{"id":104698,"date":"2025-04-05T14:31:41","date_gmt":"2025-04-05T10:01:41","guid":{"rendered":"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/"},"modified":"2025-04-05T14:31:41","modified_gmt":"2025-04-05T10:01:41","slug":"diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p","status":"publish","type":"post","link":"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/","title":{"rendered":"\ud83e\udde0 Diagram2graph: \u062a\u0646\u0638\u06cc\u0645 \u062f\u0642\u06cc\u0642 \u06cc\u06a9 \u0645\u062f\u0644 \u0632\u0628\u0627\u0646 \u0628\u06cc\u0646\u0627\u06cc\u06cc \u0628\u0631\u0627\u06cc \u0627\u0633\u062a\u062e\u0631\u0627\u062c \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627\u06cc \u062f\u0627\u0646\u0634 \u0627\u0632 \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627"},"content":{"rendered":"<div data-article-id=\"2383400\" id=\"article-body\">\n<p><em>\u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 QWEN2.5-VL + PEFT + NEO4J<\/em><\/p>\n<p><em>\u062a\u0646\u0638\u06cc\u0645 \u0634\u062f\u0647 \u062a\u0648\u0633\u0637 \u0645\u062d\u0645\u062f \u0635\u0641\u0648\u0627\u0646 | \u0631\u0627\u0647 \u062d\u0644 \u0647\u0627\u06cc Zackriya<\/em><\/p>\n<hr\/>\n<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_84 counter-hierarchy ez-toc-counter-rtl ez-toc-grey ez-toc-container-direction\">\n<div class=\"ez-toc-title-container\">\n<p class=\"ez-toc-title\" style=\"cursor:inherit\">\u0641\u0647\u0631\u0633\u062a \u0645\u0637\u0627\u0644\u0628<\/p>\n<span class=\"ez-toc-title-toggle\"><a href=\"#\" class=\"ez-toc-pull-right ez-toc-btn ez-toc-btn-xs ez-toc-btn-default ez-toc-toggle\" aria-label=\"Toggle Table of Content\"><span class=\"ez-toc-js-icon-con\"><span class=\"\"><span class=\"eztoc-hide\" style=\"display:none;\">Toggle<\/span><span class=\"ez-toc-icon-toggle-span\"><svg style=\"fill: #999;color:#999\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" class=\"list-377408\" width=\"20px\" height=\"20px\" viewBox=\"0 0 24 24\" fill=\"none\"><path d=\"M6 6H4v2h2V6zm14 0H8v2h12V6zM4 11h2v2H4v-2zm16 0H8v2h12v-2zM4 16h2v2H4v-2zm16 0H8v2h12v-2z\" fill=\"currentColor\"><\/path><\/svg><svg style=\"fill: #999;color:#999\" class=\"arrow-unsorted-368013\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"10px\" height=\"10px\" viewBox=\"0 0 24 24\" version=\"1.2\" baseProfile=\"tiny\"><path d=\"M18.2 9.3l-6.2-6.3-6.2 6.3c-.2.2-.3.4-.3.7s.1.5.3.7c.2.2.4.3.7.3h11c.3 0 .5-.1.7-.3.2-.2.3-.5.3-.7s-.1-.5-.3-.7zM5.8 14.7l6.2 6.3 6.2-6.3c.2-.2.3-.5.3-.7s-.1-.5-.3-.7c-.2-.2-.4-.3-.7-.3h-11c-.3 0-.5.1-.7.3-.2.2-.3.5-.3.7s.1.5.3.7z\"\/><\/svg><\/span><\/span><\/span><\/a><\/span><\/div>\n<nav><ul class='ez-toc-list ez-toc-list-level-1 ' ><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-1\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%9A%80_tl_%D8%9B_%D8%AF%DA%A9%D8%AA%D8%B1\" >\ud83d\ude80 tl \u061b \u062f\u06a9\u062a\u0631<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-2\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%A7%A9_%D9%85%D8%B4%DA%A9%D9%84\" >\ud83e\udde9 \u0645\u0634\u06a9\u0644<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-3\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%94%8D_%D8%A2%D9%86%DA%86%D9%87_%D9%85%D8%A7_%D8%B3%D8%A7%D8%AE%D8%AA%DB%8C%D9%85\" >\ud83d\udd0d \u0622\u0646\u0686\u0647 \u0645\u0627 \u0633\u0627\u062e\u062a\u06cc\u0645<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-4\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%94%97_%D9%85%D9%86%D8%A7%D8%A8%D8%B9\" >\ud83d\udd17 \u0645\u0646\u0627\u0628\u0639<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-5\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%E2%9A%99_%D9%85%D8%B9%D9%85%D8%A7%D8%B1%DB%8C_%D8%AF%D8%B1_%DB%8C%DA%A9_%D9%86%DA%AF%D8%A7%D9%87\" >\u2699 \u0645\u0639\u0645\u0627\u0631\u06cc \u062f\u0631 \u06cc\u06a9 \u0646\u06af\u0627\u0647<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-6\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%A4%96_%DA%86%D8%B1%D8%A7_%D8%A7%D8%B2_GPT-4_%DB%8C%D8%A7_%DA%A9%D9%84%D9%88%D8%AF_%D8%A7%D8%B3%D8%AA%D9%81%D8%A7%D8%AF%D9%87_%D9%86%D9%85%DB%8C_%DA%A9%D9%86%DB%8C%D9%85%D8%9F\" >\ud83e\udd16 \u0686\u0631\u0627 \u0627\u0632 GPT-4 \u06cc\u0627 \u06a9\u0644\u0648\u062f \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0646\u0645\u06cc \u06a9\u0646\u06cc\u0645\u061f<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-7\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%D9%BE%D8%B4%D8%AA%D9%87_%D9%81%D9%86%DB%8C\" >\u067e\u0634\u062a\u0647 \u0641\u0646\u06cc<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-8\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%93%8A_%D9%86%D8%AA%D8%A7%DB%8C%D8%AC\" >\ud83d\udcca \u0646\u062a\u0627\u06cc\u062c<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-9\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%A7%A0_%D8%AC%D8%B2%D8%A6%DB%8C%D8%A7%D8%AA_%D8%A2%D9%85%D9%88%D8%B2%D8%B4\" >\ud83e\udde0 \u062c\u0632\u0626\u06cc\u0627\u062a \u0622\u0645\u0648\u0632\u0634<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-10\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%A7%AA_%D8%AE%D9%88%D8%AF%D8%AA%D8%A7%D9%86_%D8%A2%D9%86_%D8%B1%D8%A7_%D8%A7%D9%85%D8%AA%D8%AD%D8%A7%D9%86_%DA%A9%D9%86%DB%8C%D8%AF\" >\ud83e\uddea \u062e\u0648\u062f\u062a\u0627\u0646 \u0622\u0646 \u0631\u0627 \u0627\u0645\u062a\u062d\u0627\u0646 \u06a9\u0646\u06cc\u062f<\/a><ul class='ez-toc-list-level-3' ><li class='ez-toc-heading-level-3'><a class=\"ez-toc-link ez-toc-heading-11\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%D9%82%D8%B7%D8%B9%D9%87_%D8%A7%D8%B3%D8%AA%D9%86%D8%A8%D8%A7%D8%B7\" >\u0642\u0637\u0639\u0647 \u0627\u0633\u062a\u0646\u0628\u0627\u0637:<\/a><\/li><\/ul><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-12\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%A7%B1_%D8%A8%D8%B9%D8%AF%DB%8C_%DA%86%DB%8C%D8%B3%D8%AA%D8%9F\" >\ud83e\uddf1 \u0628\u0639\u062f\u06cc \u0686\u06cc\u0633\u062a\u061f<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-13\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#%F0%9F%99%8C_%D9%85%D9%85%D9%86%D9%88%D9%86\" >\ud83d\ude4c \u0645\u0645\u0646\u0648\u0646<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-14\" href=\"https:\/\/nabfollower.com\/blog\/diagram2graph-fine-tuning-a-vision-language-model-to-extract-knowledge-graphs-from-diagrams-5f7p\/#thoughts_%D8%A7%D9%81%DA%A9%D8%A7%D8%B1_%D9%86%D9%87%D8%A7%DB%8C%DB%8C\" >thoughts \u0627\u0641\u06a9\u0627\u0631 \u0646\u0647\u0627\u06cc\u06cc<\/a><\/li><\/ul><\/nav><\/div>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%9A%80_tl_%D8%9B_%D8%AF%DA%A9%D8%AA%D8%B1\"><\/span>\n<p>  \ud83d\ude80 tl \u061b \u062f\u06a9\u062a\u0631<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u062a\u0628\u062f\u06cc\u0644 \u062a\u0635\u0627\u0648\u06cc\u0631 \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627\u06cc \u062c\u0631\u06cc\u0627\u0646 \u06cc\u0627 \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627\u06cc \u0641\u0631\u0622\u06cc\u0646\u062f \u0628\u0647 \u0637\u0648\u0631 \u0645\u0633\u062a\u0642\u06cc\u0645 <strong>JSON \u0633\u0627\u0632\u06af\u0627\u0631 \u0628\u0627 NEO4J<\/strong> \u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 \u06cc\u06a9 \u062a\u0646\u0638\u06cc\u0645 \u062f\u0642\u06cc\u0642 <strong>\u0645\u062f\u0644 \u0632\u0628\u0627\u0646 \u0628\u06cc\u0646\u0627\u06cc\u06cc<\/strong> (VLM). \u0645\u0627 \u0628\u0627 \u06a9\u0644\u0648\u062f 3.5 \u0634\u0631\u0648\u0639 \u06a9\u0631\u062f\u06cc\u0645 \u060c \u0627\u0645\u0627 \u062a\u0646\u0638\u06cc\u0645 \u062f\u0642\u06cc\u0642 <strong>QWEN2.5-VL-3B<\/strong> \u0645\u062f\u0644 \u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 Peft (Lora) \u0648 Got <strong>+23 \u066a \u0628\u0647\u0628\u0648\u062f \u062f\u0631 \u062a\u0634\u062e\u06cc\u0635 \u0644\u0628\u0647<\/strong>\u0628\u0634\u0631<\/p>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%A7%A9_%D9%85%D8%B4%DA%A9%D9%84\"><\/span>\n<p>  \ud83e\udde9 \u0645\u0634\u06a9\u0644<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u0645\u0627 \u063a\u0627\u0644\u0628\u0627\u064b \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627\u06cc \u0641\u0646\u06cc \u060c \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627\u06cc \u0646\u0645\u0648\u062f\u0627\u0631 \u0648 \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627\u06cc \u0645\u0633\u062f\u0648\u062f \u06a9\u0646\u0646\u062f\u0647 \u062f\u0631 PDF \u060c \u062a\u062e\u062a\u0647 \u0647\u0627\u06cc \u0633\u0641\u06cc\u062f \u06cc\u0627 \u0627\u0633\u0646\u0627\u062f \u0627\u0633\u06a9\u0646 \u0634\u062f\u0647 \u067e\u06cc\u062f\u0627 \u0645\u06cc \u06a9\u0646\u06cc\u0645. \u0622\u0646\u0647\u0627 \u062d\u0627\u0648\u06cc \u0645\u0646\u0637\u0642 \u0648 \u0631\u0648\u0627\u0628\u0637 \u0627\u0631\u0632\u0634\u0645\u0646\u062f \u0647\u0633\u062a\u0646\u062f \u060c \u0627\u0645\u0627 \u0622\u0646\u0647\u0627 \u0642\u0627\u0628\u0644 \u067e\u0631\u0633\u0634 \u06cc\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0646\u06cc\u0633\u062a\u0646\u062f \u0645\u06af\u0631 \u0627\u06cc\u0646\u06a9\u0647 \u0628\u0647 \u0635\u0648\u0631\u062a \u062f\u0633\u062a\u06cc \u0627\u0633\u062a\u062e\u0631\u0627\u062c \u0634\u0648\u0646\u062f.<\/p>\n<blockquote>\n<p>\u0622\u06cc\u0627 \u0645\u06cc \u062a\u0648\u0627\u0646\u06cc\u0645 \u0627\u06cc\u0646 \u0627\u0633\u062a\u062e\u0631\u0627\u062c \u0646\u0645\u0648\u062f\u0627\u0631 \u0628\u0647 \u0646\u0645\u0648\u062f\u0627\u0631 \u0631\u0627 \u062e\u0648\u062f\u06a9\u0627\u0631 \u06a9\u0646\u06cc\u0645\u061f<\/p>\n<\/blockquote>\n<p>\u0627\u06cc\u0646 \u0686\u06cc\u0632\u06cc \u0627\u0633\u062a \u06a9\u0647 <strong>\u0646\u0645\u0648\u062f\u0627\u0631<\/strong> \u0627\u0646\u062c\u0627\u0645 \u0645\u06cc \u062f\u0647\u062f.<\/p>\n<p>\u062f\u0631 \u0627\u06cc\u0646 \u067e\u0633\u062a \u060c \u0645\u0627 \u0646\u0634\u0627\u0646 \u0645\u06cc \u062f\u0647\u06cc\u0645 \u06a9\u0647 \u0686\u06af\u0648\u0646\u0647 \u0645\u06cc \u062a\u0648\u0627\u0646 \u0627\u0632 \u06cc\u06a9 \u0645\u062f\u0644 \u0632\u0628\u0627\u0646 \u0628\u06cc\u0646\u0627\u06cc\u06cc \u062a\u0646\u0638\u06cc\u0645 \u0634\u062f\u0647 (VLM) \u0628\u0631\u0627\u06cc \u062a\u0628\u062f\u06cc\u0644 \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627\u06cc \u0641\u0646\u06cc \u0648 \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627 \u0628\u0647 JSON \u0633\u0627\u062e\u062a\u0627\u0631\u06cc \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u06a9\u0631\u062f. \u0627\u06cc\u0646 \u062e\u0631\u0648\u062c\u06cc \u0645\u06cc \u062a\u0648\u0627\u0646\u062f \u0628\u0647 \u0637\u0648\u0631 \u0645\u0633\u062a\u0642\u06cc\u0645 \u0628\u0627 NEO4J \u06cc\u0627 \u0633\u0627\u06cc\u0631 \u0633\u06cc\u0633\u062a\u0645 \u0639\u0627\u0645\u0644 \u0647\u0627\u06cc \u0646\u0645\u0648\u062f\u0627\u0631 \u062f\u0627\u0646\u0634 \u0645\u0648\u0631\u062f \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0642\u0631\u0627\u0631 \u06af\u06cc\u0631\u062f \u0648 \u0633\u06cc\u0633\u062a\u0645 \u0647\u0627\u06cc AI \u0631\u0627 \u0642\u0627\u062f\u0631 \u0645\u06cc \u0633\u0627\u0632\u062f \u062a\u0627 \u0627\u0637\u0644\u0627\u0639\u0627\u062a \u0628\u0635\u0631\u06cc \u0631\u0627 \u0627\u0633\u062a\u062f\u0644\u0627\u0644 \u06a9\u0646\u0646\u062f.<\/p>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%94%8D_%D8%A2%D9%86%DA%86%D9%87_%D9%85%D8%A7_%D8%B3%D8%A7%D8%AE%D8%AA%DB%8C%D9%85\"><\/span>\n<p>  \ud83d\udd0d \u0622\u0646\u0686\u0647 \u0645\u0627 \u0633\u0627\u062e\u062a\u06cc\u0645<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u06cc\u06a9 \u062a\u0646\u0638\u06cc\u0645 \u062f\u0642\u06cc\u0642 <strong>\u0645\u062f\u0644 \u0632\u0628\u0627\u0646 \u0628\u06cc\u0646\u0627\u06cc\u06cc<\/strong> \u06a9\u0647:<\/p>\n<ul>\n<li>\u062a\u0635\u0648\u06cc\u0631\u06cc \u0627\u0632 \u0646\u0645\u0648\u062f\u0627\u0631 \u0631\u0627 \u0645\u06cc \u067e\u0630\u06cc\u0631\u062f<\/li>\n<li>\u0639\u0635\u0627\u0631\u0647 <strong>\u06af\u0631\u0647<\/strong>\u0628\u0627 <strong>\u0644\u0628\u0647<\/strong>\u0648\u062a <strong>\u0627\u0628\u0631\u062f\u0627\u062f\u0647<\/strong>\n<\/li>\n<li>\u062e\u0631\u0648\u062c\u06cc \u0647\u0627\u06cc \u0633\u0627\u062e\u062a\u0627\u0631\u06cc <strong>json<\/strong>\n<\/li>\n<li>\u0633\u0627\u0632\u06af\u0627\u0631 \u0628\u0627 <strong>\u0646\u0626\u0648 4J<\/strong> \u0628\u0631\u0627\u06cc \u067e\u0631\u0633 \u0648 \u062c\u0648 \u0646\u0645\u0648\u062f\u0627\u0631<\/li>\n<\/ul>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%94%97_%D9%85%D9%86%D8%A7%D8%A8%D8%B9\"><\/span>\n<p>  \ud83d\udd17 \u0645\u0646\u0627\u0628\u0639<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%E2%9A%99_%D9%85%D8%B9%D9%85%D8%A7%D8%B1%DB%8C_%D8%AF%D8%B1_%DB%8C%DA%A9_%D9%86%DA%AF%D8%A7%D9%87\"><\/span>\n<p>  \u2699 \u0645\u0639\u0645\u0627\u0631\u06cc \u062f\u0631 \u06cc\u06a9 \u0646\u06af\u0627\u0647<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<div class=\"highlight js-code-highlight\">\n<pre class=\"highlight plaintext\"><code>+-------------------+         +---------------------------+\n|   Diagram Image   +-------&gt; |  Fine-Tuned VLM (Qwen2.5) |\n+-------------------+         +-------------+-------------+\n                                            |\n                                            v\n                           +------------------------------+\n                           |  JSON (Nodes + Edges + Meta) |\n                           +------------------------------+\n                                            |\n                                            v\n                              +------------------------+\n                              |     Neo4J Integration  |\n                              +------------------------+\n<\/code><\/pre>\n<div class=\"highlight__panel js-actions-panel\">\n<div class=\"highlight__panel-action js-fullscreen-code-action\">\n    <svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" class=\"highlight-action crayons-icon highlight-action--fullscreen-on\"><title>\u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0631\u0627 \u0648\u0627\u0631\u062f \u06a9\u0646\u06cc\u062f<\/title>\n    <path d=\"M16 3h6v6h-2V5h-4V3zM2 3h6v2H4v4H2V3zm18 16v-4h2v6h-6v-2h4zM4 19h4v2H2v-6h2v4z\"\/>\n<\/svg><\/p>\n<p>    <svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" class=\"highlight-action crayons-icon highlight-action--fullscreen-off\"><title>\u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    <path d=\"M18 7h4v2h-6V3h2v4zM8 9H2V7h4V3h2v6zm10 8v4h-2v-6h6v2h-4zM8 15v6H6v-4H2v-2h6z\"\/>\n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%A4%96_%DA%86%D8%B1%D8%A7_%D8%A7%D8%B2_GPT-4_%DB%8C%D8%A7_%DA%A9%D9%84%D9%88%D8%AF_%D8%A7%D8%B3%D8%AA%D9%81%D8%A7%D8%AF%D9%87_%D9%86%D9%85%DB%8C_%DA%A9%D9%86%DB%8C%D9%85%D8%9F\"><\/span>\n<p>  \ud83e\udd16 \u0686\u0631\u0627 \u0627\u0632 GPT-4 \u06cc\u0627 \u06a9\u0644\u0648\u062f \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0646\u0645\u06cc \u06a9\u0646\u06cc\u0645\u061f<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u0622\u0646\u0647\u0627 \u06a9\u0627\u0631 \u0645\u06cc \u06a9\u0646\u0646\u062f &#8211; \u0627\u0645\u0627 \u0622\u0646\u0647\u0627:<\/p>\n<ul>\n<li>\n<strong>\u0648\u0627\u0628\u0633\u062a\u0647 \u0628\u0647 API<\/strong> (\u0646\u06af\u0631\u0627\u0646\u06cc \u0647\u0627\u06cc \u0645\u0631\u0628\u0648\u0637 \u0628\u0647 \u062d\u0631\u06cc\u0645 \u062e\u0635\u0648\u0635\u06cc)<\/li>\n<li>\n<strong>\u0639\u0645\u0648\u0645\u06cc<\/strong> (\u0645\u0633\u062a\u0639\u062f \u062a\u0648\u0647\u0645)<\/li>\n<li>\n<strong>\u06af\u0631\u0627\u0646<\/strong> (\u062a\u0648\u06a9\u0646 + \u0645\u062d\u0627\u0633\u0628\u0647)<\/li>\n<\/ul>\n<p>\u0645\u0627 \u06cc\u06a9 \u0645\u062f\u0644 \u0632\u0628\u0627\u0646 \u0628\u06cc\u0646\u0627\u06cc\u06cc \u062e\u0627\u0635 \u06a9\u0627\u0631 (QWEN2.5-VL 3B) \u0631\u0627 \u0628\u0631\u0627\u06cc \u062f\u0631\u06a9 \u0646\u0645\u0648\u062f\u0627\u0631 \u0648 \u0627\u0633\u062a\u062e\u0631\u0627\u062c \u0646\u0645\u0648\u062f\u0627\u0631 \u062f\u0627\u0646\u0634 \u062a\u0646\u0638\u06cc\u0645 \u06a9\u0631\u062f\u06cc\u0645.<\/p>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%D9%BE%D8%B4%D8%AA%D9%87_%D9%81%D9%86%DB%8C\"><\/span>\n<p>  \u067e\u0634\u062a\u0647 \u0641\u0646\u06cc<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<ul>\n<li>\u0645\u062f\u0644: <code>Qwen2.5-VL-3B<\/code>\n<\/li>\n<li>\u062a\u0646\u0638\u06cc\u0645 \u062f\u0642\u06cc\u0642: Peft (\u0644\u0648\u0631\u0627) \u060c <code>f32<\/code>\u060c \u0635\u0627\u0639\u0642\u0647 \u067e\u06cc\u062a\u0648\u0631\u0686<\/li>\n<li>\u0645\u062c\u0645\u0648\u0639\u0647 \u062f\u0627\u062f\u0647 \u0647\u0627: 218 \u062a\u0635\u0627\u0648\u06cc\u0631 \u0646\u0645\u0648\u062f\u0627\u0631 \u0628\u0631\u0686\u0633\u0628<\/li>\n<li>Backend: Fastapi + Neo4J (\u0627\u0632 \u0637\u0631\u06cc\u0642 Cypher)<\/li>\n<li>\u0627\u0633\u062a\u0646\u0628\u0627\u0637: \u0628\u063a\u0644 \u06a9\u0631\u062f\u0646 \u062a\u0631\u0627\u0646\u0633\u0641\u0648\u0631\u0645\u0627\u062a\u0648\u0631\u0647\u0627\u06cc \u0635\u0648\u0631\u062a<\/li>\n<li>Frontend: NextJS (WIP)<\/li>\n<li>\u0627\u0633\u062a\u0642\u0631\u0627\u0631: Kaggle + Lightning.ai<\/li>\n<\/ul>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%93%8A_%D9%86%D8%AA%D8%A7%DB%8C%D8%AC\"><\/span>\n<p>  \ud83d\udcca \u0646\u062a\u0627\u06cc\u062c<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<div class=\"table-wrapper-paragraph\">\n<table>\n<thead>\n<tr>\n<th>\u0648\u0638\u06cc\u0641\u0647<\/th>\n<th>\u0645\u062f\u0644 \u067e\u0627\u06cc\u0647 (\u06a9\u0644\u0648\u062f 3.5)<\/th>\n<th>QWEN2.5-VL-3B (\u062a\u0646\u0638\u06cc\u0645 \u062f\u0642\u06cc\u0642)<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>\u062a\u0634\u062e\u06cc\u0635 \u06af\u0631\u0647<\/td>\n<td>74.9 \u066a F1<\/td>\n<td>\n<strong>89.1 \u066a F1<\/strong> (+14.2 \u066a)<\/td>\n<\/tr>\n<tr>\n<td>\u062a\u0634\u062e\u06cc\u0635 \u0644\u0628\u0647<\/td>\n<td>46.05 \u066a F1<\/td>\n<td>\n<strong>69.45 \u066a F1<\/strong> (+23.4 \u066a)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%A7%A0_%D8%AC%D8%B2%D8%A6%DB%8C%D8%A7%D8%AA_%D8%A2%D9%85%D9%88%D8%B2%D8%B4\"><\/span>\n<p>  \ud83e\udde0 \u062c\u0632\u0626\u06cc\u0627\u062a \u0622\u0645\u0648\u0632\u0634<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<div class=\"table-wrapper-paragraph\">\n<table>\n<thead>\n<tr>\n<th>\u067e\u06cc\u06a9\u0631\u0628\u0646\u062f\u06cc<\/th>\n<th>\u0627\u0631\u0632\u0634<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>\u062f\u0648\u0631\u0647<\/td>\n<td>10<\/td>\n<\/tr>\n<tr>\n<td>\u062a\u0635\u0627\u0648\u06cc\u0631<\/td>\n<td>200 (\u0628\u0631\u0686\u0633\u0628 \u062f\u0633\u062a\u06cc)<\/td>\n<\/tr>\n<tr>\n<td>\u0627\u0646\u062f\u0627\u0632\u0647 \u062f\u0633\u062a\u0647<\/td>\n<td>2<\/td>\n<\/tr>\n<tr>\n<td>\u0631\u0648\u0634<\/td>\n<td>\u0644\u0648\u0631\u0627 (\u067e\u0641\u062a)<\/td>\n<\/tr>\n<tr>\n<td>\u062f\u0642\u062a<\/td>\n<td>BF16<\/td>\n<\/tr>\n<tr>\n<td>GPU<\/td>\n<td>L40S (48 \u06af\u06cc\u06af\u0627\u0628\u0627\u06cc\u062a VRAM)<\/td>\n<\/tr>\n<tr>\n<td>\u0645\u062a\u0631\u06cc\u06a9 \u0627\u0631\u0632\u06cc\u0627\u0628\u06cc<\/td>\n<td>\u0648\u06cc\u0631\u0627\u06cc\u0634 \u0641\u0627\u0635\u0644\u0647 + F1<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%A7%AA_%D8%AE%D9%88%D8%AF%D8%AA%D8%A7%D9%86_%D8%A2%D9%86_%D8%B1%D8%A7_%D8%A7%D9%85%D8%AA%D8%AD%D8%A7%D9%86_%DA%A9%D9%86%DB%8C%D8%AF\"><\/span>\n<p>  \ud83e\uddea \u062e\u0648\u062f\u062a\u0627\u0646 \u0622\u0646 \u0631\u0627 \u0627\u0645\u062a\u062d\u0627\u0646 \u06a9\u0646\u06cc\u062f<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u0646\u0648\u062a \u0628\u0648\u06a9 COLAB (\u0641\u0642\u0637 \u0627\u0633\u062a\u0646\u062a\u0627\u062c):<\/p>\n<p>\ud83d\udc49 \u0646\u0648\u062a \u0628\u0648\u06a9 \u0628\u0627\u0632 \u06a9\u0646\u06cc\u062f<\/p>\n<div class=\"highlight js-code-highlight\">\n<pre class=\"highlight shell\"><code>pip <span class=\"nb\">install<\/span> <span class=\"nt\">-q<\/span> transformers accelerate qwen-vl-utils[decord]<span class=\"o\">==<\/span>0.0.8\n<\/code><\/pre>\n<div class=\"highlight__panel js-actions-panel\">\n<div class=\"highlight__panel-action js-fullscreen-code-action\">\n    <svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" class=\"highlight-action crayons-icon highlight-action--fullscreen-on\"><title>\u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0631\u0627 \u0648\u0627\u0631\u062f \u06a9\u0646\u06cc\u062f<\/title>\n    <path d=\"M16 3h6v6h-2V5h-4V3zM2 3h6v2H4v4H2V3zm18 16v-4h2v6h-6v-2h4zM4 19h4v2H2v-6h2v4z\"\/>\n<\/svg><\/p>\n<p>    <svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" class=\"highlight-action crayons-icon highlight-action--fullscreen-off\"><title>\u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    <path d=\"M18 7h4v2h-6V3h2v4zM8 9H2V7h4V3h2v6zm10 8v4h-2v-6h6v2h-4zM8 15v6H6v-4H2v-2h6z\"\/>\n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n<h3><span class=\"ez-toc-section\" id=\"%D9%82%D8%B7%D8%B9%D9%87_%D8%A7%D8%B3%D8%AA%D9%86%D8%A8%D8%A7%D8%B7\"><\/span>\n<p>  \u0642\u0637\u0639\u0647 \u0627\u0633\u062a\u0646\u0628\u0627\u0637:<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h3>\n<div class=\"highlight js-code-highlight\">\n<pre class=\"highlight python\"><code><span class=\"kn\">import<\/span> <span class=\"n\">torch<\/span>\n<span class=\"kn\">from<\/span> <span class=\"n\">transformers<\/span> <span class=\"kn\">import<\/span> <span class=\"n\">Qwen2_5_VLForConditionalGeneration<\/span><span class=\"p\">,<\/span> <span class=\"n\">Qwen2_5_VLProcessor<\/span>\n\n\n<span class=\"n\">MODEL_ID<\/span><span class=\"o\">=<\/span><span class=\"sh\">\"<\/span><span class=\"s\">zackriya\/diagram2graph-adapters<\/span><span class=\"sh\">\"<\/span>\n<span class=\"n\">MAX_PIXELS<\/span> <span class=\"o\">=<\/span> <span class=\"mi\">1280<\/span> <span class=\"o\">*<\/span> <span class=\"mi\">28<\/span> <span class=\"o\">*<\/span> <span class=\"mi\">28<\/span>\n<span class=\"n\">MIN_PIXELS<\/span> <span class=\"o\">=<\/span> <span class=\"mi\">256<\/span> <span class=\"o\">*<\/span> <span class=\"mi\">28<\/span> <span class=\"o\">*<\/span> <span class=\"mi\">28<\/span>\n\n\n<span class=\"n\">model<\/span> <span class=\"o\">=<\/span> <span class=\"n\">Qwen2_5_VLForConditionalGeneration<\/span><span class=\"p\">.<\/span><span class=\"nf\">from_pretrained<\/span><span class=\"p\">(<\/span>\n    <span class=\"n\">MODEL_ID<\/span><span class=\"p\">,<\/span>\n    <span class=\"n\">device_map<\/span><span class=\"o\">=<\/span><span class=\"sh\">\"<\/span><span class=\"s\">auto<\/span><span class=\"sh\">\"<\/span><span class=\"p\">,<\/span>\n    <span class=\"n\">torch_dtype<\/span><span class=\"o\">=<\/span><span class=\"n\">torch<\/span><span class=\"p\">.<\/span><span class=\"n\">bfloat16<\/span>\n<span class=\"p\">)<\/span>\n\n<span class=\"n\">processor<\/span> <span class=\"o\">=<\/span> <span class=\"n\">Qwen2_5_VLProcessor<\/span><span class=\"p\">.<\/span><span class=\"nf\">from_pretrained<\/span><span class=\"p\">(<\/span>\n    <span class=\"n\">MODEL_ID<\/span><span class=\"p\">,<\/span>\n    <span class=\"n\">min_pixels<\/span><span class=\"o\">=<\/span><span class=\"n\">MIN_PIXELS<\/span><span class=\"p\">,<\/span>\n    <span class=\"n\">max_pixels<\/span><span class=\"o\">=<\/span><span class=\"n\">MAX_PIXELS<\/span>\n<span class=\"p\">)<\/span>\n<\/code><\/pre>\n<div class=\"highlight__panel js-actions-panel\">\n<div class=\"highlight__panel-action js-fullscreen-code-action\">\n    <svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" class=\"highlight-action crayons-icon highlight-action--fullscreen-on\"><title>\u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0631\u0627 \u0648\u0627\u0631\u062f \u06a9\u0646\u06cc\u062f<\/title>\n    <path d=\"M16 3h6v6h-2V5h-4V3zM2 3h6v2H4v4H2V3zm18 16v-4h2v6h-6v-2h4zM4 19h4v2H2v-6h2v4z\"\/>\n<\/svg><\/p>\n<p>    <svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" class=\"highlight-action crayons-icon highlight-action--fullscreen-off\"><title>\u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    <path d=\"M18 7h4v2h-6V3h2v4zM8 9H2V7h4V3h2v6zm10 8v4h-2v-6h6v2h-4zM8 15v6H6v-4H2v-2h6z\"\/>\n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n<div class=\"highlight js-code-highlight\">\n<pre class=\"highlight python\"><code><span class=\"kn\">from<\/span> <span class=\"n\">qwen_vl_utils<\/span> <span class=\"kn\">import<\/span> <span class=\"n\">process_vision_info<\/span>\n\n<span class=\"n\">SYSTEM_MESSAGE<\/span> <span class=\"o\">=<\/span> <span class=\"sh\">\"\"\"<\/span><span class=\"s\">You are a Vision Language Model specialized in extracting structured data from visual representations of process and flow diagrams.\nYour task is to analyze the provided image of a diagram and extract the relevant information into a well-structured JSON format.\nThe diagram includes details such as nodes and edges. each of them have their own attributes.\nFocus on identifying key data fields and ensuring the output adheres to the requested JSON structure.\nProvide only the JSON output based on the extracted information. Avoid additional explanations or comments.<\/span><span class=\"sh\">\"\"\"<\/span>\n\n<span class=\"k\">def<\/span> <span class=\"nf\">run_inference<\/span><span class=\"p\">(<\/span><span class=\"n\">image<\/span><span class=\"p\">):<\/span>\n  <span class=\"sh\">\"\"\"<\/span><span class=\"s\">\n  Inference with the Model\n  <\/span><span class=\"sh\">\"\"\"<\/span>\n  <span class=\"n\">messages<\/span><span class=\"o\">=<\/span> <span class=\"p\">[<\/span>\n      <span class=\"p\">{<\/span>\n          <span class=\"sh\">\"<\/span><span class=\"s\">role<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"sh\">\"<\/span><span class=\"s\">system<\/span><span class=\"sh\">\"<\/span><span class=\"p\">,<\/span>\n          <span class=\"sh\">\"<\/span><span class=\"s\">content<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"p\">[{<\/span><span class=\"sh\">\"<\/span><span class=\"s\">type<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"sh\">\"<\/span><span class=\"s\">text<\/span><span class=\"sh\">\"<\/span><span class=\"p\">,<\/span> <span class=\"sh\">\"<\/span><span class=\"s\">text<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"n\">SYSTEM_MESSAGE<\/span><span class=\"p\">}],<\/span>\n      <span class=\"p\">},<\/span>\n      <span class=\"p\">{<\/span>\n          <span class=\"sh\">\"<\/span><span class=\"s\">role<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"sh\">\"<\/span><span class=\"s\">user<\/span><span class=\"sh\">\"<\/span><span class=\"p\">,<\/span>\n          <span class=\"sh\">\"<\/span><span class=\"s\">content<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"p\">[<\/span>\n              <span class=\"p\">{<\/span>\n                  <span class=\"sh\">\"<\/span><span class=\"s\">type<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"sh\">\"<\/span><span class=\"s\">image<\/span><span class=\"sh\">\"<\/span><span class=\"p\">,<\/span>\n                  <span class=\"c1\"># this image is handled by qwen_vl_utils's process_visio_Info so no need to worry about pil image or path\n<\/span>                  <span class=\"sh\">\"<\/span><span class=\"s\">image<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"n\">image<\/span><span class=\"p\">,<\/span>\n              <span class=\"p\">},<\/span>\n              <span class=\"p\">{<\/span>\n                  <span class=\"sh\">\"<\/span><span class=\"s\">type<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"sh\">\"<\/span><span class=\"s\">text<\/span><span class=\"sh\">\"<\/span><span class=\"p\">,<\/span>\n                  <span class=\"sh\">\"<\/span><span class=\"s\">text<\/span><span class=\"sh\">\"<\/span><span class=\"p\">:<\/span> <span class=\"sh\">\"<\/span><span class=\"s\">Extract data in JSON format, Only give the JSON<\/span><span class=\"sh\">\"<\/span><span class=\"p\">,<\/span>\n              <span class=\"p\">},<\/span>\n          <span class=\"p\">],<\/span>\n      <span class=\"p\">},<\/span>\n  <span class=\"p\">]<\/span>\n\n  <span class=\"n\">text<\/span> <span class=\"o\">=<\/span> <span class=\"n\">processor<\/span><span class=\"p\">.<\/span><span class=\"nf\">apply_chat_template<\/span><span class=\"p\">(<\/span><span class=\"n\">messages<\/span><span class=\"p\">,<\/span> <span class=\"n\">tokenize<\/span><span class=\"o\">=<\/span><span class=\"bp\">False<\/span><span class=\"p\">,<\/span> <span class=\"n\">add_generation_prompt<\/span><span class=\"o\">=<\/span><span class=\"bp\">True<\/span><span class=\"p\">)<\/span>\n  <span class=\"n\">image_inputs<\/span><span class=\"p\">,<\/span> <span class=\"n\">_<\/span> <span class=\"o\">=<\/span> <span class=\"nf\">process_vision_info<\/span><span class=\"p\">(<\/span><span class=\"n\">messages<\/span><span class=\"p\">)<\/span>\n\n  <span class=\"n\">inputs<\/span> <span class=\"o\">=<\/span> <span class=\"nf\">processor<\/span><span class=\"p\">(<\/span>\n      <span class=\"n\">text<\/span><span class=\"o\">=<\/span><span class=\"p\">[<\/span><span class=\"n\">text<\/span><span class=\"p\">],<\/span>\n      <span class=\"n\">images<\/span><span class=\"o\">=<\/span><span class=\"n\">image_inputs<\/span><span class=\"p\">,<\/span>\n      <span class=\"n\">return_tensors<\/span><span class=\"o\">=<\/span><span class=\"sh\">\"<\/span><span class=\"s\">pt<\/span><span class=\"sh\">\"<\/span><span class=\"p\">,<\/span>\n  <span class=\"p\">)<\/span>\n  <span class=\"n\">inputs<\/span> <span class=\"o\">=<\/span> <span class=\"n\">inputs<\/span><span class=\"p\">.<\/span><span class=\"nf\">to<\/span><span class=\"p\">(<\/span><span class=\"sh\">'<\/span><span class=\"s\">cuda<\/span><span class=\"sh\">'<\/span><span class=\"p\">)<\/span>\n\n  <span class=\"n\">generated_ids<\/span> <span class=\"o\">=<\/span> <span class=\"n\">model<\/span><span class=\"p\">.<\/span><span class=\"nf\">generate<\/span><span class=\"p\">(<\/span><span class=\"o\">**<\/span><span class=\"n\">inputs<\/span><span class=\"p\">,<\/span> <span class=\"n\">max_new_tokens<\/span><span class=\"o\">=<\/span><span class=\"mi\">1024<\/span><span class=\"p\">)<\/span>\n  <span class=\"n\">generated_ids_trimmed<\/span> <span class=\"o\">=<\/span> <span class=\"p\">[<\/span>\n      <span class=\"n\">out_ids<\/span><span class=\"p\">[<\/span><span class=\"nf\">len<\/span><span class=\"p\">(<\/span><span class=\"n\">in_ids<\/span><span class=\"p\">):]<\/span>\n      <span class=\"k\">for<\/span> <span class=\"n\">in_ids<\/span><span class=\"p\">,<\/span> <span class=\"n\">out_ids<\/span>\n      <span class=\"ow\">in<\/span> <span class=\"nf\">zip<\/span><span class=\"p\">(<\/span><span class=\"n\">inputs<\/span><span class=\"p\">.<\/span><span class=\"n\">input_ids<\/span><span class=\"p\">,<\/span> <span class=\"n\">generated_ids<\/span><span class=\"p\">)<\/span>\n  <span class=\"p\">]<\/span>\n\n  <span class=\"n\">output_text<\/span> <span class=\"o\">=<\/span> <span class=\"n\">processor<\/span><span class=\"p\">.<\/span><span class=\"nf\">batch_decode<\/span><span class=\"p\">(<\/span>\n      <span class=\"n\">generated_ids_trimmed<\/span><span class=\"p\">,<\/span>\n      <span class=\"n\">skip_special_tokens<\/span><span class=\"o\">=<\/span><span class=\"bp\">True<\/span><span class=\"p\">,<\/span>\n      <span class=\"n\">clean_up_tokenization_spaces<\/span><span class=\"o\">=<\/span><span class=\"bp\">False<\/span>\n  <span class=\"p\">)<\/span>\n  <span class=\"k\">return<\/span> <span class=\"n\">output_text<\/span>\n\n<\/code><\/pre>\n<div class=\"highlight__panel js-actions-panel\">\n<div class=\"highlight__panel-action js-fullscreen-code-action\">\n    <svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" class=\"highlight-action crayons-icon highlight-action--fullscreen-on\"><title>\u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0631\u0627 \u0648\u0627\u0631\u062f \u06a9\u0646\u06cc\u062f<\/title>\n    <path d=\"M16 3h6v6h-2V5h-4V3zM2 3h6v2H4v4H2V3zm18 16v-4h2v6h-6v-2h4zM4 19h4v2H2v-6h2v4z\"\/>\n<\/svg><\/p>\n<p>    <svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" class=\"highlight-action crayons-icon highlight-action--fullscreen-off\"><title>\u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    <path d=\"M18 7h4v2h-6V3h2v4zM8 9H2V7h4V3h2v6zm10 8v4h-2v-6h6v2h-4zM8 15v6H6v-4H2v-2h6z\"\/>\n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%A7%B1_%D8%A8%D8%B9%D8%AF%DB%8C_%DA%86%DB%8C%D8%B3%D8%AA%D8%9F\"><\/span>\n<p>  \ud83e\uddf1 \u0628\u0639\u062f\u06cc \u0686\u06cc\u0633\u062a\u061f<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<ul>\n<li>[ ]  \u0627\u062f\u063a\u0627\u0645 Neo4J \u0627\u0632 \u0637\u0631\u06cc\u0642 Cypher Parser<\/li>\n<li>[ ]  \u0645\u062f\u0644 \u06a9\u0645\u06a9\u06cc \u0628\u0631\u0627\u06cc \u062f\u0633\u062a\u06af\u0627\u0647 \u0647\u0627\u06cc \u0644\u0628\u0647<\/li>\n<li>[ ]  Ollama \/ Python SDK \u0628\u0631\u0627\u06cc \u0627\u0641\u0632\u0648\u0646\u0647 \u0648 \u0628\u0627\u0632\u06cc<\/li>\n<li>[ ]  Frontend \u0628\u0631\u0627\u06cc \u0628\u0627\u0631\u06af\u0630\u0627\u0631\u06cc + \u0646\u0645\u0627\u06cc\u0634 \u062f\u0627\u062f\u0647 \u0634\u062f\u06af\u0627\u0646 \u0637\u0628\u06cc\u0639\u06cc<\/li>\n<\/ul>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"%F0%9F%99%8C_%D9%85%D9%85%D9%86%D9%88%D9%86\"><\/span>\n<p>  \ud83d\ude4c \u0645\u0645\u0646\u0648\u0646<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u0641\u0631\u06cc\u0627\u062f \u0628\u0647:<\/p>\n<ul>\n<li>\n<strong>\u0648\u0627\u0628\u0633\u062a\u0647 \u0628\u0647 \u0627\u0646\u0633\u0627\u0646 \u0634\u0646\u0627\u0633\u06cc<\/strong> \u0628\u0631\u0627\u06cc Claude API<\/li>\n<li>\n<strong>\u0628\u063a\u0644 \u06a9\u0631\u062f\u0646 \u0635\u0648\u0631\u062a<\/strong> \u0628\u0631\u0627\u06cc \u0645\u0627\u062f\u0648\u0646 \u0642\u0631\u0645\u0632 \u0645\u0646\u0628\u0639 \u0628\u0627\u0632<\/li>\n<li>\n<strong>Lightning.ai<\/strong> \u0628\u0631\u0627\u06cc GPU<\/li>\n<li>\n<strong>\u0631\u062d\u0645<\/strong> \u0628\u0631\u0627\u06cc \u0627\u0644\u0647\u0627\u0645 \u0628\u062e\u0634 \u062f\u0627\u062f\u0647<\/li>\n<\/ul>\n<hr\/>\n<h2><span class=\"ez-toc-section\" id=\"thoughts_%D8%A7%D9%81%DA%A9%D8%A7%D8%B1_%D9%86%D9%87%D8%A7%DB%8C%DB%8C\"><\/span>\n<p>  thoughts \u0627\u0641\u06a9\u0627\u0631 \u0646\u0647\u0627\u06cc\u06cc<br \/>\n<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>VLM \u0647\u0627\u06cc \u062e\u0627\u0635 \u0648\u0638\u06cc\u0641\u0647 \u0645\u0627\u0646\u0646\u062f <strong>\u0646\u0645\u0648\u062f\u0627\u0631<\/strong> \u06cc\u06a9 \u0632\u0645\u06cc\u0646 \u0645\u062a\u0648\u0633\u0637 \u200b\u200b\u0639\u0627\u0644\u06cc \u0647\u0633\u062a\u0646\u062f:<\/p>\n<p>\u06a9\u0648\u0686\u06a9\u062a\u0631 \u060c \u0633\u0631\u06cc\u0639\u062a\u0631 \u060c \u0627\u0631\u0632\u0627\u0646 \u062a\u0631 \u0648 \u0628\u0647 \u0637\u0631\u0632 \u0634\u06af\u0641\u062a \u0622\u0648\u0631 \u062f\u0642\u06cc\u0642.<\/p>\n<p>\u0628\u0647 \u062c\u0627\u06cc \u0627\u06cc\u0646\u06a9\u0647 \u0645\u0646\u062a\u0638\u0631 \u0645\u062f\u0644 \u0647\u0627\u06cc \u0628\u0646\u06cc\u0627\u062f\u06cc \u0628\u0627\u0634\u06cc\u062f \u062a\u0627 &#8220;\u0628\u0647\u062a\u0631 \u0634\u0648\u0646\u062f&#8221; \u060c \u0628\u06cc\u0627\u06cc\u06cc\u062f \u0622\u0646\u0647\u0627 \u0631\u0627 \u0622\u0645\u0648\u0632\u0634 \u062f\u0647\u06cc\u0645 <em>\u0648\u0638\u0627\u06cc\u0641 \u0645\u0627<\/em>\u0628\u0634\u0631<\/p>\n<blockquote>\n<p>\u0645\u062f\u0644 \u0631\u0627 \u062a\u0646\u0638\u06cc\u0645 \u06a9\u0646\u06cc\u062f. \u0635\u0627\u062d\u0628 \u06af\u0631\u062f\u0634 \u06a9\u0627\u0631 \u0628\u0627\u0634\u06cc\u062f.<\/p>\n<\/blockquote>\n<p>\u0627\u06af\u0631 \u0627\u06cc\u0646 \u0645\u0641\u06cc\u062f \u0628\u0648\u062f \u060c \u0628\u0647 repo github \u2b50 \u2b50 \u2b50 \u2b50 \u0648 \u0628\u0647 \u0631\u0648\u0632\u0631\u0633\u0627\u0646\u06cc \u0647\u0627 \u0631\u0627 \u062f\u0646\u0628\u0627\u0644 \u06a9\u0646\u06cc\u062f!<\/p>\n<p>\u0627\u06af\u0631 \u0645\u06cc \u062e\u0648\u0627\u0647\u06cc\u062f \u062f\u0631 \u0632\u0645\u06cc\u0646\u0647 \u0633\u0627\u062e\u062a \u0645\u062d\u0635\u0648\u0644\u0627\u062a AI \u0647\u0645\u06a9\u0627\u0631\u06cc \u06a9\u0646\u06cc\u062f &#8211; \u0628\u0627 \u0645\u0627 \u062a\u0645\u0627\u0633 \u0628\u06af\u06cc\u0631\u06cc\u062f &#8211; \u0628\u0627 \u0645\u0627 \u062a\u0645\u0627\u0633 \u0628\u06af\u06cc\u0631\u06cc\u062f<\/p>\n<\/p><\/div>\n","protected":false},"excerpt":{"rendered":"<p>\u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 QWEN2.5-VL + PEFT + NEO4J \u062a\u0646\u0638\u06cc\u0645 \u0634\u062f\u0647 \u062a\u0648\u0633\u0637 \u0645\u062d\u0645\u062f \u0635\u0641\u0648\u0627\u0646 | \u0631\u0627\u0647 \u062d\u0644 \u0647\u0627\u06cc Zackriya \ud83d\ude80 tl \u061b \u062f\u06a9\u062a\u0631 \u062a\u0628\u062f\u06cc\u0644 \u062a\u0635\u0627\u0648\u06cc\u0631 \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627\u06cc \u062c\u0631\u06cc\u0627\u0646 \u06cc\u0627 \u0646\u0645\u0648\u062f\u0627\u0631\u0647\u0627\u06cc \u0641\u0631\u0622\u06cc\u0646\u062f \u0628\u0647 \u0637\u0648\u0631 \u0645\u0633\u062a\u0642\u06cc\u0645 JSON \u0633\u0627\u0632\u06af\u0627\u0631 \u0628\u0627 NEO4J \u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 \u06cc\u06a9 \u062a\u0646\u0638\u06cc\u0645 \u062f\u0642\u06cc\u0642 \u0645\u062f\u0644 \u0632\u0628\u0627\u0646 \u0628\u06cc\u0646\u0627\u06cc\u06cc (VLM). \u0645\u0627 \u0628\u0627 \u06a9\u0644\u0648\u062f 3.5 \u0634\u0631\u0648\u0639 \u06a9\u0631\u062f\u06cc\u0645 \u060c \u0627\u0645\u0627 \u062a\u0646\u0638\u06cc\u0645 &hellip;<\/p>\n","protected":false},"author":2,"featured_media":104699,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"","fifu_image_alt":"","footnotes":""},"categories":[339],"tags":[],"class_list":["post-104698","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-dev"],"_links":{"self":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/posts\/104698","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/comments?post=104698"}],"version-history":[{"count":0,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/posts\/104698\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/media\/104699"}],"wp:attachment":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/media?parent=104698"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/categories?post=104698"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/tags?post=104698"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}