{"id":58,"date":"2024-07-21T17:02:46","date_gmt":"2024-07-21T09:02:46","guid":{"rendered":"https:\/\/www.onetts.com\/ai\/?p=58"},"modified":"2024-07-22T09:29:49","modified_gmt":"2024-07-22T01:29:49","slug":"qwen-vl-chat","status":"publish","type":"post","link":"https:\/\/www.onetts.com\/ai\/models\/qwen-vl-chat\/","title":{"rendered":"Qwen-VL-Chat"},"content":{"rendered":"<p data-spm-anchor-id=\"a2c6h.13066512.0.i3.763536afIDGuJU\">Qwen-VL-Chat \u662f\u963f\u91cc\u4e91\u63a8\u51fa\u7684\u4e00\u6b3e\u5148\u8fdb\u7684\u5927\u89c4\u6a21\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff08LVLM\uff09\uff0c\u65e8\u5728\u901a\u8fc7\u7ed3\u5408\u56fe\u50cf\u548c\u6587\u672c\u7684\u8f93\u5165\uff0c\u63d0\u4f9b\u66f4\u4e30\u5bcc\u7684\u591a\u6a21\u6001\u4ea4\u4e92\u4f53\u9a8c\u3002\u8be5\u6a21\u578b\u4e0d\u4ec5\u5728\u6280\u672f\u4e0a\u5177\u6709\u521b\u65b0\u6027\uff0c\u8fd8\u5728\u591a\u8bed\u8a00\u652f\u6301\u548c\u7ec6\u7c92\u5ea6\u8bc6\u522b\u65b9\u9762\u8868\u73b0\u51fa\u8272\u3002<\/p>\n<h3 id=\"-\">\u5173\u952e\u7279\u6027<\/h3>\n<ul>\n<li><strong>\u591a\u6a21\u6001\u8f93\u5165\u8f93\u51fa<\/strong>\uff1aQwen-VL-Chat \u80fd\u591f\u5904\u7406\u56fe\u50cf\u3001\u6587\u672c\u548c\u68c0\u6d4b\u6846\u4f5c\u4e3a\u8f93\u5165\uff0c\u5e76\u751f\u6210\u6587\u672c\u548c\u68c0\u6d4b\u6846\u4f5c\u4e3a\u8f93\u51fa\u3002<\/li>\n<li><strong>\u591a\u8bed\u8a00\u5bf9\u8bdd<\/strong>\uff1a\u652f\u6301\u4e2d\u82f1\u53cc\u8bed\u7684\u957f\u6587\u672c\u8bc6\u522b\uff0c\u589e\u5f3a\u4e86\u8de8\u8bed\u8a00\u7684\u4ea4\u4e92\u80fd\u529b\u3002<\/li>\n<li><strong>\u591a\u56fe\u4ea4\u4e92<\/strong>\uff1a\u652f\u6301\u591a\u56fe\u8f93\u5165\u548c\u6bd4\u8f83\uff0c\u80fd\u591f\u8fdb\u884c\u56fe\u7247\u95f4\u7684\u95ee\u7b54\u548c\u6587\u5b66\u521b\u4f5c\u3002<\/li>\n<li><strong>\u4e2d\u6587\u5b9a\u4f4d\u652f\u6301<\/strong>\uff1a\u901a\u8fc7\u4e2d\u6587\u5f00\u653e\u57df\u8bed\u8a00\u8868\u8fbe\u8fdb\u884c\u68c0\u6d4b\u6846\u6807\u6ce8\uff0c\u63d0\u5347\u4e86\u6a21\u578b\u7684\u672c\u5730\u5316\u5e94\u7528\u80fd\u529b\u3002<\/li>\n<li><strong>\u9ad8\u5206\u8fa8\u7387\u5904\u7406<\/strong>\uff1a448\u5206\u8fa8\u7387\u7684\u56fe\u50cf\u8f93\u5165\uff0c\u63d0\u5347\u4e86\u6a21\u578b\u5728\u7ec6\u7c92\u5ea6\u6587\u5b57\u8bc6\u522b\u548c\u6587\u6863\u95ee\u7b54\u65b9\u9762\u7684\u8868\u73b0\u3002<\/li>\n<\/ul>\n<h2 id=\"-\">\u6a21\u578b\u8bc4\u6d4b<\/h2>\n<p>Qwen-VL-Chat \u5728\u591a\u9879\u6807\u51c6\u82f1\u6587\u8bc4\u6d4b\u57fa\u51c6\u4e0a\u5c55\u793a\u4e86\u5176\u5353\u8d8a\u7684\u6027\u80fd\uff0c\u5305\u62ec Zero-shot Caption\u3001General VQA\u3001Text-based VQA \u548c Referring Expression Comprehension\u3002\u6b64\u5916\uff0c\u5b83\u5728 TouchStone \u8bc4\u6d4b\u4e2d\uff0c\u4e00\u4e2a\u57fa\u4e8e GPT4 \u6253\u5206\u7cfb\u7edf\u7684\u56fe\u6587\u5bf9\u8bdd\u80fd\u529b\u8bc4\u6d4b\uff0c\u4e5f\u53d6\u5f97\u4e86\u4f18\u5f02\u7684\u6210\u7ee9\uff0c\u8bc1\u660e\u4e86\u5176\u4e0e\u4eba\u7c7b\u5bf9\u9f50\u7684\u9ad8\u6c34\u5e73\u3002<\/p>\n<h3 id=\"-\">\u8bc4\u6d4b\u4eae\u70b9<\/h3>\n<ul>\n<li><strong>Zero-shot Captioning<\/strong>\uff1a\u5728 Flickr30K \u6570\u636e\u96c6\u4e0a\u53d6\u5f97 SOTA \u7ed3\u679c\u3002<\/li>\n<li><strong>General VQA<\/strong>\uff1a\u5728\u540c\u7b49\u89c4\u6a21 LVLM \u6a21\u578b\u4e2d\u53d6\u5f97 SOTA \u7ed3\u679c\u3002<\/li>\n<li><strong>TouchStone<\/strong>\uff1a\u5728\u4e2d\u82f1\u6587\u8bc4\u6d4b\u4e2d\u5747\u83b7\u5f97\u6700\u4f73\u5bf9\u9f50\u8bc4\u5206\u3002<\/li>\n<\/ul>\n<h2 id=\"-\">\u90e8\u7f72\u4f7f\u7528<\/h2>\n<p>Qwen-VL-Chat \u63d0\u4f9b\u4e86\u7075\u6d3b\u7684\u90e8\u7f72\u9009\u9879\uff0c\u652f\u6301\u901a\u8fc7 Web\u3001API\u3001WeChat\u3001Discord \u7b49\u591a\u79cd\u65b9\u5f0f\u8fdb\u884c\u4ea4\u4e92\u3002\u7528\u6237\u53ef\u4ee5\u901a\u8fc7\u63d0\u4f9b\u7684 Python \u4ee3\u7801\u5feb\u901f\u5f00\u59cb\u4f7f\u7528\u6a21\u578b\uff0c\u8fdb\u884c\u56fe\u50cf\u548c\u6587\u672c\u7684\u8f93\u5165\u5904\u7406\u3002<\/p>\n<h3 id=\"-\">\u5feb\u901f\u90e8\u7f72<\/h3>\n<ul>\n<li><strong>\u73af\u5883\u4f9d\u8d56<\/strong>\uff1aPython 3.8+\uff0cPyTorch 1.12+\uff0c\u63a8\u8350 CUDA 11.4+\u3002<\/li>\n<li><strong>\u5b89\u88c5\u547d\u4ee4<\/strong>\uff1a\n<pre><code class=\"lang-bash\">pip <span class=\"token function\">install<\/span> modelscope -U\r\npip <span class=\"token function\">install<\/span> transformers accelerate tiktoken -U\r\npip <span class=\"token function\">install<\/span> einops transformers_stream_generator -U\r\npip <span class=\"token function\">install<\/span> <span class=\"token string\">\"pillow==9.*\"<\/span> -U\r\npip <span class=\"token function\">install<\/span> torchvision\r\npip <span class=\"token function\">install<\/span> matplotlib -U\r\n<\/code><\/pre>\n<\/li>\n<li><strong>\u4ee3\u7801\u793a\u4f8b<\/strong>\uff1a\n<pre><code class=\"lang-python\"><span class=\"token keyword\">from<\/span> modelscope <span class=\"token keyword\">import<\/span> <span class=\"token punctuation\">(<\/span>\r\n    snapshot_download<span class=\"token punctuation\">,<\/span> AutoModelForCausalLM<span class=\"token punctuation\">,<\/span> AutoTokenizer<span class=\"token punctuation\">,<\/span> GenerationConfig\r\n<span class=\"token punctuation\">)<\/span>\r\n<span class=\"token keyword\">import<\/span> torch\r\n\r\nmodel_id <span class=\"token operator\">=<\/span> <span class=\"token string\">'qwen\/Qwen-VL-Chat'<\/span>\r\nrevision <span class=\"token operator\">=<\/span> <span class=\"token string\">'v1.1.0'<\/span>\r\n\r\nmodel_dir <span class=\"token operator\">=<\/span> snapshot_download<span class=\"token punctuation\">(<\/span>model_id<span class=\"token punctuation\">,<\/span> revision<span class=\"token operator\">=<\/span>revision<span class=\"token punctuation\">)<\/span>\r\ntorch<span class=\"token punctuation\">.<\/span>manual_seed<span class=\"token punctuation\">(<\/span><span class=\"token number\">1234<\/span><span class=\"token punctuation\">)<\/span>\r\n\r\ntokenizer <span class=\"token operator\">=<\/span> AutoTokenizer<span class=\"token punctuation\">.<\/span>from_pretrained<span class=\"token punctuation\">(<\/span>model_dir<span class=\"token punctuation\">,<\/span> trust_remote_code<span class=\"token operator\">=<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">)<\/span>\r\nmodel <span class=\"token operator\">=<\/span> AutoModelForCausalLM<span class=\"token punctuation\">.<\/span>from_pretrained<span class=\"token punctuation\">(<\/span>model_dir<span class=\"token punctuation\">,<\/span> device_map<span class=\"token operator\">=<\/span><span class=\"token string\">\"auto\"<\/span><span class=\"token punctuation\">,<\/span> trust_remote_code<span class=\"token operator\">=<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">,<\/span> fp16<span class=\"token operator\">=<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>eval<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span>\r\n\r\nquery <span class=\"token operator\">=<\/span> tokenizer<span class=\"token punctuation\">.<\/span>from_list_format<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">[<\/span>\r\n    <span class=\"token punctuation\">{<\/span><span class=\"token string\">'image'<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">'https:\/\/qianwen-res.oss-cn-beijing.aliyuncs.com\/Qwen-VL\/assets\/demo.jpeg'<\/span><span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">,<\/span>\r\n    <span class=\"token punctuation\">{<\/span><span class=\"token string\">'text'<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">'\u8fd9\u662f\u4ec0\u4e48'<\/span><span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">,<\/span>\r\n<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span>\r\nresponse<span class=\"token punctuation\">,<\/span> history <span class=\"token operator\">=<\/span> model<span class=\"token punctuation\">.<\/span>chat<span class=\"token punctuation\">(<\/span>tokenizer<span class=\"token punctuation\">,<\/span> query<span class=\"token operator\">=<\/span>query<span class=\"token punctuation\">,<\/span> history<span class=\"token operator\">=<\/span>None<span class=\"token punctuation\">)<\/span>\r\n<span class=\"token keyword\">print<\/span><span class=\"token punctuation\">(<\/span>response<span class=\"token punctuation\">)<\/span>\r\n<\/code><\/pre>\n<\/li>\n<\/ul>\n<h2 id=\"-\">\u5e38\u89c1\u95ee\u9898<\/h2>\n<p>Qwen-VL-Chat \u7684\u7528\u6237\u53ef\u80fd\u4f1a\u9047\u5230\u7684\u95ee\u9898\u5305\u62ec\u5b89\u88c5\u914d\u7f6e\u3001\u6a21\u578b\u8c03\u7528\u3001\u6027\u80fd\u4f18\u5316\u7b49\u3002\u5b98\u65b9\u63d0\u4f9b\u4e86 FAQ \u4ee5\u53ca issue \u533a\u57df\uff0c\u7528\u6237\u53ef\u4ee5\u5728\u8fd9\u4e9b\u8d44\u6e90\u4e2d\u5bfb\u627e\u89e3\u51b3\u65b9\u6848\u6216\u63d0\u4ea4\u65b0\u95ee\u9898\u3002<\/p>\n<ul>\n<li style=\"list-style-type: none;\">\n<ul>\n<li><strong>Q:<\/strong>\u00a0\u5982\u4f55\u5b89\u88c5\u548c\u914d\u7f6e Qwen-VL-Chat\uff1f\n<ul>\n<li><strong>A:<\/strong>\u00a0\u8bf7\u6309\u7167\u4e0a\u8ff0\u63d0\u4f9b\u7684\u5b89\u88c5\u547d\u4ee4\u8fdb\u884c\u5b89\u88c5\uff0c\u5e76\u786e\u4fdd\u4f60\u7684\u7cfb\u7edf\u6ee1\u8db3\u6240\u6709\u73af\u5883\u4f9d\u8d56\u3002<\/li>\n<\/ul>\n<\/li>\n<li><strong>Q:<\/strong>\u00a0\u5982\u4f55\u5904\u7406\u6a21\u578b\u8c03\u7528\u4e2d\u7684\u9519\u8bef\uff1f\n<ul>\n<li><strong>A:<\/strong>\u00a0\u68c0\u67e5\u4f60\u7684\u8f93\u5165\u683c\u5f0f\u662f\u5426\u6b63\u786e\uff0c\u5e76\u53c2\u8003\u6587\u6863\u4e2d\u7684\u4ee3\u7801\u793a\u4f8b\u3002\u5982\u679c\u95ee\u9898\u4f9d\u7136\u5b58\u5728\uff0c\u53ef\u4ee5\u5728 issue \u533a\u57df\u63d0\u4ea4\u8be6\u7ec6\u7684\u95ee\u9898\u63cf\u8ff0\u3002<\/li>\n<\/ul>\n<\/li>\n<li><strong>Q:<\/strong>\u00a0\u5982\u4f55\u4f18\u5316\u6a21\u578b\u7684\u6027\u80fd\uff1f\n<ul>\n<li><strong>A:<\/strong>\u00a0\u53ef\u4ee5\u5c1d\u8bd5\u4f7f\u7528\u4e0d\u540c\u7684\u786c\u4ef6\u914d\u7f6e\u6216\u8c03\u6574\u6a21\u578b\u7684\u7cbe\u5ea6\u8bbe\u7f6e\uff08\u5982 bf16 \u6216 fp16\uff09\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<h3 id=\"issue-\">Issue \u533a\u57df<\/h3>\n<ul>\n<li>\u7528\u6237\u53ef\u4ee5\u901a\u8fc7\u00a0<a target=\"_blank\" rel=\"noopener noreferrer nofollow\" href=\"https:\/\/modelscope.cn\/models\/qwen\/Qwen-VL-Chat\/issues\">ModelScope Qwen-VL-Chat Issue<\/a>\u00a0\u63d0\u4ea4\u95ee\u9898\u548c\u53cd\u9988\u3002<\/li>\n<\/ul>\n<h2 id=\"-\">\u76f8\u5173\u8d44\u6e90<\/h2>\n<p>Qwen-VL-Chat \u7684\u7528\u6237\u53ef\u4ee5\u8bbf\u95ee\u4ee5\u4e0b\u8d44\u6e90\u83b7\u53d6\u66f4\u591a\u4fe1\u606f\u548c\u652f\u6301\uff1a<\/p>\n<ul>\n<li><strong>\u6280\u672f\u5907\u5fd8\u5f55<\/strong>\uff1a<a target=\"_blank\" rel=\"noopener noreferrer nofollow\" href=\"https:\/\/modelscope.cn\/models\/qwen\/Qwen-VL-Chat\">Qwen-VL-Chat Technical Memo<\/a><\/li>\n<li><strong>GitHub \u4ed3\u5e93<\/strong>\uff1a<a target=\"_blank\" rel=\"noopener noreferrer nofollow\" href=\"https:\/\/github.com\/modelscope\/swift\/tree\/main\/examples\/pytorch\/llm\">Qwen-VL-Chat GitHub<\/a><\/li>\n<li><strong>ModelScope \u9875\u9762<\/strong>\uff1a<a target=\"_blank\" rel=\"noopener noreferrer nofollow\" href=\"https:\/\/modelscope.cn\/models\/qwen\/Qwen-VL-Chat\">Qwen-VL-Chat ModelScope<\/a><\/li>\n<li><strong>\u91cf\u5316\u6307\u5357<\/strong>\uff1a<a target=\"_blank\" rel=\"noopener noreferrer nofollow\" href=\"https:\/\/modelscope.cn\/models\/qwen\/Qwen-VL-Chat\">Qwen-VL-Chat Quantization Guide<\/a><\/li>\n<\/ul>\n<p>Qwen-VL-Chat \u4f5c\u4e3a\u963f\u91cc\u4e91\u63a8\u51fa\u7684\u5148\u8fdb\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff0c\u4e0d\u4ec5\u5728\u6280\u672f\u4e0a\u8fbe\u5230\u4e86\u9ad8\u6807\u51c6\uff0c\u540c\u65f6\u4e5f\u4e3a\u7528\u6237\u63d0\u4f9b\u4e86\u4e30\u5bcc\u7684\u8d44\u6e90\u548c\u652f\u6301\uff0c\u4ee5\u4fc3\u8fdb\u5176\u5728\u591a\u79cd\u5e94\u7528\u573a\u666f\u4e2d\u7684\u4f7f\u7528\u548c\u521b\u65b0\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Qwen-VL-Chat \u662f\u963f\u91cc\u4e91\u63a8\u51fa\u7684\u4e00\u6b3e\u5148\u8fdb\u7684\u5927\u89c4\u6a21\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff08LVLM\uff09\uff0c\u65e8\u5728\u901a\u8fc7\u7ed3\u5408\u56fe\u50cf\u548c\u6587\u672c\u7684\u8f93\u5165\uff0c\u63d0\u4f9b\u66f4\u4e30\u5bcc\u7684\u591a\u6a21\u6001\u4ea4\u4e92\u4f53\u9a8c\u3002\u8be5\u6a21\u578b\u4e0d\u4ec5\u5728\u6280\u672f\u4e0a\u5177\u6709\u521b\u65b0\u6027\uff0c\u8fd8\u5728\u591a\u8bed\u8a00\u652f\u6301\u548c\u7ec6\u7c92\u5ea6\u8bc6\u522b\u65b9\u9762\u8868\u73b0\u51fa\u8272\u3002 \u5173\u952e\u7279\u6027 \u591a\u6a21\u6001\u8f93\u5165\u8f93\u51fa\uff1aQwen-VL-Chat \u80fd\u591f\u5904\u7406\u56fe\u50cf\u3001\u6587\u672c\u548c\u68c0\u6d4b\u6846\u4f5c\u4e3a\u8f93\u5165\uff0c\u5e76\u751f\u6210\u6587\u672c\u548c\u68c0\u6d4b\u6846\u4f5c\u4e3a\u8f93\u51fa\u3002 \u591a\u8bed\u8a00\u5bf9\u8bdd\uff1a\u652f\u6301\u4e2d\u82f1\u53cc\u8bed\u7684\u957f\u6587\u672c\u8bc6\u522b\uff0c\u589e\u5f3a\u4e86\u8de8\u8bed\u8a00\u7684\u4ea4\u4e92\u80fd\u529b\u3002 \u591a\u56fe\u4ea4\u4e92\uff1a\u652f\u6301\u591a\u56fe\u8f93\u5165\u548c\u6bd4\u8f83\uff0c\u80fd\u591f\u8fdb\u884c\u56fe\u7247\u95f4\u7684\u95ee\u7b54\u548c\u6587\u5b66\u521b\u4f5c\u3002 \u4e2d\u6587\u5b9a\u4f4d\u652f\u6301\uff1a\u901a\u8fc7\u4e2d\u6587\u5f00\u653e\u57df\u8bed\u8a00\u8868\u8fbe\u8fdb\u884c\u68c0\u6d4b\u6846\u6807\u6ce8\uff0c\u63d0\u5347\u4e86\u6a21\u578b<\/p>\n","protected":false},"author":1,"featured_media":59,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[12],"tags":[],"collection":[48],"company":[6],"rank":[],"class_list":["post-58","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-multimodal","collection-qwen-vl","company-alibaba"],"_links":{"self":[{"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/posts\/58","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/comments?post=58"}],"version-history":[{"count":1,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/posts\/58\/revisions"}],"predecessor-version":[{"id":60,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/posts\/58\/revisions\/60"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/media\/59"}],"wp:attachment":[{"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/media?parent=58"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/categories?post=58"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/tags?post=58"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/collection?post=58"},{"taxonomy":"company","embeddable":true,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/company?post=58"},{"taxonomy":"rank","embeddable":true,"href":"https:\/\/www.onetts.com\/ai\/wp-json\/wp\/v2\/rank?post=58"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}