u7406\u89e3\u3057\u3084\u3059\u304f\u3059\u308b\u305f\u3081\u306b\u3001\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u306b\u8aac\u660e<strong>\u3068\u30bf\u30b0<\/strong>\u3092\u8ffd\u52a0\u3067\u304d\u307e\u3059\u3002 \u8aac\u660e\u306b\u306f\u3001\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3001\u6a5f\u80fd\u306e\u9078\u629e\u3001\u30cf\u30a4\u30d1\u30fc\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u306e\u8abf\u6574\u306a\u3069\u3001\u30e2\u30c7\u30eb\u306e\u8a73\u7d30\u3092\u542b\u3081\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u30e9\u30d9\u30eb\u306f\u3001\u56de\u5e30\u3001\u5206\u985e\u3001\u30ec\u30b3\u30e1\u30f3\u30c0\u30fc \u30b7\u30b9\u30c6\u30e0\u306a\u3069\u306e\u30e2\u30c7\u30eb\u3092\u5206\u985e\u3059\u308b\u305f\u3081\u306b\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30c1\u30fc\u30e0\u30e1\u30f3\u30d0\u30fc\u9593\u306e\u30b3\u30df\u30e5\u30cb\u30b1\u30fc\u30b7\u30e7\u30f3\u306e\u52b9\u7387\u304c\u5411\u4e0a\u3057\u3001\u30e2\u30c7\u30eb\u306e\u91cd\u8981\u306a\u60c5\u5831\u3092\u3059\u3070\u3084\u304f\u7406\u89e3\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9<strong>\u3092\u6bd4\u8f03\u3059\u308b<\/strong>: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u6210\u679c\u7269\u306a\u3069\u3001\u3055\u307e\u3056\u307e\u306a\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u6bd4\u8f03\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u5909\u5316\u3092\u7406\u89e3\u3057\u3001\u30c7\u30d7\u30ed\u30a4\u306b\u6700\u9069\u306a\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u6c7a\u5b9a\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u306f\u3001MLflow UI \u3067\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u7c21\u5358\u306b\u6bd4\u8f03\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u4e00\u5143\u7684\u306a\u65b9\u6cd5\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u8ffd\u8de1\u3001\u30e2\u30c7\u30eb\u306e\u30b9\u30c6\u30fc\u30b8\u306e\u7ba1\u7406\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u3068\u306e\u5171\u540c\u4f5c\u696d\u3001\u30e2\u30c7\u30eb\u306e\u6574\u7406\u3068\u7ba1\u7406\u3092\u7c21\u5358\u306b\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30c1\u30fc\u30e0\u306f\u5171\u6709\u3057\u3001<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u30e2\u30c7\u30eb\u3092\u7ba1\u7406\u3057\u3066\u3001\u5168\u4f53\u7684\u306a\u751f\u7523\u6027\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u5411\u4e0a\u3055\u305b\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>\u30e2\u30c7\u30eb\u627f\u8a8d\u30d7\u30ed\u30bb\u30b9<\/strong><\/p>\n\n\n\n<p>\u30e2\u30c7\u30eb\u306e\u958b\u767a\u4e2d\u306b\u3001\u7279\u5b9a\u306e\u30e2\u30c7\u30eb\u627f\u8a8d\u30d7\u30ed\u30bb\u30b9\u306b\u5f93\u3046\u5fc5\u8981\u304c\u3042\u308b\u5834\u5408\u304c\u3042\u308a\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u306f\u3001\u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u524d\u306b\u30e2\u30c7\u30eb\u3092\u5b8c\u5168\u306b\u78ba\u8a8d\u3057\u3066\u691c\u8a3c\u3067\u304d\u307e\u3059\u3002 \u8003\u3048\u3089\u308c\u308b\u30e2\u30c7\u30eb\u627f\u8a8d\u30d7\u30ed\u30bb\u30b9\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u958b\u767a\u30d5\u30a7\u30fc\u30ba<\/strong>: \u30e2\u30c7\u30eb\u958b\u767a\u8005\u306f\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u3001MLflow \u306b\u4fdd\u5b58\u3057\u307e\u3059\u3002 \u6b21\u306b\u3001\u30e2\u30c7\u30eb\u3092\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306b\u767b\u9332\u3057\u3001\u30e2\u30c7\u30eb \u30b9\u30c6\u30fc\u30b8\u3092\u958b\u767a\u306b\u8a2d\u5b9a\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30ec\u30d3\u30e5\u30fc \u30d5\u30a7\u30fc\u30ba<\/strong>: \u30e2\u30c7\u30eb\u306e\u958b\u767a\u304c\u5b8c\u4e86\u3059\u308b\u3068\u3001\u4ed6\u306e\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u30ec\u30d3\u30e5\u30fc\u3067\u304d\u307e\u3059\u3002 \u30ec\u30d3\u30e5\u30fc\u62c5\u5f53\u8005\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u304a\u3088\u3073\u6210\u679c\u7269\u3092\u8868\u793a\u3057\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3068\u6f5c\u5728\u7684\u306a\u6539\u5584\u70b9\u306b\u3064\u3044\u3066\u958b\u767a\u8005\u3068\u8a71\u3057\u5408\u3046\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u6bb5\u968e\u3067\u3001\u30ec\u30d3\u30e5\u30fc\u62c5\u5f53\u8005\u306f\u30e2\u30c7\u30eb\u3092\u5909\u66f4\u3059\u308b\u5fc5\u8981\u304c\u3042\u308b\u304b\u3001\u30c6\u30b9\u30c8\u3092\u7d9a\u884c\u3059\u308b\u5fc5\u8981\u304c\u3042\u308b\u304b\u3092\u6c7a\u5b9a\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30c6\u30b9\u30c8<\/strong>&nbsp;\u30b9\u30c6\u30fc\u30b8: \u30e2\u30c7\u30eb\u304c\u30ec\u30d3\u30e5\u30fc\u306b\u5408\u683c\u3057\u305f\u3089\u3001\u305d\u306e\u30b9\u30c6\u30fc\u30b8\u3092\u30b9\u30c6\u30fc\u30b8\u30f3\u30b0\u306b\u8a2d\u5b9a\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u6bb5\u968e\u3067\u3001\u30e2\u30c7\u30eb\u306f\u30c6\u30b9\u30c8\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3055\u308c\u3001\u3055\u3089\u306b\u691c\u8a3c\u3055\u308c\u307e\u3059\u3002 \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u306f\u3001\u3053\u306e\u6bb5\u968e\u3067\u3001\u30e2\u30c7\u30eb\u304c\u671f\u5f85\u3055\u308c\u308b\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9 \u30e1\u30c8\u30ea\u30c3\u30af\u3092\u6e80\u305f\u3057\u3066\u3044\u308b\u304b\u3069\u3046\u304b\u3092\u5224\u65ad\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u5b9f\u7a3c\u50cd<\/strong>\u30b9\u30c6\u30fc\u30b8: \u30e2\u30c7\u30eb\u304c\u30c6\u30b9\u30c8\u306b\u5408\u683c\u3057\u3001\u30c7\u30d7\u30ed\u30a4\u306b\u9069\u3057\u3066\u3044\u308b\u3068\u898b\u306a\u3055\u308c\u305f\u3089\u3001\u305d\u306e\u30b9\u30c6\u30fc\u30b8\u3092 [\u5b9f\u7a3c\u50cd] \u306b\u8a2d\u5b9a\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u6bb5\u968e\u3067\u3001\u30e2\u30c7\u30eb\u306f\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3055\u308c\u3001\u5b9f\u969b\u306e\u30e6\u30fc\u30b6\u30fc\u3078\u306e\u30b5\u30fc\u30d3\u30b9\u63d0\u4f9b\u304c\u958b\u59cb\u3055\u308c\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u7d99\u7d9a\u7684\u306a\u76e3\u8996\u3068\u30e1\u30f3\u30c6\u30ca\u30f3\u30b9<\/strong>: \u904b\u7528\u30d5\u30a7\u30fc\u30ba\u3067\u306f\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u306f\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u7d99\u7d9a\u7684\u306b\u76e3\u8996\u3057\u3066\u3001\u30e2\u30c7\u30eb\u304c\u5e38\u306b\u671f\u5f85\u3055\u308c\u308b\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9 \u30e1\u30c8\u30ea\u30c3\u30af\u3092\u6e80\u305f\u3057\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u4f4e\u4e0b\u3084\u305d\u306e\u4ed6\u306e\u554f\u984c\u306b\u6c17\u4ed8\u3044\u305f\u5834\u5408\u306f\u3001\u958b\u767a\u30d5\u30a7\u30fc\u30ba\u306b\u623b\u3063\u3066\u30e2\u30c7\u30eb\u3092\u6700\u9069\u5316\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u3053\u306e\u30e2\u30c7\u30eb\u627f\u8a8d\u30d7\u30ed\u30bb\u30b9\u306b\u5f93\u3046\u3053\u3068\u3067\u3001\u30c7\u30d7\u30ed\u30a4\u524d\u306b\u30e2\u30c7\u30eb\u304c\u5b8c\u5168\u306b\u30ec\u30d3\u30e5\u30fc\u304a\u3088\u3073\u691c\u8a3c\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3067\u304d\u307e\u3059\u3002 \u307e\u305f\u3001\u30c1\u30fc\u30e0\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u6539\u5584\u3057\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u30e2\u30c7\u30eb\u306e\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u5168\u4f53\u3067\u77e5\u8b58\u3068\u7d4c\u9a13\u3092\u5171\u6709\u3067\u304d\u308b\u3088\u3046\u306b\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30c1\u30fc\u30e0\u306e\u30e2\u30c7\u30eb \u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u3088\u308a\u52b9\u7387\u7684\u306b\u7ba1\u7406\u3059\u308b\u306e\u306b\u5f79\u7acb\u3064\u307b\u3069\u5f37\u529b\u3067\u3059\u3002 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u5358\u72ec\u306e\u958b\u767a\u8005\u3067\u3082\u30c1\u30fc\u30e0\u3067\u3082\u3001\u30e2\u30c7\u30eb\u306e\u5909\u66f4\u3092\u3088\u308a\u9069\u5207\u306b\u8ffd\u8de1\u3057\u3001\u540c\u50da\u3068\u77e5\u8b58\u3092\u5171\u6709\u3057\u3001\u5e38\u306b\u6700\u9069\u306a\u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u306e\u306b\u5f79\u7acb\u3064\u8cb4\u91cd\u306a\u30c4\u30fc\u30eb\u3067\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u3092TensorFlow\u3084PyTorch\u306a\u3069\u306e\u30c7\u30a3\u30fc\u30d7\u30e9\u30fc\u30cb\u30f3\u30b0\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3068\u7d71\u5408\u3059\u308b\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u306f\u3001TensorFlow \u3084 PyTorch \u306a\u3069\u306e\u3055\u307e\u3056\u307e\u306a\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0 \u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3068\u7d71\u5408\u3067\u304d\u308b\u305f\u3081\u3001\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0\u5b9f\u9a13\u3092\u7c21\u5358\u306b\u6587\u66f8\u5316\u3057\u3066\u7ba1\u7406\u3067\u304d\u307e\u3059\u3002 MLflow \u3092 TensorFlow \u304a\u3088\u3073 PyTorch \u3068\u7d71\u5408\u3059\u308b\u65b9\u6cd5\u306e\u8a73\u7d30\u306a\u8aac\u660e\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1. \u30c6\u30f3\u30bd\u30eb\u30d5\u30ed\u30fc\u306e\u7d71\u5408<\/strong><\/p>\n\n\n\n<p>MLflow \u3092 TensorFlow \u3068\u7d71\u5408\u3059\u308b\u306b\u306f\u3001TensorFlow \u3068 MLflow \u7528\u306e TensorFlow \u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u307e\u305a\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066TensorFlow\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">Copy code<code>pip install tensorflow\n<\/code><\/pre>\n\n\n\n<p>\u6b21\u306b\u3001MLflow \u306e TensorFlow \u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">cssCopy code<code>pip install mlflow[tensorflow]\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u3067\u3001MLflow \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u306e\u7d50\u679c\u3092 TensorFlow \u30b3\u30fc\u30c9\u306b\u8a18\u9332\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3057\u305f\u3002 \u6b21\u306b\u7c21\u5358\u306a\u4f8b\u3092\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import tensorflow as tf\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.layers import Dense\nimport mlflow\nimport mlflow.tensorflow\n\n# \u5f00\u542f MLflow \u81ea\u52a8\u8bb0\u5f55\nmlflow.tensorflow.autolog()\n\n# \u51c6\u5907\u6570\u636e\n(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()\nx_train, x_test = x_train \/ 255.0, x_test \/ 255.0\n\n# \u6784\u5efa\u6a21\u578b\nmodel = Sequential([\n    tf.keras.layers.Flatten(input_shape=(28, 28)),\n    Dense(128, activation='relu'),\n    Dense(10, activation='softmax')\n])\n\n# \u7f16\u8bd1\u6a21\u578b\nmodel.compile(optimizer='adam',\n              loss='sparse_categorical_crossentropy',\n              metrics=['accuracy'])\n\n# \u8bad\u7ec3\u6a21\u578b\nmodel.fit(x_train, y_train, epochs=5)\n<\/code><\/pre>\n\n\n\n<p>\u3053\u306e\u4f8b\u3067\u306f\u3001MLflow \u306e\u81ea\u52d5\u8a18\u9332\u6a5f\u80fd\u3092\u6709\u52b9\u306b\u3059\u308b\u305f\u3081\u306b\u4f7f\u7528\u3057\u307e\u3059\u3002 \u6b21\u306b\u3001\u5358\u7d14\u306a TensorFlow \u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u3057\u3066\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u307e\u3057\u305f\u3002 MLflow \u306f\u3001\u5b9f\u9a13\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u6210\u679c\u7269\u3092\u81ea\u52d5\u7684\u306b\u8a18\u9332\u3057\u307e\u3059\u3002<code>mlflow.tensorflow.autolog()<\/code><\/p>\n\n\n\n<p><strong>2. PyTorch \u3068\u7d71\u5408\u3059\u308b<\/strong><\/p>\n\n\n\n<p>MLflow \u3092 PyTorch \u3068\u7d71\u5408\u3059\u308b\u306b\u306f\u3001PyTorch \u3068 MLflow \u7528\u306e PyTorch \u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u307e\u305a\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066 PyTorch \u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">Copy code<code>pip install torch torchvision\n<\/code><\/pre>\n\n\n\n<p>\u6b21\u306b\u3001MLflow \u306e PyTorch \u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">cssCopy code<code>pip install mlflow[pytorch]\n<\/code><\/pre>\n\n\n\n<p>\u6b21\u306b\u3001MLflow \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u306e\u7d50\u679c\u3092 PyTorch \u30b3\u30fc\u30c9\u306b\u8a18\u9332\u3067\u304d\u307e\u3059\u3002 \u6b21\u306b\u7c21\u5358\u306a\u4f8b\u3092\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import torch\nimport torch.nn as nn\nimport torch.optim as optim\nimport torchvision\nimport torchvision.transforms as transforms\nimport mlflow\nimport mlflow.pytorch\n\n# \u51c6\u5907\u6570\u636e\ntransform = transforms.Compose([\n    transforms.ToTensor(),\n    transforms.Normalize((0.5,), (0.5,))\n])\n\ntrainset = torchvision.datasets.MNIST(root='.\/data', train=True, download=True, transform=transform)\ntrainloader = torch.utils.data.DataLoader(trainset, batch_size=100, shuffle=True)\n\n# \u6784\u5efa\u6a21\u578b\nclass Net(nn.Module):\n    def __init__(self):\n        super(Net, self).__init__()\n        self.fc1 = nn.Linear(28 * 28, 128)\n        self.fc2 = nn.Linear(128, 10)\n\n    def forward(self, x):\n        x = x.view(-1, 28 * 28)\n        x = torch.relu(self.fc1(x))\n       \n<\/code><\/pre>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">kotlinCopy code<code>    x = self.fc2(x)\n    return x\n<\/code><\/pre>\n\n\n\n<p>\u30e2\u30c7\u30eb = \u30cd\u30c3\u30c8()<\/p>\n\n\n\n<h1 class=\"wp-block-heading\">\u640d\u5931\u95a2\u6570\u3068\u30aa\u30d7\u30c6\u30a3\u30de\u30a4\u30b6\u306e\u5b9a\u7fa9<\/h1>\n\n\n\n<p>\u57fa\u6e96= nn\u3002CrossEntropyLoss() \u30aa\u30d7\u30c6\u30a3\u30de\u30a4\u30b6\u30fc = optim.Adam(model.parameters(), lr=0.001)<\/p>\n\n\n\n<h1 class=\"wp-block-heading\">\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3059\u308b<\/h1>\n\n\n\n<p>mlflow.start_run():\u7bc4\u56f2(5)\u306e\u30a8\u30dd\u30c3\u30af\u306e\u5834\u5408:#5\u4e2a\u30a8\u30dd\u30c3\u30afrunning_loss = 0.0 i\u306e\u5834\u5408\u3001\u5217\u6319(\u30c8\u30ec\u30a4\u30f3\u30ed\u30fc\u30c0\u30fc\u30010)\u306e\u30c7\u30fc\u30bf:\u5165\u529b\u3001\u30e9\u30d9\u30eb=\u30c7\u30fc\u30bf<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">scssCopy code<code>        optimizer.zero_grad()\n\n        outputs = model(inputs)\n        loss = criterion(outputs, labels)\n        loss.backward()\n        optimizer.step()\n\n        running_loss += loss.item()\n\n    # \u8ba1\u7b97\u5e76\u8bb0\u5f55\u6bcf\u4e2aepoch\u7684\u5e73\u5747\u635f\u5931\n    avg_loss = running_loss \/ (i + 1)\n    mlflow.log_metric(\"loss\", avg_loss, step=epoch)\n    print(f\"Epoch: {epoch + 1}, Loss: {avg_loss:.4f}\")\n\n# \u4fdd\u5b58 PyTorch \u6a21\u578b\nmlflow.pytorch.log_model(model, \"model\")\n<\/code><\/pre>\n\n\n\n<pre class=\"wp-block-preformatted\">goCopy code<code>\n\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u4f7f\u7528 `mlflow.start_run()` \u521b\u5efa\u4e86\u4e00\u4e2a\u65b0\u7684 MLflow \u8fd0\u884c\uff0c\u5e76\u5728\u8bad\u7ec3\u5faa\u73af\u5185\u90e8\u8bb0\u5f55\u4e86\u6bcf\u4e2a epoch \u7684\u635f\u5931\u3002\u5728\u8bad\u7ec3\u7ed3\u675f\u540e\uff0c\u6211\u4eec\u4f7f\u7528 `mlflow.pytorch.log_model()` \u5c06\u8bad\u7ec3\u597d\u7684 PyTorch \u6a21\u578b\u4fdd\u5b58\u5230 MLflow\u3002\n\n\u901a\u8fc7\u4ee5\u4e0a\u7684\u793a\u4f8b\uff0c\u60a8\u53ef\u4ee5\u4e86\u89e3\u5982\u4f55\u5c06 MLflow \u4e0e TensorFlow \u548c PyTorch \u96c6\u6210\u3002\u65e0\u8bba\u60a8\u4f7f\u7528\u54ea\u79cd\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\uff0cMLflow \u90fd\u53ef\u4ee5\u5e2e\u52a9\u60a8\u8f7b\u677e\u8bb0\u5f55\u548c\u7ba1\u7406\u5b9e\u9a8c\u6570\u636e\uff0c\u63d0\u9ad8\u5de5\u4f5c\u6548\u7387\u3002\n<\/code><\/pre>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u4e0a\u8a18\u306e\u4f8b\u3067\u306f\u3001MLflow \u3092 TensorFlow \u304a\u3088\u3073 PyTorch \u3068\u7d71\u5408\u3059\u308b\u65b9\u6cd5\u3092\u793a\u3057\u307e\u3059\u3002 \u6b21\u306b\u3001MLflow \u306e\u4ed6\u306e\u6a5f\u80fd\u3092\u6d3b\u7528\u3057\u3066\u3001\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0 \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u7ba1\u7406\u52b9\u7387\u3092\u5411\u4e0a\u3055\u305b\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1. MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u30c7\u30a3\u30fc\u30d7\u30e9\u30fc\u30cb\u30f3\u30b0\u5b9f\u9a13\u3092\u7ba1\u7406\u3059\u308b<\/strong><\/p>\n\n\n\n<p>TensorFlow \u304a\u3088\u3073 PyTorch \u3068\u7d71\u5408\u3057\u305f\u5f8c\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f7f\u7528\u3057\u3066\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0\u5b9f\u9a13\u3092\u7ba1\u7406\u3067\u304d\u307e\u3059\u3002 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306f\u3001\u30b3\u30fc\u30c9\u3092\u6574\u7406\u3057\u3066\u5b9f\u884c\u3059\u308b\u305f\u3081\u306e\u6a19\u6e96\u7684\u306a\u65b9\u6cd5\u3067\u3059\u3002 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u3059\u3079\u3066\u306e\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u540c\u3058\u4f9d\u5b58\u95a2\u4fc2\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30b3\u30fc\u30c9\u69cb\u9020\u3092\u4f7f\u7528\u3059\u308b\u3088\u3046\u306b\u3057\u3001\u5b9f\u9a13\u306e\u518d\u73fe\u3068\u5171\u6709\u3092\u7c21\u7565\u5316\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<p>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f5c\u6210\u306f\u7c21\u5358\u3067\u3059\u3002 \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u540d\u524d\u3001\u74b0\u5883\u69cb\u6210\u3001\u304a\u3088\u3073\u30a8\u30f3\u30c8\u30ea \u30dd\u30a4\u30f3\u30c8\u3092\u542b\u3080\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30eb\u30fc\u30c8\u306e\u4e0b\u306b\u540d\u524d\u306e\u4ed8\u3044\u305f\u30d5\u30a1\u30a4\u30eb\u3092\u4f5c\u6210\u3059\u308b\u3060\u3051\u3067\u3059\u3002 \u4f8b\u3048\u3070\uff1a<code>MLproject<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">yamlCopy code<code>name: my-deep-learning-project\n\nconda_env: conda.yaml\n\nentry_points:\n  main:\n    parameters:\n      epochs: int\n      batch_size: int\n    command: \"python train.py --epochs {epochs} --batch_size {batch_size}\"\n<\/code><\/pre>\n\n\n\n<p>\u6b21\u306b\u3001\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0\u5b9f\u9a13\u306e\u30b3\u30fc\u30c9\u3068\u30c7\u30fc\u30bf\u3092\u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30d5\u30a9\u30eb\u30c0\u30fc\u306b\u683c\u7d0d\u3057\u3001\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002 \u4f8b\u3048\u3070\uff1a<code>mlflow run<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">cssCopy code<code>mlflow run . -P epochs=5 -P batch_size=100\n<\/code><\/pre>\n\n\n\n<p><strong>2. MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3057\u3066\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0 \u30e2\u30c7\u30eb\u3092\u7ba1\u7406\u3059\u308b<\/strong><\/p>\n\n\n\n<p>TensorFlow \u304a\u3088\u3073 PyTorch \u3068\u306e\u7d71\u5408\u306b\u3088\u308a\u3001MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3057\u3066\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0 \u30e2\u30c7\u30eb\u3092\u7ba1\u7406\u3059\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306f\u3001\u30e2\u30c7\u30eb\u3092\u683c\u7d0d\u3001\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u3001\u304a\u3088\u3073\u7ba1\u7406\u3059\u308b\u305f\u3081\u306e\u4e00\u5143\u7684\u306a\u5834\u6240\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30e2\u30c7\u30eb\u306e\u5c65\u6b74\u306e\u8ffd\u8de1\u3001\u30e2\u30c7\u30eb\u306e\u65b0\u3057\u3044\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u30c7\u30d7\u30ed\u30a4\u3001\u5171\u540c\u4f5c\u696d\u304c\u3088\u308a\u7c21\u5358\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6e08\u307f\u306e\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0 \u30e2\u30c7\u30eb\u3092 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306b\u767b\u9332\u3059\u308b\u306b\u306f\u3001\u95a2\u6570\u3092\u4f7f\u7528\u3059\u308b\u3060\u3051\u3067\u3059\u3002 \u4f8b\u3048\u3070\uff1a<code>mlflow.register_model()<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>result = mlflow.register_model(\"runs:\/&lt;run_id&gt;\/model\", \"my-deep-learning-model\")\n<\/code><\/pre>\n\n\n\n<p>\u305d\u306e\u5f8c\u3001MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306e Web UI \u307e\u305f\u306f API \u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb\u3092\u8868\u793a\u3001\u7ba1\u7406\u3001\u30c7\u30d7\u30ed\u30a4\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u306f TensorFlow \u3084 PyTorch \u306a\u3069\u306e\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0 \u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3068\u306e\u7d71\u5408\u3092\u63d0\u4f9b\u3057\u3001\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0\u5b9f\u9a13\u306e\u8a18\u9332\u3001\u7ba1\u7406\u3001\u5171\u6709\u3092\u5bb9\u6613\u306b\u3057\u307e\u3059\u3002 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3068 MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u30c7\u30a3\u30fc\u30d7 \u30e9\u30fc\u30cb\u30f3\u30b0 \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3067\u306e\u7ba1\u7406\u52b9\u7387\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u3055\u3089\u306b\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow\u3068Jupyter\u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u306e\u4f7f\u3044\u65b9\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u306f Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3068\u30b7\u30fc\u30e0\u30ec\u30b9\u306b\u7d71\u5408\u3055\u308c\u3066\u3044\u308b\u305f\u3081\u3001\u5bfe\u8a71\u578b\u74b0\u5883\u3067\u5b9f\u9a13\u7d50\u679c\u3092\u7c21\u5358\u306b\u8a18\u9332\u304a\u3088\u3073\u8868\u793a\u3067\u304d\u307e\u3059\u3002 Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3067 MLflow \u3092\u4f7f\u7528\u3059\u308b\u65b9\u6cd5\u306e\u8a73\u7d30\u306a\u624b\u9806\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1. Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3068 MLflow \u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b<\/strong><\/p>\n\n\n\n<p>Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3068 MLflow \u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002 \u307e\u3060\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u306a\u3044\u5834\u5408\u306f\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">Copy code<code>pip install jupyter\npip install mlflow\n<\/code><\/pre>\n\n\n\n<p><strong>2. \u30b8\u30e5\u30d4\u30bf\u30fc\u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3092\u8d77\u52d5\u3057\u307e\u3059<\/strong><\/p>\n\n\n\n<p>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3067 Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3092\u8d77\u52d5\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">Copy code<code>jupyter notebook\n<\/code><\/pre>\n\n\n\n<p>\u6b21\u306b\u3001\u65b0\u3057\u3044\u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3092\u4f5c\u6210\u3057\u3001\u5fc5\u8981\u306a\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u30a4\u30f3\u30dd\u30fc\u30c8\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow\nimport mlflow.sklearn\n<\/code><\/pre>\n\n\n\n<p><strong>3. MLflow \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u306e\u7d50\u679c\u3092\u8a18\u9332\u3059\u308b<\/strong><\/p>\n\n\n\n<p>Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3067\u306f\u3001\u901a\u5e38\u306e Python \u30b9\u30af\u30ea\u30d7\u30c8\u3068\u540c\u3058\u3088\u3046\u306b MLflow \u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001scikit-learn \u3092\u4f7f\u7528\u3057\u305f\u5358\u7d14\u306a\u7dda\u5f62\u56de\u5e30\u306e\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import numpy as np\nfrom sklearn.linear_model import LinearRegression\nfrom sklearn.metrics import mean_squared_error\n\n# \u51c6\u5907\u6570\u636e\nX = np.array([[1, 1], [1, 2], [2, 2], [2, 3]])\ny = np.dot(X, np.array([1, 2])) + 3\n\n# \u8bad\u7ec3\u6a21\u578b\nmodel = LinearRegression().fit(X, y)\n\n# \u9884\u6d4b\ny_pred = model.predict(X)\n\n# \u8ba1\u7b97\u6027\u80fd\u6307\u6807\nmse = mean_squared_error(y, y_pred)\n\n# \u4f7f\u7528 MLflow \u8bb0\u5f55\u5b9e\u9a8c\u7ed3\u679c\nwith mlflow.start_run():\n    mlflow.log_param(\"model_type\", \"LinearRegression\")\n    mlflow.log_metric(\"mse\", mse)\n    mlflow.sklearn.log_model(model, \"model\")\n<\/code><\/pre>\n\n\n\n<p><strong>4. MLflow UI \u3092\u8d77\u52d5\u3059\u308b<\/strong><\/p>\n\n\n\n<p>\u30b3\u30de\u30f3\u30c9 \u30e9\u30a4\u30f3\u304b\u3089\u30d7\u30ed\u30b8\u30a7\u30af\u30c8 \u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u306b\u79fb\u52d5\u3057\u3001MLflow UI \u3092\u8d77\u52d5\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">Copy code<code>mlflow ui\n<\/code><\/pre>\n\n\n\n<p>\u65e2\u5b9a\u3067\u306f\u3001MLflow UI \u306f\u30dd\u30fc\u30c8 5000 \u3067\u958b\u59cb\u3055\u308c\u307e\u3059\u3002 \u30d6\u30e9\u30a6\u30b6\u30fc\u3067\u958b\u304d\u3001\u5b9f\u9a13\u306e\u7d50\u679c\u3092\u8868\u793a\u3057\u307e\u3059\u3002<code>http:\/\/localhost:5000<\/code><\/p>\n\n\n\n<p><strong>5. \u5b9f\u9a13\u7d50\u679c\u306e\u30af\u30a8\u30ea\u3092\u5b9f\u884c\u3059\u308b<\/strong><\/p>\n\n\n\n<p>Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3067\u306f\u3001MLflow API \u3092\u4f7f\u7528\u3057\u3066\u5b9f\u9a13\u7d50\u679c\u3092\u30af\u30a8\u30ea\u3067\u304d\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001(MSE \u306b\u57fa\u3065\u3044\u3066)\u6700\u9069\u306a\u30e2\u30c7\u30eb\u3092\u53d6\u5f97\u3059\u308b\u65b9\u6cd5\u306e\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">pythonCopy code<code>import mlflow\nimport mlflow.sklearn\n\n# \u67e5\u8be2\u5b9e\u9a8c\u7ed3\u679c\nbest_run = mlflow.search_runs(order_by=[\"metric.mse ASC\"]).iloc[0]\n\n# \u52a0\u8f7d\u6700\u4f73\u6a21\u578b\nbest_model = mlflow.sklearn.load_model(\"runs:\/{}\/model\".format(best_run.run_id))\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306f\u3001Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3067 MLflow \u3092\u4f7f\u7528\u3059\u308b\u57fa\u672c\u7684\u306a\u30d7\u30ed\u30bb\u30b9\u3067\u3059\u3002 MLflow \u306f Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3068\u7c21\u5358\u306b\u7d71\u5408\u3067\u304d\u308b\u305f\u3081\u3001\u5b9f\u9a13\u7d50\u679c\u3092\u8a18\u9332\u3057\u305f\u308a\u3001\u5b9f\u9a13\u7d50\u679c\u3092\u8868\u793a\u3057\u305f\u308a\u3001\u4fdd\u5b58\u3055\u308c\u305f\u30e2\u30c7\u30eb\u3092\u5bfe\u8a71\u578b\u74b0\u5883\u3067\u8aad\u307f\u8fbc\u3093\u3060\u308a\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u5b9f\u9a13\u3092\u3088\u308a\u9069\u5207\u306b\u6574\u7406\u304a\u3088\u3073\u7ba1\u7406\u3057\u3001\u751f\u7523\u6027\u3092\u9ad8\u3081\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>AWS SageMaker\u3001Azure ML\u306a\u3069\u306e\u30af\u30e9\u30a6\u30c9\u30b5\u30fc\u30d3\u30b9\u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u3068\u306e\u7d71\u5408\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u306f\u3001AWS SageMaker \u3084 Azure ML \u306a\u3069\u306e\u3055\u307e\u3056\u307e\u306a\u30af\u30e9\u30a6\u30c9\u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u3068\u7d71\u5408\u3057\u3066\u3001\u3088\u308a\u5b8c\u5168\u306a\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u3068\u3088\u308a\u5805\u7262\u306a\u30c7\u30d7\u30ed\u30a4\u30aa\u30d7\u30b7\u30e7\u30f3\u3092\u63d0\u4f9b\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u30bb\u30af\u30b7\u30e7\u30f3\u3067\u306f\u3001MLflow \u3092 AWS SageMaker \u304a\u3088\u3073 Azure ML \u3068\u305d\u308c\u305e\u308c\u7d71\u5408\u3059\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1. AWS SageMaker \u3068\u7d71\u5408\u3059\u308b<\/strong><\/p>\n\n\n\n<p>AWS SageMaker \u306f\u3001ML \u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3001\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u305f\u3081\u306e\u30d5\u30eb\u30de\u30cd\u30fc\u30b8\u30c9\u578b\u306e\u6a5f\u68b0\u5b66\u7fd2\u30b5\u30fc\u30d3\u30b9\u3067\u3059\u3002 MLflow \u3092 AWS SageMaker \u3068\u7d71\u5408\u3059\u308b\u3053\u3068\u3067\u3001MLflow \u30e2\u30c7\u30eb\u3092 SageMaker \u306b\u7c21\u5358\u306b\u30c7\u30d7\u30ed\u30a4\u3067\u304d\u307e\u3059\u3002 MLflow \u30e2\u30c7\u30eb\u3092 SageMaker \u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u305f\u3081\u306e\u57fa\u672c\u7684\u306a\u624b\u9806\u306f\u6b21\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u307e\u305a\u3001\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u6e08\u307f\u3067\u3042\u308a\u3001(AWS SDK for Python): \u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3059\u308b<code>mlflow<\/code><code>boto3<\/code><code>pip install mlflow pip install boto3<\/code><\/li>\n\n\n\n<li>AWS \u30a2\u30af\u30bb\u30b9\u30ad\u30fc\u3092\u8a2d\u5b9a\u3057\u307e\u3059\u3002 AWS \u30b3\u30f3\u30bd\u30fc\u30eb\u3067 IAM \u30e6\u30fc\u30b6\u30fc\u3092\u4f5c\u6210\u3057\u3001\u9069\u5207\u306a\u30a2\u30af\u30bb\u30b9\u8a31\u53ef\u3092\u4ed8\u4e0e\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u6b21\u306b\u3001\u30a2\u30af\u30bb\u30b9\u30ad\u30fc\u3068\u30b7\u30fc\u30af\u30ec\u30c3\u30c8\u30a2\u30af\u30bb\u30b9\u30ad\u30fc\u3092\u74b0\u5883\u5909\u6570\u306b\u8a2d\u5b9a\u3057\u307e\u3059\u3002<code>export AWS_ACCESS_KEY_ID=&lt;your_access_key&gt; export AWS_SECRET_ACCESS_KEY=&lt;your_secret_key&gt;<\/code><\/li>\n\n\n\n<li>Python \u30b9\u30af\u30ea\u30d7\u30c8\u307e\u305f\u306f Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3067\u3001sageMaker \u306b\u65b0\u3057\u3044\u30a8\u30f3\u30c9\u30dd\u30a4\u30f3\u30c8\u3092\u4f5c\u6210\u3057\u3001MLflow \u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b module:pythonCopy \u30b3\u30fc\u30c9\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092 SageMaker \u306b\u30c7\u30d7\u30ed\u30a4\u3057\u307e\u3059\u3002 \u30c7\u30d7\u30ed\u30a4\u3055\u308c\u305f\u30e2\u30c7\u30eb\u3068\u30a8\u30f3\u30c9\u30dd\u30a4\u30f3\u30c8\u306f\u3001AWS \u30b3\u30f3\u30bd\u30fc\u30eb\u306e SageMaker \u30b5\u30fc\u30d3\u30b9\u3067\u8868\u793a\u3067\u304d\u307e\u3059\u3002<code>mlflow.sagemaker<\/code><code>import mlflow.sagemaker as mfs # \u6307\u5b9a\u6a21\u578b\u7684 URI\uff08\u4f8b\u5982\uff1as3:\/\/your_bucket\/your_model_path\uff09 model_uri = \"runs:\/&lt;run_id&gt;\/model\" # \u90e8\u7f72\u6a21\u578b mfs.deploy(app_name=\"mlflow-sagemaker-example\", model_uri=model_uri, execution_role_arn=\"arn:aws:iam::&lt;account_id&gt;:role\/service-role\/AmazonSageMaker-ExecutionRole\", region_name=\"us-west-2\", mode=mfs.DeploymentMode.CREATE)<\/code><\/li>\n\n\n\n<li>\u30c7\u30d7\u30ed\u30a4\u3055\u308c\u305f\u30e2\u30c7\u30eb\u3092\u547c\u3073\u51fa\u3059\u306b\u306f\u3001(AWS SDK for Python) :p ythonCopy \u30b3\u30fc\u30c9\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<code>boto3<\/code><code>import boto3 # \u521b\u5efa SageMaker \u8fd0\u884c\u65f6\u5ba2\u6237\u7aef client = boto3.client(\"sagemaker-runtime\", region_name=\"us-west-2\") # \u51c6\u5907\u8f93\u5165\u6570\u636e input_data = ... # \u6309\u7167\u6a21\u578b\u6240\u9700\u683c\u5f0f\u51c6\u5907\u8f93\u5165\u6570\u636e # \u8c03\u7528\u6a21\u578b response = client.invoke_endpoint(EndpointName=\"mlflow-sagemaker-example\", ContentType=\"application\/json\", Body=input_data) # \u89e3\u6790\u8f93\u51fa\u7ed3\u679c output_data = response[\"Body\"].read().decode(\"utf-8\")<\/code><\/li>\n<\/ul>\n\n\n\n<p><strong>2. Azure ML \u3092\u7d71\u5408\u3059\u308b<\/strong><\/p>\n\n\n\n<p>Azure ML \u306f\u3001ML \u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3001\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u305f\u3081\u306e\u30c4\u30fc\u30eb\u3068\u30b5\u30fc\u30d3\u30b9\u3092\u63d0\u4f9b\u3059\u308b\u30af\u30e9\u30a6\u30c9\u4e0a\u306e\u6a5f\u68b0\u5b66\u7fd2\u30b5\u30fc\u30d3\u30b9\u3067\u3059\u3002 MLflow \u3092 Azure ML \u3068\u7d71\u5408\u3059\u308b\u3053\u3068\u3067\u3001\u6b21\u306e\u3053\u3068\u304c\u53ef\u80fd\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u306f\u3001Azure ML \u30ef\u30fc\u30af\u30b9\u30da\u30fc\u30b9\u3067\u7c21\u5358\u306b\u7ba1\u7406\u304a\u3088\u3073\u30c7\u30d7\u30ed\u30a4\u3067\u304d\u307e\u3059\u3002 MLflow \u30e2\u30c7\u30eb\u3092 Azure ML \u306b\u7d71\u5408\u3059\u308b\u305f\u3081\u306e\u57fa\u672c\u7684\u306a\u624b\u9806\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u307e\u305a\u3001 :Copy \u30b3\u30fc\u30c9\u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059<code>mlflow<\/code><code>azureml-sdk<\/code><code>pip install mlflow pip install azureml-sdk<\/code><\/li>\n\n\n\n<li>Azure \u30dd\u30fc\u30bf\u30eb\u3067 Azure ML \u30ef\u30fc\u30af\u30b9\u30da\u30fc\u30b9\u3092\u4f5c\u6210\u3057\u3001\u30b5\u30d6\u30b9\u30af\u30ea\u30d7\u30b7\u30e7\u30f3 ID\u3001\u30ea\u30bd\u30fc\u30b9 \u30b0\u30eb\u30fc\u30d7\u3001\u30ef\u30fc\u30af\u30b9\u30da\u30fc\u30b9\u540d\u3092\u53d6\u5f97\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>Python \u30b9\u30af\u30ea\u30d7\u30c8\u307e\u305f\u306f Jupyter \u30ce\u30fc\u30c8\u30d6\u30c3\u30af\u3067\u3001\u6b21\u306e\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066 Azure ML \u30ef\u30fc\u30af\u30b9\u30da\u30fc\u30b9\u306b\u63a5\u7d9a\u3057\u307e\u3059\u3002<code>azureml.core<\/code><code>from azureml.core import Workspace subscription_id = \"&lt;your_subscription_id&gt;\" resource_group = \"&lt;your_resource_group&gt;\" workspace_name = \"&lt;your_workspace_name&gt;\" ws = Workspace.get(name=workspace_name, subscription_id=subscription_id, resource_group=resource_group)<\/code><\/li>\n\n\n\n<li>\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066 MLflow \u30e2\u30c7\u30eb\u3092 Azure ML \u306b\u767b\u9332\u3057\u307e\u3059: pythonCopy \u30b3\u30fc\u30c9\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001Azure ML \u30ef\u30fc\u30af\u30b9\u30da\u30fc\u30b9\u306b\u65b0\u3057\u3044\u767b\u9332\u6e08\u307f\u30e2\u30c7\u30eb\u304c\u4f5c\u6210\u3055\u308c\u307e\u3059\u3002 \u767b\u9332\u6e08\u307f\u306e\u30e2\u30c7\u30eb\u306f\u3001Azure \u30dd\u30fc\u30bf\u30eb\u306e Azure ML \u30b5\u30fc\u30d3\u30b9\u3067\u8868\u793a\u3067\u304d\u307e\u3059\u3002<code>mlflow.azureml<\/code><code>import mlflow.azureml model_uri = \"runs:\/&lt;run_id&gt;\/model\" model_name = \"my_mlflow_model\" registered_model = mlflow.azureml.register_model(ws, model_name=model_name, model_uri=model_uri, description=\"MLflow model example\")<\/code><\/li>\n\n\n\n<li>\u767b\u9332\u6e08\u307f\u306e MLflow \u30e2\u30c7\u30eb\u3092 Azure Kubernetes Service (AKS) \u307e\u305f\u306f Azure Container Instances (ACI) \u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u306b\u306f\u3001\u30e2\u30b8\u30e5\u30fc\u30eb\u3068\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002 ACI \u3078\u306e\u30c7\u30d7\u30ed\u30a4\u306e\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059: python\u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3059\u308b\u3053\u308c\u306b\u3088\u308a\u3001MLflow \u30e2\u30c7\u30eb\u304c Azure ML \u306b\u30c7\u30d7\u30ed\u30a4\u3055\u308c\u307e\u3059\u3002 \u30c7\u30d7\u30ed\u30a4\u3055\u308c\u305f\u30e2\u30c7\u30eb\u3068 Web \u30b5\u30fc\u30d3\u30b9\u306f\u3001Azure \u30dd\u30fc\u30bf\u30eb\u306e Azure ML Services \u3067\u8868\u793a\u3067\u304d\u307e\u3059\u3002<code>azureml.core<\/code><code>azureml.mlflow<\/code><code>from azureml.core import Environment from azureml.core.webservice import AciWebservice, Webservice from azureml.core.model import InferenceConfig, Model from azureml.mlflow import get_mlflow import mlflow.azureml # \u521b\u5efa\u63a8\u7406\u73af\u5883 env = get_mlflow(ws, model_name) # \u521b\u5efa\u63a8\u7406\u914d\u7f6e inference_config = InferenceConfig(entry_script=\"score.py\", environment=env) # \u914d\u7f6e\u90e8\u7f72\u76ee\u6807\uff08\u5728\u672c\u4f8b\u4e2d\u4e3a ACI\uff09 deployment_config = AciWebservice.deploy_configuration(cpu_cores=1, memory_gb=1) # \u90e8\u7f72\u6a21\u578b service = Model.deploy(ws, \"my-aci-service\", [registered_model], inference_config, deployment_config) service.wait_for_deployment(show_output=True)<\/code><\/li>\n\n\n\n<li>\u30c7\u30d7\u30ed\u30a4\u3055\u308c\u305f\u30e2\u30c7\u30eb\u3092\u547c\u3073\u51fa\u3059\u306b\u306f\u3001\u30e2\u30b8\u30e5\u30fc\u30eb\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059: pythonCopy \u30b3\u30fc\u30c9<code>azureml.core.webservice<\/code><code>import json # \u51c6\u5907\u8f93\u5165\u6570\u636e input_data = ... # \u6309\u7167\u6a21\u578b\u6240\u9700\u683c\u5f0f\u51c6\u5907\u8f93\u5165\u6570\u636e input_json = json.dumps({\"data\": input_data}) # \u8c03\u7528\u6a21\u578b output_json = service.run(input_json) # \u89e3\u6790\u8f93\u51fa\u7ed3\u679c output_data = json.loads(output_json)<\/code><\/li>\n<\/ul>\n\n\n\n<p>MLflow \u3092\u7d44\u307f\u8fbc\u3080\u3053\u3068\u3067<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>AWS SageMaker \u304a\u3088\u3073 Azure ML \u3068\u7d71\u5408\u3059\u308b\u3053\u3068\u3067\u3001\u81ea\u52d5\u5316\u3055\u308c\u305f\u30e2\u30c7\u30eb\u306e\u30c7\u30d7\u30ed\u30a4\u3001\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3001\u7ba1\u7406\u306a\u3069\u3001\u30af\u30e9\u30a6\u30c9\u30b5\u30fc\u30d3\u30b9\u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u304c\u63d0\u4f9b\u3059\u308b\u6a5f\u80fd\u3092\u5229\u7528\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u57fa\u306b\u306a\u308b\u30a4\u30f3\u30d5\u30e9\u30b9\u30c8\u30e9\u30af\u30c1\u30e3\u3084\u30c7\u30d7\u30ed\u30a4\u3067\u306f\u306a\u304f\u3001\u30e2\u30c7\u30eb\u306e\u69cb\u7bc9\u3068\u6700\u9069\u5316\u306b\u96c6\u4e2d\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u3055\u3089\u306b\u3001MLflow \u306f\u3001Google Cloud AI Platform \u306a\u3069\u306e\u4ed6\u306e\u30af\u30e9\u30a6\u30c9 \u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u3068\u7d71\u5408\u3059\u308b\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002 \u9078\u629e\u3057\u305f\u30af\u30e9\u30a6\u30c9 \u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u306b MLflow \u3092\u7d71\u5408\u3059\u308b\u65b9\u6cd5\u306e\u8a73\u7d30\u306b\u3064\u3044\u3066\u306f\u3001MLflow \u306e\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u3068\u30af\u30e9\u30a6\u30c9 \u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u306e\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u3092\u53c2\u7167\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u3092\u30af\u30e9\u30a6\u30c9 \u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u3068\u7d71\u5408\u3059\u308b\u3068\u3001\u30af\u30e9\u30a6\u30c9\u4e0a\u3067\u306e\u6a5f\u68b0\u5b66\u7fd2\u30e2\u30c7\u30eb\u306e\u7ba1\u7406\u3001\u30c7\u30d7\u30ed\u30a4\u3001\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u304c\u5bb9\u6613\u306b\u306a\u308a\u307e\u3059\u3002 \u3053\u306e\u7d71\u5408\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb\u306e\u7ba1\u7406\u3068\u30c7\u30d7\u30ed\u30a4\u306e\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u304c\u5408\u7406\u5316\u3055\u308c\u3001\u751f\u7523\u6027\u304c\u5411\u4e0a\u3057\u3001\u5b9f\u9a13\u304b\u3089\u904b\u7528\u307e\u3067\u306e\u30e2\u30c7\u30eb\u306e\u9ad8\u901f\u5316\u304c\u53ef\u80fd\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u65b9\u6cd5\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u306b\u306f\u3001\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0 \u30d5\u30a7\u30fc\u30ba\u3068\u5b9f\u9a13\u30d5\u30a7\u30fc\u30ba\u304b\u3089\u5b9f\u969b\u306e\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u306b\u79fb\u52d5\u3057\u3066\u3001\u5b9f\u969b\u306e\u30c7\u30fc\u30bf\u3067\u4e88\u6e2c\u3092\u884c\u3046\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 MLflow \u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u306b\u306f\u3055\u307e\u3056\u307e\u306a\u65b9\u6cd5\u304c\u3042\u308a\u307e\u3059\u304c\u3001\u3053\u3053\u3067\u306f\u3044\u304f\u3064\u304b\u306e\u4e00\u822c\u7684\u306a\u30c7\u30d7\u30ed\u30a4\u65b9\u6cd5\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1. MLflow \u306e\u7d44\u307f\u8fbc\u307f\u30c7\u30d7\u30ed\u30a4 \u30c4\u30fc\u30eb\u3092\u4f7f\u7528\u3059\u308b<\/strong><\/p>\n\n\n\n<p>MLflow \u306b\u306f\u3001\u30e2\u30c7\u30eb\u3092 REST API \u30b5\u30fc\u30d3\u30b9\u3068\u3057\u3066\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u305f\u3081\u306e\u7d44\u307f\u8fbc\u307f\u306e\u30c7\u30d7\u30ed\u30a4 \u30c4\u30fc\u30eb\u304c\u3044\u304f\u3064\u304b\u7528\u610f\u3055\u308c\u3066\u3044\u307e\u3059\u3002 MLflow \u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u7c21\u5358\u306a\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<code>mlflow models serve<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">javascriptCopy code<code>mlflow models serve -m runs:\/&lt;run_id&gt;\/model -p 1234\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u30dd\u30fc\u30c8 1234 \u3067 REST API \u30b5\u30fc\u30d3\u30b9\u304c\u958b\u59cb\u3055\u308c\u307e\u3059\u3002 POST \u8981\u6c42\u3092\u9001\u4fe1\u3059\u308b\u3053\u3068\u3067\u3001\u30b5\u30fc\u30d3\u30b9\u3092\u547c\u3073\u51fa\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u3053\u306e\u65b9\u6cd5\u306f\u3001\u5c0f\u898f\u6a21\u306a\u5c55\u958b\u3068\u30c6\u30b9\u30c8\u306b\u306e\u307f\u9069\u3057\u3066\u304a\u308a\u3001\u904b\u7528\u74b0\u5883\u3067\u306e\u4f7f\u7528\u306f\u63a8\u5968\u3055\u308c\u307e\u305b\u3093\u3002<code>http:\/\/&lt;server_url&gt;:1234\/invocations<\/code><\/p>\n\n\n\n<p><strong>2. \u30c9\u30c3\u30ab\u30fc\u3092\u4f7f\u7528\u3057\u3066\u30c7\u30d7\u30ed\u30a4\u3059\u308b<\/strong><\/p>\n\n\n\n<p>MLflow \u3067\u306f\u3001\u4efb\u610f\u306e Docker \u5bfe\u5fdc\u74b0\u5883\u3067\u5b9f\u884c\u3067\u304d\u308b Docker \u30b3\u30f3\u30c6\u30ca\u30fc\u3078\u306e\u30e2\u30c7\u30eb\u306e\u30d1\u30c3\u30b1\u30fc\u30b8\u5316\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 MLflow \u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u30e2\u30c7\u30eb\u3092 Docker \u30b3\u30f3\u30c6\u30ca\u30fc\u306b\u30d1\u30c3\u30b1\u30fc\u30b8\u5316\u3059\u308b\u4f8b\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<code>mlflow models build-docker<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">javascriptCopy code<code>mlflow models build-docker -m runs:\/&lt;run_id&gt;\/model -n my_mlflow_model\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u3068\u3044\u3046\u540d\u524d\u306e Docker \u30a4\u30e1\u30fc\u30b8\u304c\u751f\u6210\u3055\u308c\u307e\u3059\u3002 \u305d\u306e\u5f8c\u3001\u6b21\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066 Docker \u30b3\u30f3\u30c6\u30ca\u30fc\u3092\u5b9f\u884c\u3067\u304d\u307e\u3059\u3002<code>my_mlflow_model<\/code><\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">arduinoCopy code<code>docker run -p 1234:8080 my_mlflow_model\n<\/code><\/pre>\n\n\n\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u30dd\u30fc\u30c8 1234 \u3067 REST API \u30b5\u30fc\u30d3\u30b9\u304c\u958b\u59cb\u3055\u308c\u307e\u3059\u3002 \u4ee5\u524d\u3068\u540c\u69d8\u306b\u3001POST \u8981\u6c42\u3092\u9001\u4fe1\u3059\u308b\u3053\u3068\u3067\u30b5\u30fc\u30d3\u30b9\u3092\u547c\u3073\u51fa\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<code>http:\/\/&lt;server_url&gt;:1234\/invocations<\/code><\/p>\n\n\n\n<p><strong>3. \u30af\u30e9\u30a6\u30c9 \u30b5\u30fc\u30d3\u30b9 \u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u3078\u306e\u30c7\u30d7\u30ed\u30a4<\/strong><\/p>\n\n\n\n<p>\u524d\u8ff0\u306e\u3088\u3046\u306b\u3001MLflow \u306f AWS SageMaker\u3001Azure ML\u3001Google Cloud AI Platform \u306a\u3069\u306e\u30af\u30e9\u30a6\u30c9\u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u3068\u7d71\u5408\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u30b5\u30fc\u30d3\u30b9\u306f\u3001\u591a\u304f\u306e\u5834\u5408\u3001MLflow \u30e2\u30c7\u30eb\u3092\u30b9\u30b1\u30fc\u30e9\u30d6\u30eb\u306a REST API \u30b5\u30fc\u30d3\u30b9\u3068\u3057\u3066\u7c21\u5358\u306b\u30c7\u30d7\u30ed\u30a4\u3067\u304d\u308b\u30d5\u30eb \u30de\u30cd\u30fc\u30b8\u30c9\u306e\u30c7\u30d7\u30ed\u30a4 \u30bd\u30ea\u30e5\u30fc\u30b7\u30e7\u30f3\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002 MLflow \u30e2\u30c7\u30eb\u3092 AWS SageMaker \u304a\u3088\u3073 Azure ML \u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u306f\u3001\u524d\u306e\u56de\u7b54\u3092\u53c2\u7167\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n\n<p><strong>4. Kubernetes \u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b<\/strong><\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u3092 Kubernetes \u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u3068\u3001\u62e1\u5f35\u6027\u3068\u67d4\u8edf\u6027\u306b\u512a\u308c\u305f\u30c7\u30d7\u30ed\u30a4\u304c\u53ef\u80fd\u306b\u306a\u308a\u307e\u3059\u3002 \u7c21\u7565\u5316\u3055\u308c\u305f\u5c55\u958b\u30d7\u30ed\u30bb\u30b9\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u3053\u306e\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066\u3001Docker \u30a4\u30e1\u30fc\u30b8\u3092\u4f5c\u6210\u3057\u307e\u3059 (\u524d\u8ff0\u306e\u3068\u304a\u308a)\u3002<code>mlflow models build-docker<\/code><\/li>\n\n\n\n<li>Docker \u30a4\u30e1\u30fc\u30b8\u3092 Docker Hub\u3001Google Container Registry\u3001AWS Elastic Container Registry \u306a\u3069\u306e\u30b3\u30f3\u30c6\u30ca\u30ea\u30dd\u30b8\u30c8\u30ea\u306b\u30d7\u30c3\u30b7\u30e5\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>Kubernetes \u306e\u30c7\u30d7\u30ed\u30a4\u3068\u30b5\u30fc\u30d3\u30b9 \u30ea\u30bd\u30fc\u30b9\u3092\u4f5c\u6210\u3057\u3001\u30b3\u30f3\u30c6\u30ca\u30fc \u30a4\u30e1\u30fc\u30b8\u3068\u9069\u5207\u306a\u30dd\u30fc\u30c8 \u30de\u30c3\u30d4\u30f3\u30b0\u3092\u6307\u5b9a\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30b3\u30de\u30f3\u30c9\u3092\u4f7f\u7528\u3057\u3066 Kubernetes \u30ea\u30bd\u30fc\u30b9\u3092\u30c7\u30d7\u30ed\u30a4\u3057\u307e\u3059\u3002<code>kubectl apply<\/code><\/li>\n<\/ul>\n\n\n\n<p>\u5c55\u958b<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u5834\u5408\u306f\u3001\u6b21\u306e\u8981\u7d20\u3082\u8003\u616e\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>1.\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3068\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3<\/strong><\/p>\n\n\n\n<p>\u904b\u7528\u74b0\u5883\u306b MLflow \u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u5834\u5408\u306f\u3001\u30e2\u30c7\u30eb\u304c\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u8981\u4ef6\u3092\u6e80\u305f\u3057\u3001\u30b9\u30b1\u30fc\u30e9\u30d6\u30eb\u3067\u3042\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u9069\u5207\u306a\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2 \u30ea\u30bd\u30fc\u30b9\u3001\u30c7\u30d7\u30ed\u30a4\u65b9\u6cd5\u3092\u9078\u629e\u3057\u3001\u30e2\u30c7\u30eb\u3068\u30b3\u30fc\u30c9\u3092\u6700\u9069\u5316\u3059\u308b\u3068\u3001\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u5411\u4e0a\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u30af\u30e9\u30a6\u30c9 \u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u307e\u305f\u306f Kubernetes \u306a\u3069\u306e\u30b3\u30f3\u30c6\u30ca\u30fc \u30aa\u30fc\u30b1\u30b9\u30c8\u30ec\u30fc\u30b7\u30e7\u30f3 \u30c4\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u3001\u5fc5\u8981\u306b\u5fdc\u3058\u3066\u30b5\u30fc\u30d3\u30b9\u3092\u52d5\u7684\u306b\u30b9\u30b1\u30fc\u30eb\u30a2\u30c3\u30d7\u307e\u305f\u306f\u30b9\u30b1\u30fc\u30eb\u30c0\u30a6\u30f3\u3059\u308b\u3053\u3068\u3092\u691c\u8a0e\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n\n<p><strong>2. \u76e3\u8996\u3068\u30ed\u30ae\u30f3\u30b0<\/strong><\/p>\n\n\n\n<p>\u76e3\u8996\u3068\u30ed\u30b0\u8a18\u9332\u306f\u3001\u904b\u7528\u74b0\u5883\u306b MLflow \u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u3068\u304d\u306b\u91cd\u8981\u3067\u3059\u3002 \u30c7\u30d7\u30ed\u30a4 \u30b7\u30ca\u30ea\u30aa\u3067\u3001\u8981\u6c42\u306e\u5f85\u6a5f\u6642\u9593\u3001\u30a8\u30e9\u30fc\u7387\u3001\u30b9\u30eb\u30fc\u30d7\u30c3\u30c8\u306a\u3069\u306e\u4e3b\u8981\u306a\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u304a\u3088\u3073\u30e2\u30c7\u30eb\u3068\u30b5\u30fc\u30d3\u30b9\u306e\u30ed\u30b0\u306e\u53ce\u96c6\u304c\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u6f5c\u5728\u7684\u306a\u554f\u984c\u3092\u30bf\u30a4\u30e0\u30ea\u30fc\u306b\u7279\u5b9a\u3057\u3066\u89e3\u6c7a\u3057\u3001\u512a\u308c\u305f\u30b5\u30fc\u30d3\u30b9\u54c1\u8cea\u3092\u78ba\u4fdd\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>3. \u30bb\u30ad\u30e5\u30ea\u30c6\u30a3<\/strong><\/p>\n\n\n\n<p>\u30e2\u30c7\u30eb\u30fb\u30c7\u30d7\u30ed\u30a4\u30e1\u30f3\u30c8\u306e\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3092\u78ba\u4fdd\u3059\u308b\u3053\u3068\u306f\u3001\u5b9f\u7a3c\u50cd\u74b0\u5883\u306b\u3068\u3063\u3066\u91cd\u8981\u3067\u3059\u3002 \u7279\u5b9a\u306e\u30cb\u30fc\u30ba\u306b\u5fdc\u3058\u3066\u3001\u8a8d\u8a3c\u3001\u627f\u8a8d\u3001\u30c7\u30fc\u30bf\u6697\u53f7\u5316\u306a\u3069\u306e\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u5bfe\u7b56\u3092\u5b9f\u88c5\u3059\u308b\u5fc5\u8981\u304c\u3042\u308b\u5834\u5408\u304c\u3042\u308a\u307e\u3059\u3002 \u3055\u3089\u306b\u3001GDPR\u3001HIPAA\u306a\u3069\u306e\u30b3\u30f3\u30d7\u30e9\u30a4\u30a2\u30f3\u30b9\u8981\u4ef6\u3082\u3001\u5c55\u958b\u6226\u7565\u3068\u30d7\u30e9\u30af\u30c6\u30a3\u30b9\u306b\u5f71\u97ff\u3092\u4e0e\u3048\u308b\u53ef\u80fd\u6027\u304c\u3042\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>4. \u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u3068\u7d99\u7d9a\u7684\u30a4\u30f3\u30c6\u30b0\u30ec\u30fc\u30b7\u30e7\u30f3\/\u7d99\u7d9a\u7684\u30c7\u30d7\u30ed\u30a4 (CI\/CD)<\/strong><\/p>\n\n\n\n<p>MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u3067\u5e38\u306b\u6700\u65b0\u306e\u72b6\u614b\u306b\u4fdd\u3064\u306b\u306f\u3001\u30e2\u30c7\u30eb\u3092 Git \u306a\u3069\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u30b7\u30b9\u30c6\u30e0\u3068\u7d71\u5408\u3057\u3001Jenkins\u3001Travis CI\u3001GitHub Actions \u306a\u3069\u306e\u7d99\u7d9a\u7684\u30a4\u30f3\u30c6\u30b0\u30ec\u30fc\u30b7\u30e7\u30f3\/\u7d99\u7d9a\u7684\u30c7\u30d7\u30ed\u30a4 (CI\/CD) \u30c4\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u30c7\u30d7\u30ed\u30a4 \u30d7\u30ed\u30bb\u30b9\u3092\u81ea\u52d5\u5316\u3059\u308b\u3053\u3068\u3092\u304a\u52e7\u3081\u3057\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb\u66f4\u65b0\u306e\u9069\u6642\u6027\u3068\u4e00\u8cab\u6027\u304c\u78ba\u4fdd\u3055\u308c\u3001\u4eba\u70ba\u7684\u30df\u30b9\u306e\u53ef\u80fd\u6027\u304c\u6e1b\u5c11\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u306b\u306f\u3001\u3044\u304f\u3064\u304b\u306e\u8003\u616e\u4e8b\u9805\u304c\u3042\u308a\u307e\u3059\u3002 \u9069\u5207\u306a\u30c7\u30d7\u30ed\u30a4 \u30a2\u30d7\u30ed\u30fc\u30c1\u3092\u9078\u629e\u3057\u3001\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3068\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3\u3092\u78ba\u4fdd\u3057\u3001\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3068\u76e3\u8996\u306b\u91cd\u70b9\u3092\u7f6e\u304d\u3001\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u3068 CI\/CD \u30c4\u30fc\u30eb\u3092\u6d3b\u7528\u3059\u308b\u3053\u3068\u3067\u3001MLflow \u30e2\u30c7\u30eb\u3092\u5b9f\u969b\u306e\u30d3\u30b8\u30cd\u30b9 \u30b7\u30ca\u30ea\u30aa\u306b\u3046\u307e\u304f\u9069\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u672c\u756a\u74b0\u5883\u3067\u306eMLflow\u306e\u30d9\u30b9\u30c8\u30d7\u30e9\u30af\u30c6\u30a3\u30b9\u3092\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u904b\u7528\u74b0\u5883\u3067 MLflow \u3092\u4f7f\u7528\u3059\u308b\u305f\u3081\u306e\u30d9\u30b9\u30c8 \u30d7\u30e9\u30af\u30c6\u30a3\u30b9\u306f\u3001\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u3092\u3088\u308a\u52b9\u679c\u7684\u306b\u7ba1\u7406\u3057\u3001\u30b5\u30fc\u30d3\u30b9\u54c1\u8cea\u3092\u78ba\u4fdd\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u3053\u3053\u306b\u3044\u304f\u3064\u304b\u306e\u91cd\u8981\u306a\u30d9\u30b9\u30c8\u30d7\u30e9\u30af\u30c6\u30a3\u30b9\u304c\u3042\u308a\u307e\u3059:<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u30e2\u30c7\u30eb<strong>\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u3068\u8ffd\u8de1<\/strong>: MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3068\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u7ba1\u7406\u3057\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u3054\u3068\u306b\u500b\u5225\u306e\u30ec\u30b8\u30b9\u30c8\u30ea \u30a8\u30f3\u30c8\u30ea\u3092\u4f5c\u6210\u3057\u3001\u91cd\u8981\u306a\u66f4\u65b0\u30d7\u30ed\u30b0\u30e9\u30e0\u3054\u3068\u306b\u65b0\u3057\u3044\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u4f5c\u6210\u3057\u3066\u304f\u3060\u3055\u3044\u3002 \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u3055\u307e\u3056\u307e\u306a\u30e2\u30c7\u30eb \u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9 \u30e1\u30c8\u30ea\u30c3\u30af\u3092\u8ffd\u8de1\u3057\u3001\u5fc5\u8981\u306b\u5fdc\u3058\u3066\u53e4\u3044\u30d0\u30fc\u30b8\u30e7\u30f3\u306b\u30ed\u30fc\u30eb\u30d0\u30c3\u30af\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u5b9f\u9a13\u306e\u660e\u78ba\u306a\u7de8\u6210<\/strong>: MLflow \u3067\u5b9f\u9a13\u3092\u6574\u7406\u3057\u3066\u3001\u5404\u5b9f\u9a13\u306b\u660e\u78ba\u306a\u76ee\u6a19\u3068\u30b9\u30b3\u30fc\u30d7\u304c\u3042\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002 \u7570\u306a\u308b\u30bf\u30a4\u30d7\u306e\u30e2\u30c7\u30eb\u3068\u30bf\u30b9\u30af\u3092\u6df7\u5728\u3055\u305b\u306a\u3044\u3088\u3046\u306b\u3057\u3066\u304f\u3060\u3055\u3044\u3002 \u5b9f\u9a13\u306b\u8aac\u660e\u30bf\u30b0\u3092\u8ffd\u52a0\u3057\u3066\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u7406\u89e3\u3057\u3084\u3059\u304f\u3001\u898b\u3064\u3051\u3084\u3059\u304f\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u5b8c\u5168\u306a\u5b9f\u9a13\u60c5\u5831\u3092\u8a18\u9332\u3057\u3066\u30af\u30a8\u30ea\u3059\u308b:<\/strong>&nbsp;\u30e2\u30c7\u30eb\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u30a2\u30fc\u30c6\u30a3\u30d5\u30a1\u30af\u30c8\u3092\u8a18\u9332\u3057\u3066\u3001\u5b9f\u9a13\u306e\u518d\u73fe\u6027\u3092\u78ba\u4fdd\u3057\u307e\u3059\u3002 \u5b9f\u9a13\u306b\u5bfe\u3057\u3066\u30af\u30a8\u30ea\u3092\u5b9f\u884c\u3059\u308b\u3068\u304d\u306f\u3001MLflow \u306b\u3088\u3063\u3066\u63d0\u4f9b\u3055\u308c\u308b API \u3092\u4f7f\u7528\u3057\u3066\u3001\u5b9f\u9a13\u7d50\u679c\u3092\u30d5\u30a3\u30eb\u30bf\u30fc\u51e6\u7406\u304a\u3088\u3073\u6bd4\u8f03\u3057\u3001\u6700\u9069\u306a\u30e2\u30c7\u30eb\u3092\u898b\u3064\u3051\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30b3\u30fc\u30c9\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406<\/strong>: Git \u306a\u3069\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u30b7\u30b9\u30c6\u30e0\u3067 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u30b3\u30fc\u30c9\u3092\u30db\u30b9\u30c8\u3057\u3066\u3001\u30b3\u30fc\u30c9\u306e\u8ffd\u8de1\u53ef\u80fd\u6027\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u78ba\u4fdd\u3057\u307e\u3059\u3002 \u30b3\u30fc\u30c9\u3068\u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u540c\u671f\u3055\u305b\u3066\u3001\u30c8\u30ec\u30fc\u30b9\u3068\u30c7\u30d0\u30c3\u30b0\u3092\u5bb9\u6613\u306b\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u7d99\u7d9a\u7684<strong>\u30a4\u30f3\u30c6\u30b0\u30ec\u30fc\u30b7\u30e7\u30f3\/\u7d99\u7d9a\u7684\u30c7\u30d7\u30ed\u30a4 (CI\/<\/strong>CD): MLflow \u3092 Jenkins\u3001Travis CI\u3001GitHub Actions \u306a\u3069\u306e CI\/CD \u30c4\u30fc\u30eb\u3068\u7d71\u5408\u3057\u3066\u3001\u30d3\u30eb\u30c9\u3001\u30c6\u30b9\u30c8\u3001\u30c7\u30d7\u30ed\u30a4\u306e\u30d7\u30ed\u30bb\u30b9\u3092\u81ea\u52d5\u5316\u3057\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb\u306e\u30bf\u30a4\u30e0\u30ea\u30fc\u306a\u66f4\u65b0\u3068\u4e00\u8cab\u6027\u304c\u4fdd\u8a3c\u3055\u308c\u3001\u4eba\u70ba\u7684\u30df\u30b9\u306e\u53ef\u80fd\u6027\u304c\u6e1b\u5c11\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u74b0\u5883\u306e\u5206\u96e2\u3068<strong>\u4f9d\u5b58\u95a2\u4fc2\u306e\u7ba1\u7406<\/strong>: \u4eee\u60f3\u74b0\u5883 (Conda\u3001virtualenv \u306a\u3069) \u307e\u305f\u306f\u30b3\u30f3\u30c6\u30ca\u30fc \u30c6\u30af\u30ce\u30ed\u30b8 (Docker \u306a\u3069) \u3092\u4f7f\u7528\u3057\u3066\u3001\u5b9f\u9a13\u74b0\u5883\u3068\u30c7\u30d7\u30ed\u30a4\u74b0\u5883\u3092\u78ba\u5b9f\u306b\u5206\u96e2\u3057\u307e\u3059\u3002 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f7f\u7528\u3057\u3066\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u3092\u7ba1\u7406\u3057\u3001\u307e\u305f\u306f\u30d5\u30a1\u30a4\u30eb\u3067\u4f9d\u5b58\u95a2\u4fc2\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u6307\u5b9a\u3057\u3066\u3001\u74b0\u5883\u5185\u306e\u4e00\u8cab\u6027\u3092\u78ba\u4fdd\u3057\u307e\u3059\u3002<code>conda.yaml<\/code><code>requirements.txt<\/code><\/li>\n\n\n\n<li>\u6a19\u6e96\u30e2\u30c7\u30eb\u5f62\u5f0f<strong>\u3092\u4f7f\u7528\u3059\u308b<\/strong>: \u30e2\u30c7\u30eb\u3092\u4fdd\u5b58\u3059\u308b\u3068\u304d\u306f\u3001\u6a19\u6e96\u30e2\u30c7\u30eb\u5f62\u5f0f (ONNX\u3001PMML\u3001TensorFlow SavedModel \u306a\u3069) \u3092\u4f7f\u7528\u3057\u3066\u3001\u7570\u306a\u308b\u74b0\u5883\u3084\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u9593\u3067\u30e2\u30c7\u30eb\u306e\u76f8\u4e92\u904b\u7528\u6027\u3092\u4fc3\u9032\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3068\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3<\/strong>: \u30e2\u30c7\u30eb\u3068\u30b3\u30fc\u30c9\u3092\u6700\u9069\u5316\u3057\u3066\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5411\u4e0a\u3055\u305b\u3001\u30e2\u30c7\u30eb\u306e\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3\u3092\u78ba\u4fdd\u3057\u307e\u3059\u3002 \u904b\u7528\u74b0\u5883\u306b\u5c55\u958b\u3059\u308b\u5834\u5408\u306f\u3001\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u8981\u4ef6\u3092\u6e80\u305f\u3059\u9069\u5207\u306a\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2 \u30ea\u30bd\u30fc\u30b9\u3068\u5c55\u958b\u65b9\u6cd5\u3092\u9078\u629e\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3068\u30b3\u30f3\u30d7\u30e9\u30a4\u30a2\u30f3\u30b9<\/strong>:\u30e2\u30c7\u30eb\u30c7\u30d7\u30ed\u30a4\u306e\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3092\u78ba\u4fdd\u3057\u3001\u9069\u5207\u306a\u8a8d\u8a3c\u3092\u5b9f\u88c5\u3057\u3001<\/li>\n<\/ol>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u627f\u8a8d\u304a\u3088\u3073\u30c7\u30fc\u30bf\u6697\u53f7\u5316\u5bfe\u7b56\u3002 \u696d\u754c\u3068\u5730\u57df\u306b\u5fdc\u3058\u3066\u3001\u95a2\u9023\u3059\u308b\u898f\u5236\u304a\u3088\u3073\u30b3\u30f3\u30d7\u30e9\u30a4\u30a2\u30f3\u30b9\u8981\u4ef6(GDPR\u3001HIPAA\u306a\u3069)\u306b\u5f93\u3063\u3066\u304f\u3060\u3055\u3044\u3002 MLflow \u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u3068\u304d\u306f\u3001\u30c7\u30fc\u30bf\u306e\u30d7\u30e9\u30a4\u30d0\u30b7\u30fc\u3068\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u306b\u91cd\u70b9\u3092\u7f6e\u304f\u3053\u3068\u304c\u91cd\u8981\u3067\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"10\">\n<li><strong>\u76e3\u8996\u3068\u30ed\u30b0\u8a18\u9332<\/strong>: MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u5834\u5408\u306f\u3001\u8981\u6c42\u306e\u5f85\u6a5f\u6642\u9593\u3001\u30a8\u30e9\u30fc\u7387\u3001\u30b9\u30eb\u30fc\u30d7\u30c3\u30c8\u306a\u3069\u306e\u4e3b\u8981\u306a\u30e1\u30c8\u30ea\u30c3\u30af\u3092\u76e3\u8996\u3057\u3001\u30e2\u30c7\u30eb\u3068\u30b5\u30fc\u30d3\u30b9\u306e\u30ed\u30b0\u304c\u53ce\u96c6\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u6f5c\u5728\u7684\u306a\u554f\u984c\u3092\u30bf\u30a4\u30e0\u30ea\u30fc\u306b\u7279\u5b9a\u3057\u3066\u89e3\u6c7a\u3057\u3001\u512a\u308c\u305f\u30b5\u30fc\u30d3\u30b9\u54c1\u8cea\u3092\u78ba\u4fdd\u3067\u304d\u307e\u3059\u3002 MLflow \u3092 Elasticsearch\u3001Kibana\u3001Grafana \u306a\u3069\u306e\u30ed\u30ae\u30f3\u30b0\u304a\u3088\u3073\u76e3\u8996\u30c4\u30fc\u30eb\u3068\u7d71\u5408\u3057\u3066\u3001\u30ed\u30b0\u3068\u30e1\u30c8\u30ea\u30af\u30b9\u3092\u4e00\u5143\u7ba1\u7406\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30e2\u30c7\u30eb\u306e\u691c\u8a3c\u3068\u30c6\u30b9\u30c8: \u30e2\u30c7\u30eb\u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u524d\u306e\u8a73\u7d30\u306a<strong>\u691c\u8a3c\u3068\u30c6\u30b9\u30c8<\/strong>\u3002 \u30c8\u30ec\u30fc\u30cb\u30f3\u30b0 \u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3068\u691c\u8a3c\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3067\u306e\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u691c\u8a3c\u3059\u308b\u3060\u3051\u3067\u306a\u304f\u3001\u3055\u307e\u3056\u307e\u306a\u30a8\u30c3\u30b8 \u30b1\u30fc\u30b9\u3067\u30e2\u30c7\u30eb\u304c\u3069\u306e\u3088\u3046\u306b\u5b9f\u884c\u3055\u308c\u308b\u304b\u3092\u30c6\u30b9\u30c8\u3057\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u6f5c\u5728\u7684\u306a\u554f\u984c\u3092\u767a\u751f\u524d\u306b\u7279\u5b9a\u3057\u3001\u904b\u7528\u74b0\u5883\u306e\u30ea\u30b9\u30af\u3092\u8efd\u6e1b\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30c1\u30fc\u30e0\u306e\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u77e5\u8b58\u306e\u5171\u6709<\/strong>: \u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c MLflow \u306e\u5b9f\u9a13\u3068\u30e2\u30c7\u30eb\u306b\u7c21\u5358\u306b\u30a2\u30af\u30bb\u30b9\u3057\u3066\u5171\u6709\u3067\u304d\u308b\u3088\u3046\u306b\u3057\u307e\u3059\u3002 \u5171\u6709 MLflow \u30b5\u30fc\u30d0\u30fc\u3092\u4f7f\u7528\u3057\u3066\u3001\u30c1\u30fc\u30e0 \u30e1\u30f3\u30d0\u30fc\u304c\u5b9f\u9a13\u7d50\u679c\u3068\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u8868\u793a\u3067\u304d\u308b\u3088\u3046\u306b\u3057\u307e\u3059\u3002 \u30c1\u30fc\u30e0\u9593\u3067 MLflow \u306e\u4f7f\u7528\u3092\u4fc3\u9032\u3057\u3066\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3084\u90e8\u9580\u9593\u3067\u306e\u77e5\u8b58\u306e\u5171\u6709\u3092\u4fc3\u9032\u3057\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u8981\u7d04\u3059\u308b\u3068\u3001\u4e0a\u8a18\u306e\u30d9\u30b9\u30c8 \u30d7\u30e9\u30af\u30c6\u30a3\u30b9\u306b\u5f93\u3046\u3068\u3001\u904b\u7528\u74b0\u5883\u3067 MLflow \u3092\u3088\u308a\u52b9\u679c\u7684\u306b\u4f7f\u7528\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u30e2\u30c7\u30eb\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u3084\u5b9f\u9a13\u306e\u7de8\u6210\u304b\u3089\u74b0\u5883\u306e\u7ba1\u7406\u3084\u76e3\u8996\u307e\u3067\u3001\u3053\u308c\u3089\u306e\u30d7\u30e9\u30af\u30c6\u30a3\u30b9\u306b\u3088\u308a\u3001\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u306e\u5805\u7262\u6027\u3001\u62e1\u5f35\u6027\u3001\u5b89\u5168\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLFLOW\u306e\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u6700\u9069\u5316\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u3092\u4f7f\u7528\u3059\u308b\u5834\u5408\u3001\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3092\u78ba\u4fdd\u3057\u3001\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u6700\u9069\u5316\u3059\u308b\u3053\u3068\u304c\u91cd\u8981\u3067\u3059\u3002 MLflow \u306e\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u6700\u9069\u5316\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><strong>\u5b89\u5168<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u8a8d\u8a3c\u3068\u627f\u8a8d: \u904b\u7528\u74b0\u5883\u306b MLflow \u3092\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u5834\u5408\u306f\u3001\u8a8d\u8a3c<strong>\u3068\u627f\u8a8d<\/strong>\u306e\u65b9\u6cd5\u3092\u5b9f\u88c5\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001OAuth2\u3001LDAP\u3001\u307e\u305f\u306f\u305d\u306e\u4ed6\u306e\u8a8d\u8a3c\u30b7\u30b9\u30c6\u30e0\u3092\u4f7f\u7528\u3057\u3066\u3001\u627f\u8a8d\u3055\u308c\u305f\u30e6\u30fc\u30b6\u30fc\u306e\u307f\u304c MLflow \u30b5\u30fc\u30d3\u30b9\u306b\u30a2\u30af\u30bb\u30b9\u3067\u304d\u308b\u3088\u3046\u306b\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30c7\u30fc\u30bf\u6697\u53f7\u5316<\/strong>: \u30c7\u30fc\u30bf\u306e\u30d7\u30e9\u30a4\u30d0\u30b7\u30fc\u3092\u4fdd\u8b77\u3059\u308b\u306b\u306f\u3001\u8ee2\u9001\u4e2d\u306e\u30c7\u30fc\u30bf (HTTPS \u306a\u3069) \u3068\u30b9\u30c8\u30ec\u30fc\u30b8\u5185\u306e\u30c7\u30fc\u30bf (\u30b5\u30fc\u30d0\u30fc\u5074\u306e\u6697\u53f7\u5316\u3055\u308c\u305f\u30aa\u30d6\u30b8\u30a7\u30af\u30c8 \u30b9\u30c8\u30ec\u30fc\u30b8\u306a\u3069) \u3092\u6697\u53f7\u5316\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30cd\u30c3\u30c8\u30ef\u30fc\u30af \u30bb\u30ad\u30e5\u30ea\u30c6\u30a3<\/strong>: MLflow \u30b5\u30fc\u30d3\u30b9\u3068\u3001\u30c7\u30fc\u30bf\u30d9\u30fc\u30b9\u3084\u30b9\u30c8\u30ec\u30fc\u30b8\u306a\u3069\u306e\u4ed6\u306e\u91cd\u8981\u306a\u30b5\u30fc\u30d3\u30b9\u3068\u306e\u9593\u306e\u901a\u4fe1\u304c\u4fdd\u8b77\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002 \u30cd\u30c3\u30c8\u30ef\u30fc\u30af \u30dd\u30ea\u30b7\u30fc\u3001\u30d5\u30a1\u30a4\u30a2\u30a6\u30a9\u30fc\u30eb\u898f\u5247\u306a\u3069\u3092\u4f7f\u7528\u3057\u3066\u3001\u30a2\u30af\u30bb\u30b9\u3092\u5236\u9650\u3057\u3001\u4e0d\u6b63\u30a2\u30af\u30bb\u30b9\u3092\u9632\u6b62\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30b3\u30f3\u30d7\u30e9\u30a4\u30a2\u30f3\u30b9<\/strong>: \u696d\u754c\u3068\u5730\u57df\u306b\u5fdc\u3058\u3066\u3001\u95a2\u9023\u3059\u308b\u898f\u5236\u304a\u3088\u3073\u30b3\u30f3\u30d7\u30e9\u30a4\u30a2\u30f3\u30b9\u8981\u4ef6 (GDPR\u3001HIPAA \u306a\u3069) \u306b\u5f93\u3044\u307e\u3059\u3002 MLflow \u306e\u4f7f\u7528\u3068\u30c7\u30d7\u30ed\u30a4\u304c\u3053\u308c\u3089\u306e\u8981\u4ef6\u3092\u6e80\u305f\u3057\u3066\u3044\u308b\u3053\u3068\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p><strong>\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u6700\u9069\u5316<\/strong><\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2\u30ea\u30bd\u30fc\u30b9:\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u8981\u4ef6\u3092\u6e80\u305f\u3059\u305f\u3081\u306b\u3001\u30e2\u30c7\u30eb\u306e\u30cb\u30fc\u30ba\u306b\u5fdc\u3058\u3066\u9069\u5207\u306a<strong>\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2<\/strong>\u30ea\u30bd\u30fc\u30b9(CPU\u3001GPU\u3001\u30e1\u30e2\u30ea\u306a\u3069)\u3092\u9078\u629e\u3057\u307e\u3059\u3002 \u30ea\u30bd\u30fc\u30b9\u5272\u308a\u5f53\u3066\u306f\u3001\u30b9\u30eb\u30fc\u30d7\u30c3\u30c8\u3001\u5f85\u6a5f\u6642\u9593\u306a\u3069\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9 \u30e1\u30c8\u30ea\u30c3\u30af\u306b\u57fa\u3065\u3044\u3066\u8abf\u6574\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30b3\u30fc\u30c9\u306e\u6700\u9069\u5316<\/strong>: \u6a5f\u68b0\u5b66\u7fd2\u30b3\u30fc\u30c9\u3092\u6700\u9069\u5316\u3057\u3066\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3068\u63a8\u8ad6\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5411\u4e0a\u3055\u305b\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u30d9\u30af\u30c8\u30eb\u5316\u6f14\u7b97\u3001\u4e26\u5217\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u3001\u304a\u3088\u3073\u305d\u306e\u4ed6\u306e\u624b\u6cd5\u3092\u4f7f\u7528\u3057\u3066\u3001\u8a08\u7b97\u3092\u9ad8\u901f\u5316\u3057\u3001\u4e0d\u8981\u306a\u8a08\u7b97\u30aa\u30fc\u30d0\u30fc\u30d8\u30c3\u30c9\u3092\u524a\u6e1b\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30e2\u30c7\u30eb\u306e\u6700\u9069\u5316<\/strong>: \u7279\u5b9a\u306e\u30b7\u30ca\u30ea\u30aa\u306b\u5f93\u3063\u3066\u30e2\u30c7\u30eb\u3092\u6700\u9069\u5316\u3057\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u91cf\u5b50\u5316\u3084\u30d7\u30eb\u30fc\u30cb\u30f3\u30b0\u306a\u3069\u306e\u30e2\u30c7\u30eb\u5727\u7e2e\u624b\u6cd5\u3092\u4f7f\u7528\u3057\u3066\u3001\u30e2\u30c7\u30eb\u306e\u30b5\u30a4\u30ba\u3068\u8a08\u7b97\u306e\u8907\u96d1\u3055\u3092\u8efd\u6e1b\u3057\u3001\u63a8\u8ad6\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5411\u4e0a\u3055\u305b\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u8ca0\u8377\u5206\u6563<\/strong>: MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u5834\u5408\u306f\u3001\u30ed\u30fc\u30c9 \u30d0\u30e9\u30f3\u30b5\u30fc\u3092\u4f7f\u7528\u3057\u3066\u30c8\u30e9\u30d5\u30a3\u30c3\u30af\u3092\u5206\u6563\u3057\u3001\u30b5\u30fc\u30d3\u30b9\u306e\u5b89\u5b9a\u6027\u3068\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3\u3092\u78ba\u4fdd\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3<\/strong>: \u30af\u30e9\u30a6\u30c9 \u30b5\u30fc\u30d3\u30b9 \u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u307e\u305f\u306f Kubernetes \u306a\u3069\u306e\u30b3\u30f3\u30c6\u30ca\u30fc \u30aa\u30fc\u30b1\u30b9\u30c8\u30ec\u30fc\u30b7\u30e7\u30f3 \u30c4\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u3001\u5909\u5316\u3059\u308b\u9700\u8981\u306b\u5408\u308f\u305b\u3066\u30b5\u30fc\u30d3\u30b9\u3092\u52d5\u7684\u306b\u30b9\u30b1\u30fc\u30eb\u30a2\u30c3\u30d7\u307e\u305f\u306f\u30b9\u30b1\u30fc\u30eb\u30c0\u30a6\u30f3\u3057\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u3055\u307e\u3056\u307e\u306a\u8ca0\u8377\u8981\u6c42\u3092\u51e6\u7406\u3059\u308b\u305f\u3081\u306b\u30b5\u30fc\u30d3\u30b9\u3092\u9069\u5207\u306b\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30ad\u30e3\u30c3\u30b7\u30e5\u3068\u30d0\u30c3\u30c1\u51e6\u7406<\/strong>: \u5fc5\u8981\u306b\u5fdc\u3058\u3066\u3001\u30ad\u30e3\u30c3\u30b7\u30e5\u3068\u30d0\u30c3\u30c1\u51e6\u7406\u306e\u624b\u6cd5\u3092\u4f7f\u7528\u3057\u3066\u3001\u5f85\u6a5f\u6642\u9593\u3092\u77ed\u7e2e\u3057\u3001\u30b9\u30eb\u30fc\u30d7\u30c3\u30c8\u3092\u5411\u4e0a\u3055\u305b\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u983b\u7e41\u306b\u4f7f\u7528\u3059\u308b\u8a08\u7b97\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3057\u3066\u3001\u7e70\u308a\u8fd4\u3057\u8a08\u7b97\u306e\u30aa\u30fc\u30d0\u30fc\u30d8\u30c3\u30c9\u3092\u6e1b\u3089\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u63a8\u8ad6\u8981\u6c42\u306e\u30d0\u30c3\u30c1\u51e6\u7406\u306b\u3088\u308b GPU \u4f7f\u7528\u7387\u3068\u30b9\u30eb\u30fc\u30d7\u30c3\u30c8\u306e\u5411\u4e0a<\/li>\n<\/ol>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"7\">\n<li><strong>\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9 \u30e1\u30c8\u30ea\u30c3\u30af\u3092\u76e3\u8996\u3059\u308b<\/strong>: \u5f85\u6a5f\u6642\u9593\u3001\u30b9\u30eb\u30fc\u30d7\u30c3\u30c8\u3001\u30ea\u30bd\u30fc\u30b9\u4f7f\u7528\u7387\u306a\u3069\u306e\u4e3b\u8981\u696d\u7e3e\u8a55\u4fa1\u6307\u6a19\u3092\u5b9a\u671f\u7684\u306b\u76e3\u8996\u3057\u3066\u3001\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u30dc\u30c8\u30eb\u30cd\u30c3\u30af\u3092\u7279\u5b9a\u3057\u3066\u5bfe\u51e6\u3067\u304d\u308b\u3088\u3046\u306b\u3057\u307e\u3059\u3002 \u30ea\u30a2\u30eb\u30bf\u30a4\u30e0\u306e\u76e3\u8996\u3068\u8996\u899a\u5316\u306f\u3001\u30d7\u30ed\u30e1\u30c6\u30a6\u30b9\u3001\u30b0\u30e9\u30d5\u30a1\u30ca\u306a\u3069\u306e\u76e3\u8996\u30c4\u30fc\u30eb\u3092\u4f7f\u7528\u3057\u3066\u5b9f\u73fe\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u975e\u540c\u671f<\/strong>\u51e6\u7406: \u9069\u5207\u306a\u30b7\u30ca\u30ea\u30aa\u3067\u306f\u3001\u975e\u540c\u671f\u51e6\u7406\u624b\u6cd5\u3092\u4f7f\u7528\u3057\u3066\u30b7\u30b9\u30c6\u30e0 \u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5411\u4e0a\u3055\u305b\u307e\u3059\u3002 \u305f\u3068\u3048\u3070\u3001\u6642\u9593\u304c\u304b\u304b\u308b\u63a8\u8ad6\u30ea\u30af\u30a8\u30b9\u30c8\u3092\u51e6\u7406\u3059\u308b\u5834\u5408\u3001\u30e1\u30c3\u30bb\u30fc\u30b8\u30ad\u30e5\u30fc (RabbitMQ\u3001Kafka \u306a\u3069) \u3092\u4f7f\u7528\u3057\u3066\u30ea\u30af\u30a8\u30b9\u30c8\u3092\u30ad\u30e5\u30fc\u306b\u5165\u308c\u3001\u975e\u540c\u671f\u7684\u306b\u51e6\u7406\u3057\u3066\u7d50\u679c\u3092\u8fd4\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u5206\u6563<\/strong>\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0: \u5927\u898f\u6a21\u306a\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3059\u308b\u5834\u5408\u3001MLflow \u3092\u4f7f\u7528\u3057\u3066 Horovod\u3001TensorFlow \u5206\u6563\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u306a\u3069\u306e\u5206\u6563\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0 \u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3068\u7d71\u5408\u3057\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0 \u30d7\u30ed\u30bb\u30b9\u3092\u9ad8\u901f\u5316\u3057\u3001\u30ea\u30bd\u30fc\u30b9\u4f7f\u7528\u7387\u3092\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u7d99\u7d9a\u7684<strong>\u30a4\u30f3\u30c6\u30b0\u30ec\u30fc\u30b7\u30e7\u30f3\/\u7d99\u7d9a\u7684\u30c7\u30d7\u30ed\u30a4 (CI\/<\/strong>CD): MLflow \u3092 Jenkins\u3001Travis CI\u3001GitHub Actions \u306a\u3069\u306e CI\/CD \u30c4\u30fc\u30eb\u3068\u7d71\u5408\u3057\u3066\u3001\u30d3\u30eb\u30c9\u3001\u30c6\u30b9\u30c8\u3001\u30c7\u30d7\u30ed\u30a4\u306e\u30d7\u30ed\u30bb\u30b9\u3092\u81ea\u52d5\u5316\u3057\u307e\u3059\u3002 \u3053\u308c\u306b\u3088\u308a\u3001\u30e2\u30c7\u30eb\u306e\u30bf\u30a4\u30e0\u30ea\u30fc\u306a\u66f4\u65b0\u3068\u4e00\u8cab\u6027\u304c\u4fdd\u8a3c\u3055\u308c\u3001\u4eba\u70ba\u7684\u30df\u30b9\u306e\u53ef\u80fd\u6027\u304c\u6e1b\u5c11\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u6b63\u5e38\u6027\u30c1\u30a7\u30c3\u30af\u3068\u969c\u5bb3\u5fa9\u65e7<\/strong>: \u6b63\u5e38\u6027\u30c1\u30a7\u30c3\u30af\u3092\u5b9f\u88c5\u3057\u3066\u3001MLflow \u30b5\u30fc\u30d3\u30b9\u306e\u6b63\u5e38\u6027\u3092\u76e3\u8996\u3057\u307e\u3059\u3002 \u30b5\u30fc\u30d3\u30b9\u969c\u5bb3\u304c\u767a\u751f\u3057\u305f\u5834\u5408\u306b\u3001\u30b5\u30fc\u30d3\u30b9\u306e\u518d\u8d77\u52d5\u3084\u30d0\u30c3\u30af\u30a2\u30c3\u30d7\u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u3078\u306e\u5207\u308a\u66ff\u3048\u306a\u3069\u306e\u30ea\u30ab\u30d0\u30ea\u30d7\u30ed\u30bb\u30b9\u3092\u81ea\u52d5\u7684\u306b\u30c8\u30ea\u30ac\u30fc\u3057\u3066\u3001\u30b5\u30fc\u30d3\u30b9\u306e\u9ad8\u53ef\u7528\u6027\u3092\u78ba\u4fdd\u3057\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u4e0a\u8a18\u306e\u6226\u7565\u306b\u5f93\u3046\u3068\u3001\u904b\u7528\u74b0\u5883\u3067\u306e MLflow \u306e\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5411\u4e0a\u3055\u305b\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u9069\u5207\u306a\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u5bfe\u7b56\u3092\u5b9f\u88c5\u3057\u3001\u30b3\u30fc\u30c9\u3068\u30e2\u30c7\u30eb\u3092\u6700\u9069\u5316\u3057\u3001\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9 \u30e1\u30c8\u30ea\u30c3\u30af\u3092\u76e3\u8996\u3057\u3001\u9ad8\u53ef\u7528\u6027\u3092\u78ba\u4fdd\u3059\u308b\u3053\u3068\u3067\u3001MLflow \u304c\u904b\u7528\u74b0\u5883\u306e\u8981\u6c42\u3092\u78ba\u5b9f\u306b\u6e80\u305f\u3059\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u3053\u306e\u30b3\u30fc\u30b9\u306e\u5fa9\u7fd2\u3068\u8981\u7d04\u3092\u884c\u3063\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u3053\u306e\u30b3\u30fc\u30b9\u3067\u306f\u3001MLflow\u3068\u305d\u306e\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3078\u306e\u5fdc\u7528\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u307e\u3059\u3002 \u4ee5\u4e0b\u306f\u3001\u4e3b\u8981\u306a\u6982\u5ff5\u3068\u30c8\u30d4\u30c3\u30af\u306e\u30ec\u30d3\u30e5\u30fc\u3068\u8981\u7d04\u3067\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>MLflow \u306e\u6982\u8981: MLflow \u306e\u80cc\u666f\u3001\u76ee\u6a19\u3001\u76ee\u6a19\u3001\u304a\u3088\u3073\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u7ba1\u7406\u30c4\u30fc\u30eb\u3068\u3057\u3066&nbsp;<strong>MLflow<\/strong>&nbsp;\u3092\u9078\u629e\u3057\u305f\u7406\u7531\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>MLflow \u306e\u5229\u70b9\u3068\u6a5f\u80fd<\/strong>: \u30e2\u30b8\u30e5\u30fc\u30eb\u6027\u3001\u30b9\u30b1\u30fc\u30e9\u30d3\u30ea\u30c6\u30a3\u3001\u4f7f\u3044\u3084\u3059\u3055\u306a\u3069\u3001MLflow \u306e\u4e3b\u306a\u5229\u70b9\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3057\u305f\u3002<\/li>\n\n\n\n<li><strong>MLflow \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3: MLflow<\/strong>&nbsp;\u306e 4 \u3064\u306e\u4e3b\u8981\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8 (\u8ffd\u8de1\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3001\u30e2\u30c7\u30eb\u3001\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea) \u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u3001MLflow \u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u306b\u304a\u3051\u308b\u305d\u308c\u3089\u306e\u5f79\u5272\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3068\u69cb\u6210<\/strong>: MLflow \u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u65b9\u6cd5\u3068\u3001\u74b0\u5883\u3068\u30d0\u30c3\u30af\u30a8\u30f3\u30c9 \u30b9\u30c8\u30ec\u30fc\u30b8\u3092\u69cb\u6210\u3059\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3057\u305f\u3002<\/li>\n\n\n\n<li><strong>MLflow<\/strong>&nbsp;\u8ffd\u8de1: \u5b9f\u9a13\u30c7\u30fc\u30bf\u306e\u8a18\u9332\u3001\u5b9f\u9a13\u7d50\u679c\u306e\u30af\u30a8\u30ea\u3068\u6bd4\u8f03\u3001\u30d1\u30e9\u30e1\u30fc\u30bf\u30fc\u3001\u30e1\u30c8\u30ea\u30c3\u30af\u3001\u6210\u679c\u7269\u306e\u8a18\u9332\u306a\u3069\u3001MLflow \u8ffd\u8de1\u306e\u6a5f\u80fd\u3068\u4f7f\u7528\u6cd5\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u8aac\u660e\u3057\u307e\u3057\u305f\u3002<\/li>\n\n\n\n<li><strong>MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8: MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8<\/strong>\u3092\u4f5c\u6210\u3001\u7ba1\u7406\u3001\u5b9f\u884c\u3059\u308b\u65b9\u6cd5\u3068\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u4f9d\u5b58\u95a2\u4fc2\u306e\u7ba1\u7406\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3057\u305f\u3002<\/li>\n\n\n\n<li><strong>MLflow<\/strong>&nbsp;\u30e2\u30c7\u30eb: \u3055\u307e\u3056\u307e\u306a\u30e2\u30c7\u30eb\u5f62\u5f0f\u3067 MLflow \u30e2\u30c7\u30eb\u3092\u4f5c\u6210\u3001\u4fdd\u5b58\u3001\u30c7\u30d7\u30ed\u30a4\u3001\u4f7f\u7528\u3059\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3057\u305f\u3002<\/li>\n\n\n\n<li><strong>MLflow<\/strong>&nbsp;\u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea: \u30e2\u30c7\u30eb\u306e\u767b\u9332\u3001\u30d0\u30fc\u30b8\u30e7\u30f3\u7ba1\u7406\u3001\u30e2\u30c7\u30eb \u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u7ba1\u7406\u3001\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u306a\u3069\u3001MLflow \u30e2\u30c7\u30eb \u30ec\u30b8\u30b9\u30c8\u30ea\u306e\u6a5f\u80fd\u304c\u5c0e\u5165\u3055\u308c\u307e\u3057\u305f\u3002<\/li>\n\n\n\n<li><strong>\u30af\u30e9\u30a6\u30c9\u30b5\u30fc\u30d3\u30b9\u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u3068\u306e\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u7d71\u5408<\/strong>: MLflow\u3092TensorFlow\u3084PyTorch\u306a\u3069\u306e\u30c7\u30a3\u30fc\u30d7\u30e9\u30fc\u30cb\u30f3\u30b0\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3001\u304a\u3088\u3073AWS SageMaker\u3084Azure ML\u306a\u3069\u306e\u30af\u30e9\u30a6\u30c9\u30d7\u30ed\u30d0\u30a4\u30c0\u30fc\u306b\u7d71\u5408\u3059\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3057\u305f\u3002<\/li>\n\n\n\n<li>Jupyter Notebook \u3068 MLflow:&nbsp;<strong>Jupyter Notebook \u3067 MLflow<\/strong>&nbsp;\u3092\u4f7f\u7528\u3059\u308b\u65b9\u6cd5\u3092\u7d39\u4ecb\u3057\u307e\u3057\u305f\u3002<\/li>\n\n\n\n<li><strong>\u904b\u7528\u74b0\u5883\u306e\u30c7\u30d7\u30ed\u30a4\u3068\u30d9\u30b9\u30c8 \u30d7\u30e9\u30af\u30c6\u30a3\u30b9<\/strong>: MLflow \u30e2\u30c7\u30eb\u3092\u904b\u7528\u74b0\u5883\u306b\u30c7\u30d7\u30ed\u30a4\u3059\u308b\u65b9\u6cd5\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u3001\u30bb\u30ad\u30e5\u30ea\u30c6\u30a3\u3084\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u6700\u9069\u5316\u306a\u3069\u3001\u904b\u7528\u74b0\u5883\u3067\u306e\u30d9\u30b9\u30c8 \u30d7\u30e9\u30af\u30c6\u30a3\u30b9\u3092\u3044\u304f\u3064\u304b\u5171\u6709\u3057\u307e\u3057\u305f\u3002<\/li>\n<\/ol>\n\n\n\n<p>\u3053\u306e\u30b3\u30fc\u30b9\u307e\u3067\u306b\u3001MLflow \u306e\u4e3b\u8981\u306a\u6982\u5ff5\u3068\u6a5f\u80fd\u3092\u7fd2\u5f97\u3057\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u5b9f\u969b\u306e\u30b7\u30ca\u30ea\u30aa\u306b\u9069\u7528\u3059\u308b\u65b9\u6cd5\u3092\u7406\u89e3\u3057\u3066\u3044\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002 MLflow \u306e\u30d1\u30ef\u30fc\u3092\u6700\u5927\u9650\u306b\u6d3b\u7528\u3057\u3066\u3001\u6a5f\u68b0\u5b66\u7fd2\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u3092\u52b9\u679c\u7684\u306b\u7ba1\u7406\u304a\u3088\u3073\u6700\u9069\u5316\u3067\u304d\u308b\u3053\u3068\u3092\u9858\u3063\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>MLFOW\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u3068\u30ea\u30bd\u30fc\u30b9\u306b\u3064\u3044\u3066\u6559\u3048\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u306f\u3001\u30e6\u30fc\u30b6\u30fc\u306b\u8c4a\u5bcc\u306a\u30ea\u30bd\u30fc\u30b9\u3001\u30b5\u30dd\u30fc\u30c8\u3001\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u306e\u6a5f\u4f1a\u3092\u63d0\u4f9b\u3059\u308b\u3001\u6d3b\u767a\u3067\u6210\u9577\u3057\u3066\u3044\u308b\u30a8\u30b3\u30b7\u30b9\u30c6\u30e0\u3067\u3059\u3002 MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u3068\u30ea\u30bd\u30fc\u30b9\u306b\u95a2\u9023\u3059\u308b\u60c5\u5831\u3092\u6b21\u306b\u793a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>GitHub \u30ea\u30dd\u30b8\u30c8\u30ea<\/strong>: MLflow \u306e\u30bd\u30fc\u30b9 \u30b3\u30fc\u30c9\u306f GitHub \u3067\u30db\u30b9\u30c8\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u3053\u3053\u3067\u306f\u3001\u30bd\u30fc\u30b9\u30b3\u30fc\u30c9\u3001\u8ab2\u984c\u8ffd\u8de1\u3001\u6295\u7a3f\u30ac\u30a4\u30c9\u30e9\u30a4\u30f3\u306a\u3069\u3092\u898b\u3064\u3051\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 GitHub \u30ea\u30dd\u30b8\u30c8\u30ea \u30a2\u30c9\u30ec\u30b9:&nbsp;<a href=\"https:\/\/github.com\/mlflow\/mlflow\">https:\/\/github.com\/mlflow\/mlflow<\/a><\/li>\n\n\n\n<li><strong>\u516c\u5f0f<\/strong>\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8: MLflow \u306e\u516c\u5f0f\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u306b\u306f\u3001\u8a73\u7d30\u306a\u30e6\u30fc\u30b6\u30fc \u30ac\u30a4\u30c9\u3001API \u30ea\u30d5\u30a1\u30ec\u30f3\u30b9\u3001\u958b\u767a\u8005\u30ac\u30a4\u30c9\u306a\u3069\u304c\u8a18\u8f09\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u6587\u66f8\u4f4f\u6240:<a href=\"https:\/\/www.mlflow.org\/docs\/latest\/index.html\">https:\/\/www.mlflow.org\/docs\/latest\/index.html<\/a><\/li>\n\n\n\n<li><strong>\u516c\u5f0f<\/strong>\u30d6\u30ed\u30b0: MLflow \u306e\u516c\u5f0f\u30d6\u30ed\u30b0\u306b\u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u65b0\u6a5f\u80fd\u3001\u30b1\u30fc\u30b9 \u30b9\u30bf\u30c7\u30a3\u3001\u30a4\u30d9\u30f3\u30c8\u306a\u3069\u306b\u95a2\u3059\u308b\u8a18\u4e8b\u304c\u63b2\u8f09\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u30d6\u30ed\u30b0\u30a2\u30c9\u30ec\u30b9:&nbsp;<a href=\"https:\/\/medium.com\/mlflow\">https:\/\/medium.com\/mlflow<\/a><\/li>\n\n\n\n<li><strong>\u30e1\u30fc\u30ea\u30f3\u30b0<\/strong>&nbsp;\u30ea\u30b9\u30c8: MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u306b\u306f\u3001\u66f4\u65b0\u3001\u8cea\u554f\u3001\u30c7\u30a3\u30b9\u30ab\u30c3\u30b7\u30e7\u30f3\u3092\u8cfc\u8aad\u3067\u304d\u308b Google \u30b0\u30eb\u30fc\u30d7 \u30e1\u30fc\u30ea\u30f3\u30b0 \u30ea\u30b9\u30c8\u304c\u3042\u308a\u307e\u3059\u3002 \u30e1\u30fc\u30ea\u30f3\u30b0\u30ea\u30b9\u30c8\u306e\u30a2\u30c9\u30ec\u30b9:&nbsp;<a href=\"https:\/\/groups.google.com\/a\/databricks.com\/forum\/#!forum\/mlflow-users\">https:\/\/groups.google.com\/a\/databricks.com\/forum\/#!forum\/mlflow-users<\/a><\/li>\n\n\n\n<li><strong>Slack<\/strong>: MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3 Slack \u306b\u306f\u3001\u4ed6\u306e\u30e6\u30fc\u30b6\u30fc\u3084\u958b\u767a\u8005\u3068\u30b3\u30df\u30e5\u30cb\u30b1\u30fc\u30b7\u30e7\u30f3\u3092\u3068\u308b\u305f\u3081\u306b\u53c2\u52a0\u3067\u304d\u308b\u5c02\u7528\u30c1\u30e3\u30f3\u30cd\u30eb\u304c\u3042\u308a\u307e\u3059\u3002 \u30b9\u30e9\u30c3\u30af\u30c1\u30e3\u30f3\u30cd\u30eb\u30a2\u30c9\u30ec\u30b9:<a href=\"https:\/\/tinyurl.com\/y8msy8tb\">https:\/\/tinyurl.com\/y8msy8tb<\/a><\/li>\n\n\n\n<li>\u30b9\u30bf\u30c3\u30af\u30aa\u30fc\u30d0\u30fc\u30d5\u30ed\u30fc:&nbsp;<strong>\u30b9\u30bf\u30c3\u30af\u30aa\u30fc\u30d0\u30fc\u30d5\u30ed\u30fc<\/strong>\u3067\u306f\u3001\u30bf\u30b0\u3092\u4f7f\u7528\u3057\u3066\u8cea\u554f\u3057\u305f\u308a\u3001\u52a9\u3051\u3092\u6c42\u3081\u305f\u308a\u3067\u304d\u307e\u3059\u3002 \u591a\u304f\u306e\u8cea\u554f\u304c\u8a73\u7d30\u306b\u56de\u7b54\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u30b9\u30bf\u30c3\u30af\u30aa\u30fc\u30d0\u30fc\u30d5\u30ed\u30fc\u30e9\u30d9\u30eb\u30a2\u30c9\u30ec\u30b9:<a href=\"https:\/\/stackoverflow.com\/questions\/tagged\/mlflow\">https:\/\/stackoverflow.com\/questions\/tagged\/mlflow<\/a><code>mlflow<\/code><\/li>\n\n\n\n<li><strong>\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u3078\u306e\u8ca2\u732e<\/strong>: MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u306f\u3001\u30d0\u30b0\u306e\u4fee\u6b63\u3001\u65b0\u6a5f\u80fd\u306e\u8ffd\u52a0\u3001\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u306e\u6539\u5584\u306a\u3069\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u8ca2\u732e\u3059\u308b\u3059\u3079\u3066\u306e\u4eba\u3092\u6b53\u8fce\u3057\u307e\u3059\u3002 GitHub \u30ea\u30dd\u30b8\u30c8\u30ea\u306e [\u30b3\u30f3\u30c8\u30ea\u30d3\u30e5\u30fc\u30c8] \u30bb\u30af\u30b7\u30e7\u30f3\u3092\u78ba\u8a8d\u3057\u3066\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u30b3\u30f3\u30c8\u30ea\u30d3\u30e5\u30fc\u30c8\u3059\u308b\u65b9\u6cd5\u3092\u78ba\u8a8d\u3067\u304d\u307e\u3059\u3002&nbsp;<a href=\"https:\/\/github.com\/mlflow\/mlflow\/blob\/main\/CONTRIBUTING.rst\">https:\/\/github.com\/mlflow\/mlflow\/blob\/main\/CONTRIBUTING.rst<\/a><\/li>\n\n\n\n<li>\u30b1\u30fc\u30b9 \u30b9\u30bf\u30c7\u30a3\u3068\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb: \u516c\u5f0f\u30ea\u30bd\u30fc\u30b9\u306b\u52a0\u3048\u3066\u3001\u3055\u307e\u3056\u307e\u306a\u30b7\u30ca\u30ea\u30aa\u3067 MLflow \u3092\u4f7f\u7528\u3059\u308b\u65b9\u6cd5\u3092\u793a\u3059\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u304b\u3089\u306e<strong>\u591a\u304f\u306e\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u3001\u30b1\u30fc\u30b9 \u30b9\u30bf\u30c7\u30a3\u3001\u30c7\u30e2<\/strong>\u304c\u3042\u308a\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u30ea\u30bd\u30fc\u30b9\u306f\u3001\u30d6\u30ed\u30b0\u3001GitHub \u30ea\u30dd\u30b8\u30c8\u30ea\u3001YouTube \u30d3\u30c7\u30aa\u306a\u3069\u306b\u3042\u308a\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u306b\u53c2\u52a0\u3057\u3001\u3053\u308c\u3089\u306e\u30ea\u30bd\u30fc\u30b9\u3092\u6d3b\u7528\u3059\u308b\u3053\u3068\u3067\u3001MLflow \u306e\u6700\u65b0\u306e\u958b\u767a\u3092\u3088\u308a\u3088\u304f\u7406\u89e3\u3057\u3001\u767a\u751f\u3057\u305f\u554f\u984c\u3092\u89e3\u6c7a\u3057\u3001\u4ed6\u306e\u30e6\u30fc\u30b6\u30fc\u3084\u958b\u767a\u8005\u3068\u3064\u306a\u304c\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 \u540c\u6642\u306b\u3001\u3042\u306a\u305f\u306f<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"http:\/\/edu.ujhb.org\/?p=1305\" alt=\"\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/chat.openai.com\/_next\/image?url=https%3A%2F%2Flh3.googleusercontent.com%2Fa%2FAGNmyxbwVdnTP0E2eBqDrujRQjV-wFP12Thr4vFgysVv5A%3Ds96-c&amp;w=64&amp;q=75\" alt=\"\u8ecd\u4ead\"\/><\/figure>\n\n\n\n<p>\u7d9a\u3051\u3066\u304f\u3060\u3055\u3044<\/p>\n\n\n\n<p>\u8ca2\u732e\u306f\u3001MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u306e\u7d99\u7d9a\u7684\u306a\u6539\u5584\u3068\u6210\u9577\u306b\u3082\u5f79\u7acb\u3061\u307e\u3059\u3002 \u53c2\u52a0\u65b9\u6cd5\u3092\u3044\u304f\u3064\u304b\u3054\u7d39\u4ecb\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u8cea\u554f\u3084\u63d0\u6848\u3092\u3059\u308b<\/strong>: GitHub \u30ea\u30dd\u30b8\u30c8\u30ea\u3067\u554f\u984c\u3092\u9001\u4fe1\u3059\u308b\u3068\u3001\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u304c\u30d0\u30b0\u3092\u898b\u3064\u3051\u305f\u308a\u3001\u6539\u5584\u3092\u63d0\u6848\u3057\u305f\u308a\u3001\u65b0\u6a5f\u80fd\u306b\u3064\u3044\u3066\u8a71\u3057\u5408\u3063\u305f\u308a\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002 \u540c\u6642\u306b\u3001\u30b9\u30bf\u30c3\u30af\u30aa\u30fc\u30d0\u30fc\u30d5\u30ed\u30fc\u306b\u95a2\u3059\u308b\u8cea\u554f\u3092\u3059\u308b\u3053\u3068\u306f\u3001\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u30e1\u30f3\u30d0\u30fc\u306b\u3088\u3063\u3066\u3082\u52a9\u3051\u3089\u308c\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u77e5\u8b58\u3092\u5171\u6709\u3059\u308b<\/strong>: \u30d6\u30ed\u30b0\u6295\u7a3f\u3092\u66f8\u3044\u305f\u308a\u3001\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb \u30d3\u30c7\u30aa\u3092\u9332\u753b\u3057\u305f\u308a\u3001\u30b5\u30f3\u30d7\u30eb \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u3092\u4f5c\u6210\u3057\u305f\u308a\u3057\u3066\u3001MLflow \u3092\u4f7f\u7528\u3057\u305f\u7d4c\u9a13\u3084\u30dd\u30a4\u30f3\u30c8\u3092\u5171\u6709\u3067\u304d\u307e\u3059\u3002 \u77e5\u8b58\u3092\u5171\u6709\u3059\u308b\u3053\u3068\u3067\u3001\u4ed6\u306e\u30e6\u30fc\u30b6\u30fc\u304c MLflow \u3092\u3088\u308a\u3088\u304f\u7406\u89e3\u3057\u3066\u4f7f\u7528\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li><strong>\u30c7\u30a3\u30b9\u30ab\u30c3\u30b7\u30e7\u30f3\u3068\u30b5\u30dd\u30fc\u30c8\u306b\u53c2\u52a0\u3059\u308b<\/strong>: \u30e1\u30fc\u30ea\u30f3\u30b0\u30ea\u30b9\u30c8\u3001Slack \u30c1\u30e3\u30f3\u30cd\u30eb\u3001Stack Overflow \u3067\u306e\u30c7\u30a3\u30b9\u30ab\u30c3\u30b7\u30e7\u30f3\u306b\u7a4d\u6975\u7684\u306b\u53c2\u52a0\u3057\u3001\u4ed6\u306e\u30e6\u30fc\u30b6\u30fc\u306e\u8cea\u554f\u306b\u7b54\u3048\u308b\u3053\u3068\u3067\u3001\u3088\u308a\u5f37\u529b\u3067\u89aa\u3057\u307f\u3084\u3059\u3044\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u3092\u69cb\u7bc9\u3067\u304d\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u30b3\u30fc\u30c9\u3068\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u306e\u63d0\u4f9b:&nbsp;<strong>\u30b3\u30fc\u30c9\u3068\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8<\/strong>\u306e\u6539\u5584\u3092 MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u9001\u4fe1\u3059\u308b\u3053\u3068\u306f\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u6210\u9577\u3092\u652f\u63f4\u3059\u308b\u76f4\u63a5\u7684\u306a\u65b9\u6cd5\u3067\u3059\u3002 GitHub \u30ea\u30dd\u30b8\u30c8\u30ea\u306e\u672a\u89e3\u6c7a\u306e\u554f\u984c\u3068\u6295\u7a3f\u306e\u30ac\u30a4\u30c9\u30e9\u30a4\u30f3\u3092\u78ba\u8a8d\u3057\u3066\u3001\u8ca2\u732e\u3067\u304d\u308b\u30bf\u30b9\u30af\u3092\u898b\u3064\u3051\u3066\u304f\u3060\u3055\u3044\u3002<\/li>\n\n\n\n<li>\u30a4\u30d9\u30f3\u30c8<strong>\u3092\u958b\u50ac\u3057\u3066\u53c2\u52a0<\/strong>\u3059\u308b: \u30ef\u30fc\u30af\u30b7\u30e7\u30c3\u30d7\u3001\u8b1b\u7fa9\u3001\u30cf\u30c3\u30ab\u30bd\u30f3\u306a\u3069\u3001MLflow \u306b\u95a2\u9023\u3059\u308b\u30aa\u30f3\u30e9\u30a4\u30f3\u307e\u305f\u306f\u30aa\u30d5\u30e9\u30a4\u30f3\u306e\u30a4\u30d9\u30f3\u30c8\u3092\u958b\u50ac\u307e\u305f\u306f\u53c2\u52a0\u3067\u304d\u307e\u3059\u3002 \u3053\u308c\u3089\u306e\u6d3b\u52d5\u306f\u3001\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u306e\u5f71\u97ff\u529b\u3092\u62e1\u5927\u3057\u3001\u30e1\u30f3\u30d0\u30fc\u9593\u306e\u30b3\u30df\u30e5\u30cb\u30b1\u30fc\u30b7\u30e7\u30f3\u3068\u5354\u529b\u3092\u4fc3\u9032\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002<\/li>\n<\/ol>\n\n\n\n<p>MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u306b\u7a4d\u6975\u7684\u306b\u53c2\u52a0\u3059\u308b\u3053\u3068\u3067\u3001\u4ed6\u306e\u30e6\u30fc\u30b6\u30fc\u3084\u958b\u767a\u8005\u3068\u3064\u306a\u304c\u308a\u3001\u30b9\u30ad\u30eb\u3092\u5411\u4e0a\u3055\u305b\u3001MLflow \u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306e\u958b\u767a\u306b\u8ca2\u732e\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002 MLflow \u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u306e\u30b5\u30dd\u30fc\u30c8\u3068\u30b3\u30e9\u30dc\u30ec\u30fc\u30b7\u30e7\u30f3\u306b\u3088\u308a\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306f\u3088\u308a\u30b9\u30e0\u30fc\u30ba\u3067\u9769\u65b0\u7684\u306b\u306a\u308a\u307e\u3059<\/p>\n","protected":false},"excerpt":{"rendered":"<p>IT\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u306e\u30d3\u30c7\u30aa\u3092\u3084\u308a\u305f\u3044\u306e\u3067 &hellip; <\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_mi_skip_tracking":false,"footnotes":""},"categories":[23],"tags":[],"class_list":["post-1313","post","type-post","status-publish","format-standard","hentry","category-material"],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=\/wp\/v2\/posts\/1313","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1313"}],"version-history":[{"count":1,"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=\/wp\/v2\/posts\/1313\/revisions"}],"predecessor-version":[{"id":1314,"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=\/wp\/v2\/posts\/1313\/revisions\/1314"}],"wp:attachment":[{"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1313"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1313"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/edu.ujhb.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1313"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}