{"id":20851,"date":"2025-02-12T01:47:40","date_gmt":"2025-02-11T17:47:40","guid":{"rendered":"https:\/\/www.aisharenet.com\/?p=20851"},"modified":"2025-05-20T12:49:18","modified_gmt":"2025-05-20T04:49:18","slug":"deepseek-vl2","status":"publish","type":"post","link":"https:\/\/www.kdjingpai.com\/en\/deepseek-vl2\/","title":{"rendered":"DeepSeek-VL2"},"content":{"rendered":"<p>DeepSeek-VL2 \u662f\u4e00\u7cfb\u5217\u9ad8\u7ea7\u7684 Mixture-of-Experts (MoE) \u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff0c\u663e\u8457\u63d0\u5347\u4e86\u5176\u524d\u8eab DeepSeek-VL \u7684\u6027\u80fd\u3002\u8be5\u6a21\u578b\u5728\u89c6\u89c9\u95ee\u7b54\u3001\u5149\u5b66\u5b57\u7b26\u8bc6\u522b\u3001\u6587\u6863\/\u8868\u683c\/\u56fe\u8868\u7406\u89e3\u548c\u89c6\u89c9\u5b9a\u4f4d\u7b49\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u8272\u3002DeepSeek-VL2 \u7cfb\u5217\u5305\u62ec\u4e09\u79cd\u53d8\u4f53\uff1aDeepSeek-VL2-Tiny\u3001DeepSeek-VL2-Small \u548c DeepSeek-VL2\uff0c\u5206\u522b\u5177\u6709 1.0B\u30012.8B \u548c 4.5B \u4e2a\u6fc0\u6d3b\u53c2\u6570\u3002\u8be5\u6a21\u578b\u5728\u53c2\u6570\u6570\u91cf\u76f8\u4f3c\u6216\u66f4\u5c11\u7684\u60c5\u51b5\u4e0b\uff0c\u8fbe\u5230\u4e86\u4e0e\u73b0\u6709\u5f00\u6e90\u5bc6\u96c6\u548c MoE \u6a21\u578b\u76f8\u5ab2\u7f8e\u6216\u66f4\u4f18\u7684\u6027\u80fd\u3002<\/p>\n<div id=\"attachment_20853\" style=\"width: 1110px\" class=\"wp-caption aligncenter\"><img loading=\"lazy\" decoding=\"async\" aria-describedby=\"caption-attachment-20853\" class=\"wp-image-20853 size-full\" title=\"DeepSeek-VL2\uff1a\u9ad8\u7ea7\u591a\u6a21\u6001\u7406\u89e3\u7684\u4e13\u5bb6\u7ea7\u89c6\u89c9\u8bed\u8a00\u6a21\u578b-1\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/02\/3d420a5af72de60.jpg\" alt=\"DeepSeek-VL2\uff1a\u9ad8\u7ea7\u591a\u6a21\u6001\u7406\u89e3\u7684\u4e13\u5bb6\u7ea7\u89c6\u89c9\u8bed\u8a00\u6a21\u578b-1\" width=\"1100\" height=\"585\" srcset=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/02\/3d420a5af72de60.jpg 1100w, https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/02\/3d420a5af72de60-768x408.jpg 768w\" sizes=\"auto, (max-width: 1100px) 100vw, 1100px\" \/><p id=\"caption-attachment-20853\" class=\"wp-caption-text\">\u6f14\u793a\uff1ahttps:\/\/huggingface.co\/spaces\/deepseek-ai\/deepseek-vl2-small<\/p><\/div>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-full wp-image-20852\" title=\"DeepSeek-VL2\uff1a\u9ad8\u7ea7\u591a\u6a21\u6001\u7406\u89e3\u7684\u4e13\u5bb6\u7ea7\u89c6\u89c9\u8bed\u8a00\u6a21\u578b-1\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/02\/f798b56607836f0.jpg\" alt=\"DeepSeek-VL2\uff1a\u9ad8\u7ea7\u591a\u6a21\u6001\u7406\u89e3\u7684\u4e13\u5bb6\u7ea7\u89c6\u89c9\u8bed\u8a00\u6a21\u578b-1\" width=\"612\" height=\"504\" \/><\/p>\n<p>&nbsp;<\/p>\n<h2>\u529f\u80fd\u5217\u8868<\/h2>\n<ul>\n<li><strong>\u89c6\u89c9\u95ee\u7b54<\/strong>\uff1a\u652f\u6301\u590d\u6742\u7684\u89c6\u89c9\u95ee\u7b54\u4efb\u52a1\uff0c\u63d0\u4f9b\u51c6\u786e\u7684\u7b54\u6848\u3002<\/li>\n<li><strong>\u5149\u5b66\u5b57\u7b26\u8bc6\u522b (OCR)<\/strong>\uff1a\u9ad8\u6548\u8bc6\u522b\u56fe\u50cf\u4e2d\u7684\u6587\u672c\u5185\u5bb9\u3002<\/li>\n<li><strong>\u6587\u6863\u7406\u89e3<\/strong>\uff1a\u89e3\u6790\u548c\u7406\u89e3\u590d\u6742\u7684\u6587\u6863\u7ed3\u6784\u548c\u5185\u5bb9\u3002<\/li>\n<li><strong>\u8868\u683c\u7406\u89e3<\/strong>\uff1a\u8bc6\u522b\u548c\u5904\u7406\u8868\u683c\u6570\u636e\uff0c\u63d0\u53d6\u6709\u7528\u4fe1\u606f\u3002<\/li>\n<li><strong>\u56fe\u8868\u7406\u89e3<\/strong>\uff1a\u5206\u6790\u548c\u89e3\u91ca\u56fe\u8868\u4e2d\u7684\u6570\u636e\u548c\u8d8b\u52bf\u3002<\/li>\n<li><strong>\u89c6\u89c9\u5b9a\u4f4d<\/strong>\uff1a\u5728\u56fe\u50cf\u4e2d\u51c6\u786e\u5b9a\u4f4d\u76ee\u6807\u5bf9\u8c61\u3002<\/li>\n<li><strong>\u591a\u53d8\u4f53\u652f\u6301<\/strong>\uff1a\u63d0\u4f9b Tiny\u3001Small \u548c\u6807\u51c6\u7248\u4e09\u79cd\u6a21\u578b\uff0c\u6ee1\u8db3\u4e0d\u540c\u9700\u6c42\u3002<\/li>\n<li><strong>\u9ad8\u6548\u6027\u80fd<\/strong>\uff1a\u5728\u4fdd\u6301\u9ad8\u6027\u80fd\u7684\u540c\u65f6\uff0c\u51cf\u5c11\u4e86\u6fc0\u6d3b\u53c2\u6570\u6570\u91cf\u3002<\/li>\n<\/ul>\n<p>&nbsp;<\/p>\n<h2>\u4f7f\u7528\u5e2e\u52a9<\/h2>\n<h3>\u5b89\u88c5\u6d41\u7a0b<\/h3>\n<ol>\n<li>\u786e\u4fdd Python \u7248\u672c &gt;= 3.8\u3002<\/li>\n<li>\u514b\u9686 DeepSeek-VL2 \u4ed3\u5e93\uff1a<\/li>\n<\/ol>\n<pre><code>   git clone https:\/\/github.com\/deepseek-ai\/DeepSeek-VL2.git\r\n<\/code><\/pre>\n<ol start=\"3\">\n<li>\u8fdb\u5165\u9879\u76ee\u76ee\u5f55\u5e76\u5b89\u88c5\u5fc5\u8981\u4f9d\u8d56\uff1a<\/li>\n<\/ol>\n<pre><code>   cd DeepSeek-VL2\r\npip install -e .\r\n<\/code><\/pre>\n<h3>\u4f7f\u7528\u793a\u4f8b<\/h3>\n<h4>\u7b80\u5355\u63a8\u7406\u793a\u4f8b<\/h4>\n<p>\u4ee5\u4e0b\u662f\u4f7f\u7528 DeepSeek-VL2 \u8fdb\u884c\u7b80\u5355\u63a8\u7406\u7684\u793a\u4f8b\u4ee3\u7801\uff1a<\/p>\n<pre><code>import torch\r\nfrom transformers import AutoModelForCausalLM\r\nfrom deepseek_vl2.models import DeepseekVLV2Processor, DeepseekVLV2ForCausalLM\r\nfrom deepseek_vl2.utils.io import load_pil_images\r\n# \u6307\u5b9a\u6a21\u578b\u8def\u5f84\r\nmodel_path = \"deepseek-ai\/deepseek-vl2-tiny\"\r\nvl_chat_processor = DeepseekVLV2Processor.from_pretrained(model_path)\r\nvl_model = DeepseekVLV2ForCausalLM.from_pretrained(model_path)\r\n# \u52a0\u8f7d\u56fe\u50cf\r\nimages = load_pil_images([\"path_to_image.jpg\"])\r\n# \u63a8\u7406\r\ninputs = vl_chat_processor(images=images, return_tensors=\"pt\")\r\noutputs = vl_model.generate(**inputs)\r\nprint(outputs)\r\n<\/code><\/pre>\n<h3>\u8be6\u7ec6\u529f\u80fd\u64cd\u4f5c\u6d41\u7a0b<\/h3>\n<ol>\n<li><strong>\u89c6\u89c9\u95ee\u7b54<\/strong>\uff1a\n<ul>\n<li>\u52a0\u8f7d\u6a21\u578b\u548c\u5904\u7406\u5668\u3002<\/li>\n<li>\u8f93\u5165\u56fe\u50cf\u548c\u95ee\u9898\uff0c\u6a21\u578b\u5c06\u8fd4\u56de\u7b54\u6848\u3002<\/li>\n<\/ul>\n<\/li>\n<li><strong>\u5149\u5b66\u5b57\u7b26\u8bc6\u522b (OCR)<\/strong>\uff1a\n<ul>\n<li>\u4f7f\u7528 <code>DeepseekVLV2Processor<\/code> \u52a0\u8f7d\u56fe\u50cf\u3002<\/li>\n<li>\u8c03\u7528\u6a21\u578b\u8fdb\u884c\u63a8\u7406\uff0c\u63d0\u53d6\u56fe\u50cf\u4e2d\u7684\u6587\u672c\u3002<\/li>\n<\/ul>\n<\/li>\n<li><strong>\u6587\u6863\u7406\u89e3<\/strong>\uff1a\n<ul>\n<li>\u52a0\u8f7d\u5305\u542b\u6587\u6863\u56fe\u50cf\u7684\u8f93\u5165\u3002<\/li>\n<li>\u6a21\u578b\u89e3\u6790\u6587\u6863\u7ed3\u6784\uff0c\u8fd4\u56de\u89e3\u6790\u7ed3\u679c\u3002<\/li>\n<\/ul>\n<\/li>\n<li><strong>\u8868\u683c\u7406\u89e3<\/strong>\uff1a\n<ul>\n<li>\u8f93\u5165\u5305\u542b\u8868\u683c\u7684\u56fe\u50cf\u3002<\/li>\n<li>\u6a21\u578b\u8bc6\u522b\u8868\u683c\u7ed3\u6784\u548c\u5185\u5bb9\uff0c\u63d0\u53d6\u5173\u952e\u4fe1\u606f\u3002<\/li>\n<\/ul>\n<\/li>\n<li><strong>\u56fe\u8868\u7406\u89e3<\/strong>\uff1a\n<ul>\n<li>\u52a0\u8f7d\u56fe\u8868\u56fe\u50cf\u3002<\/li>\n<li>\u6a21\u578b\u5206\u6790\u56fe\u8868\u6570\u636e\uff0c\u63d0\u4f9b\u89e3\u91ca\u548c\u8d8b\u52bf\u5206\u6790\u3002<\/li>\n<\/ul>\n<\/li>\n<li><strong>\u89c6\u89c9\u5b9a\u4f4d<\/strong>\uff1a\n<ul>\n<li>\u8f93\u5165\u76ee\u6807\u5bf9\u8c61\u7684\u63cf\u8ff0\u548c\u56fe\u50cf\u3002<\/li>\n<li>\u6a21\u578b\u5728\u56fe\u50cf\u4e2d\u5b9a\u4f4d\u76ee\u6807\u5bf9\u8c61\uff0c\u8fd4\u56de\u4f4d\u7f6e\u5750\u6807\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ol>\n<p>\u901a\u8fc7\u4ee5\u4e0a\u6b65\u9aa4\uff0c\u7528\u6237\u53ef\u4ee5\u5145\u5206\u5229\u7528 DeepSeek-VL2 \u7684\u5f3a\u5927\u529f\u80fd\uff0c\u5b8c\u6210\u5404\u79cd\u590d\u6742\u7684\u89c6\u89c9\u8bed\u8a00\u4efb\u52a1\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>DeepSeek-VL2 \u662f\u4e00\u7cfb\u5217\u9ad8\u7ea7\u7684 Mixture-of-Experts (MoE) \u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff0c\u663e\u8457\u63d0\u5347\u4e86\u5176\u524d\u8eab DeepSeek-VL \u7684\u6027\u80fd\u3002\u8be5\u6a21\u578b\u5728\u89c6\u89c9\u95ee\u7b54\u3001\u5149\u5b66\u5b57\u7b26\u8bc6\u522b\u3001\u6587\u6863\/\u8868\u683c\/\u56fe\u8868\u7406\u89e3\u548c\u89c6\u89c9\u5b9a\u4f4d\u7b49\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u8272\u3002De&#8230;<\/p>\n","protected":false},"author":1,"featured_media":61820,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[392,401],"tags":[230,374],"class_list":["post-20851","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-models","category-visual-recognition-model","tag-aikaiyuanxiangmu","tag-duandaoduanyuyinzhubei"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/posts\/20851","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/comments?post=20851"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/posts\/20851\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/media\/61820"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/media?parent=20851"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/categories?post=20851"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/tags?post=20851"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}