![\"R1-Omni\uff1a\u901a\u8fc7\u89c6\u9891\u548c\u97f3\u9891\u5206\u6790\u60c5\u611f\u7684\u5f00\u6e90\u6a21\u578b-1\"](\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/03\/edef2093082ba55.png\" "\\"R1-Omni\uff1a\u901a\u8fc7\u89c6\u9891\u548c\u97f3\u9891\u5206\u6790\u60c5\u611f\u7684\u5f00\u6e90\u6a21\u578b-1\\"")
<\/p>\nR1-Omni \u662f HumanMLLM \u56e2\u961f\u5728 GitHub \u4e0a\u63a8\u51fa\u7684\u5f00\u6e90\u9879\u76ee\u3002\u5b83\u9996\u6b21\u5c06\u5f3a\u5316\u5b66\u4e60\u4e0e\u53ef\u9a8c\u8bc1\u5956\u52b1\uff08RLVR\uff09\u6280\u672f\u5e94\u7528\u4e8e\u591a\u6a21\u6001\u5927\u8bed\u8a00\u6a21\u578b\uff0c\u4e13\u6ce8\u4e8e\u60c5\u611f\u8bc6\u522b\u3002\u9879\u76ee\u901a\u8fc7\u5206\u6790\u89c6\u9891\u548c\u97f3\u9891\u6570\u636e\uff0c\u8bc6\u522b\u4eba\u7269\u7684\u60c5\u7eea\uff0c\u5982\u6124\u6012\u3001\u5feb\u4e50\u6216\u60ca\u8bb6\u7b49\u3002\u5b83\u4ee5 HumanOmni-0.5B \u4e3a\u57fa\u7840\u6a21\u578b\u5f00\u53d1\uff0c\u63d0\u4f9b\u514d\u8d39\u7684\u4ee3\u7801\u548c\u6a21\u578b\u4e0b\u8f7d\uff0c\u9002\u5408\u7814\u7a76\u4eba\u5458\u548c\u5f00\u53d1\u8005\u4f7f\u7528\u3002R1-Omni \u5728 DFEW\u3001MAFW \u7b49\u6570\u636e\u96c6\u4e0a\u8868\u73b0\u4f18\u5f02\uff0c\u5c24\u5176\u5728\u590d\u6742\u573a\u666f\u4e2d\u80fd\u66f4\u51c6\u786e\u5730\u7406\u89e3\u60c5\u7eea\u3002\u76ee\u524d\u9879\u76ee\u4ecd\u5728\u6301\u7eed\u66f4\u65b0\uff0c\u8ba1\u5212\u5f00\u653e\u66f4\u591a\u8bad\u7ec3\u6570\u636e\u548c\u529f\u80fd\u3002<\/p>\n
<\/p>\n
<\/p>\n
<\/p>\n
R1-Omni \u7684\u4f7f\u7528\u9700\u8981\u4e00\u5b9a\u7684\u6280\u672f\u57fa\u7840\u3002\u4ee5\u4e0b\u662f\u8be6\u7ec6\u7684\u5b89\u88c5\u6b65\u9aa4\uff1a<\/p>\n
\u5728\u7ec8\u7aef\u8f93\u5165\u4ee5\u4e0b\u547d\u4ee4\uff0c\u4e0b\u8f7d R1-Omni \u7684\u4ee3\u7801\uff1a<\/p>\n
git clone https:\/\/github.com\/HumanMLLM\/R1-Omni.git\r\ncd R1-Omni\r\n<\/code><\/pre>\n3. \u914d\u7f6e\u73af\u5883<\/h4>\n
R1-Omni \u57fa\u4e8e R1-V \u6846\u67b6\u5f00\u53d1\uff0c\u73af\u5883\u642d\u5efa\u53c2\u8003\u00a0R1-V \u4ed3\u5e93\u00a0\u7684\u8bf4\u660e\u3002\u5177\u4f53\u6b65\u9aa4\u5982\u4e0b\uff1a<\/p>\n
\n- \u521b\u5efa\u865a\u62df\u73af\u5883\uff1a<\/li>\n<\/ul>\n
python -m venv venv\r\nsource venv\/bin\/activate # Linux\/macOS\r\nvenv\\Scripts\\activate # Windows\r\n<\/code><\/pre>\n\n- \u5b89\u88c5\u4f9d\u8d56\u5305\uff1a\u5b98\u65b9\u63a8\u8350\u4ee5\u4e0b\u7248\u672c\uff0c\u8f93\u5165\u547d\u4ee4\u5b89\u88c5\uff1a<\/li>\n<\/ul>\n
pip install torch==2.5.1+cu124 torchvision==0.20.1+cu124 torchaudio==2.5.1+cu124 -f https:\/\/download.pytorch.org\/whl\/torch_stable.html\r\npip install transformers==4.49.0 flash-attn==2.7.4\r\n<\/code><\/pre>\n\u6ce8\u610f\uff1a\u9700\u786e\u4fdd NVIDIA \u9a71\u52a8\u7248\u672c\u4e0d\u4f4e\u4e8e 535.54\u3002\u82e5\u7f3a\u5c11\u5b8c\u6574\u4f9d\u8d56\u5217\u8868\uff0c\u53ef\u53c2\u8003 R1-V \u4ed3\u5e93\u3002<\/p>\n
4. \u4e0b\u8f7d\u6a21\u578b<\/h4>\n
R1-Omni \u63d0\u4f9b\u591a\u4e2a\u6a21\u578b\u7248\u672c\uff0c\u53ef\u4ece Hugging Face \u6216 ModelScope \u4e0b\u8f7d\u3002\u4ee5 Hugging Face \u4e3a\u4f8b\uff1a<\/p>\n
\n- \u8bbf\u95ee\u00a0Hugging Face R1-Omni<\/a>\u3002<\/li>\n
- \u4e0b\u8f7d\u00a0
R1-Omni-0.5B<\/code>\u00a0\u6a21\u578b\uff0c\u4fdd\u5b58\u5230\u672c\u5730\u8def\u5f84\uff0c\u5982\u00a0\/path\/to\/models\/R1-Omni-0.5B<\/code>\u3002<\/li>\n- \u4e0b\u8f7d\u4f9d\u8d56\u6a21\u578b\uff1a<\/li>\n
- \u97f3\u9891\u6a21\u578b\uff1awhisper-large-v3<\/a>\uff0c\u4fdd\u5b58\u5230\u00a0
\/path\/to\/models\/whisper-large-v3<\/code>\u3002<\/li>\n- \u89c6\u89c9\u6a21\u578b\uff1asiglip-base-patch16-224<\/a>\uff0c\u4fdd\u5b58\u5230\u00a0
\/path\/to\/models\/siglip-base-patch16-224<\/code>\u3002<\/li>\n<\/ul>\n5. \u4fee\u6539\u914d\u7f6e\u6587\u4ef6<\/h4>\n
\u5728\u00a0R1-Omni-0.5B<\/code>\u00a0\u6587\u4ef6\u5939\u4e2d\u627e\u5230\u00a0config.json<\/code>\uff0c\u7528\u6587\u672c\u7f16\u8f91\u5668\u6253\u5f00\uff0c\u4fee\u6539\u4ee5\u4e0b\u4e24\u884c\uff1a<\/p>\n\"mm_audio_tower\": \"\/path\/to\/models\/whisper-large-v3\",\r\n\"mm_vision_tower\": \"\/path\/to\/models\/siglip-base-patch16-224\"\r\n<\/code><\/pre>\n\u4fdd\u5b58\u5e76\u5173\u95ed\u6587\u4ef6\u3002<\/p>\n
\u529f\u80fd\u64cd\u4f5c\u6d41\u7a0b<\/h3>\n1. \u60c5\u611f\u8bc6\u522b\u63a8\u7406<\/h4>\n
R1-Omni \u63d0\u4f9b\u00a0inference.py<\/code>\u00a0\u811a\u672c\uff0c\u7528\u4e8e\u5206\u6790\u5355\u4e2a\u89c6\u9891\u7684\u60c5\u611f\u3002\u64cd\u4f5c\u6b65\u9aa4\u5982\u4e0b\uff1a<\/p>\n\n- \u51c6\u5907\u89c6\u9891\u6587\u4ef6\uff08\u5982\u00a0
video.mp4<\/code>\uff09\uff0c\u786e\u4fdd\u6709\u753b\u9762\u548c\u58f0\u97f3\u3002<\/li>\n- \u5728\u7ec8\u7aef\u8fd0\u884c\uff1a\n
python inference.py --modal video_audio \\\r\n--model_path \/path\/to\/models\/R1-Omni-0.5B \\\r\n--video_path video.mp4 \\\r\n--instruct \"As an emotional recognition expert; throughout the video, which emotion conveyed by the characters is the most obvious to you? Output the thinking process in <think> <\/think> and final emotion in <answer> <\/answer> tags.\"\r\n<\/code><\/pre>\n<\/li>\n- \u8f93\u51fa\u793a\u4f8b\uff1a\n
<think>\u89c6\u9891\u4e2d\u4e00\u540d\u7537\u5b50\u7ad9\u5728\u5f69\u8272\u58c1\u753b\u524d\uff0c\u7a7f\u68d5\u8272\u5939\u514b\uff0c\u7709\u5934\u7d27\u76b1\uff0c\u5634\u5df4\u5f20\u5f00\uff0c\u8868\u60c5\u663e\u5f97\u6fc0\u52a8\u3002\u97f3\u9891\u4e2d\u6709\u201c\u964d\u4f4e\u58f0\u97f3\u201d\u548c\u201c\u522b\u614c\u201d\u7684\u8bcd\u8bed\uff0c\u8bed\u901f\u5feb\u4e14\u8bed\u6c14\u7d27\u5f20\u3002\u7efc\u5408\u5206\u6790\uff0c\u4ed6\u7684\u60c5\u7eea\u662f\u6124\u6012\u548c\u4e0d\u5b89\u3002<\/think>\r\n<answer>angry<\/answer>\r\n<\/code><\/pre>\n<\/li>\n<\/ul>\n2. \u6d4b\u8bd5\u6a21\u578b\u6027\u80fd<\/h4>\n
\u5b98\u65b9\u5728 DFEW\u3001MAFW \u548c RAVDESS \u6570\u636e\u96c6\u4e0a\u6d4b\u8bd5\u4e86\u6a21\u578b\u6027\u80fd\u3002\u7528\u6237\u53ef\u4e0b\u8f7d\u8fd9\u4e9b\u6570\u636e\u96c6\uff08\u89c1\u00a0DFEW \u5b98\u7f51\u00a0\u6216\u00a0MAFW \u5b98\u7f51\uff09\uff0c\u7136\u540e\u7528\u4e0a\u8ff0\u547d\u4ee4\u9a8c\u8bc1\u672c\u5730\u6548\u679c\u3002\u5bf9\u6bd4\u6570\u636e\u5982\u4e0b\uff1a<\/p>\n
\n- R1-Omni \u5728 DFEW (WAR) \u4e0a\u8fbe 65.83%\uff0c\u4f18\u4e8e HumanOmni-0.5B \u7684 22.64%\u3002<\/li>\n<\/ul>\n
3. \u81ea\u5b9a\u4e49\u8bad\u7ec3<\/h4>\n\n- \u51b7\u542f\u52a8\u8bad\u7ec3<\/strong>\uff1a\u57fa\u4e8e Explainable Multimodal Emotion Reasoning\uff08232 \u4e2a\u6837\u672c\uff09\u548c HumanOmni\uff08348 \u4e2a\u6837\u672c\uff09\u6570\u636e\u96c6\u521d\u59cb\u5316\u6a21\u578b\u3002\u6570\u636e\u683c\u5f0f\u793a\u4f8b\uff1a\n
[{\"video\": \"MER24\/sample_00000967.mp4\", \"conversations\": [{\"from\": \"human\", \"value\": \"<video>\\n<audio>\\n\u8bf7\u8bc6\u522b\u89c6\u9891\u4e2d\u7684\u4e3b\u8981\u60c5\u7eea\"}, {\"from\": \"gpt\", \"value\": \"<think>\u89c6\u9891\u4e2d\u4e00\u540d\u7537\u5b50\u5728\u6253\u7535\u8bdd\uff0c\u7709\u5934\u7d27\u76b1\uff0c\u8bed\u901f\u5feb\uff0c\u8bed\u6c14\u7d27\u5f20\uff0c\u8868\u73b0\u51fa\u7126\u8651\u3002<\/think>\\n<answer>anxious<\/answer>\"}]}]\r\n<\/code><\/pre>\n\u6570\u636e\u5c1a\u672a\u5b8c\u5168\u5f00\u6e90\uff0c\u53ef\u5173\u6ce8 GitHub \u52a8\u6001\u3002<\/li>\n
- RLVR \u8bad\u7ec3<\/strong>\uff1a\u4f7f\u7528 MAFW \u548c DFEW \u6570\u636e\u96c6\uff08\u5171 15,306 \u4e2a\u89c6\u9891\uff09\u3002\u6570\u636e\u683c\u5f0f\u793a\u4f8b\uff1a\n
[{\"video\": \"DFEW\/videos\/1.mp4\", \"conversations\": [{\"from\": \"human\", \"value\": \"<video>\\n<audio>\\n\u8bf7\u8bc6\u522b\u89c6\u9891\u4e2d\u7684\u4e3b\u8981\u60c5\u7eea\"}, {\"from\": \"gpt\", \"value\": \"sad\"}]}]\r\n<\/code><\/pre>\n\u8bad\u7ec3\u7ec6\u8282\u5f85\u5b98\u65b9\u8fdb\u4e00\u6b65\u8bf4\u660e\u3002<\/li>\n<\/ul>\n
\u6ce8\u610f\u4e8b\u9879<\/h3>\n\n- \u89c6\u9891\u8981\u6c42<\/strong>\uff1a\u652f\u6301 MP4 \u7b49\u683c\u5f0f\uff0c\u9700\u5305\u542b\u6e05\u6670\u753b\u9762\u548c\u97f3\u9891\u3002<\/li>\n
- \u6a21\u578b\u9009\u62e9<\/strong>\uff1aR1-Omni \u662f\u6700\u7ec8\u7248\u672c\uff0c\u6027\u80fd\u6700\u4f18\uff1b\u5176\u4ed6\u6a21\u578b\u53ef\u7528\u4e8e\u5bf9\u6bd4\u5b9e\u9a8c\u3002<\/li>\n
- \u6280\u672f\u652f\u6301<\/strong>\uff1a\u9047\u5230\u95ee\u9898\u53ef\u5728 GitHub \u63d0\u4ea4 Issue\uff0c\u56e2\u961f\u4f1a\u6301\u7eed\u5b8c\u5584\u6587\u6863\u3002<\/li>\n<\/ul>\n
\u901a\u8fc7\u4ee5\u4e0a\u6b65\u9aa4\uff0c\u7528\u6237\u53ef\u4ee5\u5feb\u901f\u5b89\u88c5\u5e76\u4f7f\u7528 R1-Omni\uff0c\u4f53\u9a8c\u5176\u60c5\u611f\u8bc6\u522b\u529f\u80fd\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"
R1-Omni \u662f HumanMLLM \u56e2\u961f\u5728 GitHub \u4e0a\u63a8\u51fa\u7684\u5f00\u6e90\u9879\u76ee\u3002\u5b83\u9996\u6b21\u5c06\u5f3a\u5316\u5b66\u4e60\u4e0e\u53ef\u9a8c\u8bc1\u5956\u52b1\uff08RLVR\uff09\u6280\u672f\u5e94\u7528\u4e8e\u591a\u6a21\u6001\u5927\u8bed\u8a00\u6a21\u578b\uff0c\u4e13\u6ce8\u4e8e\u60c5\u611f\u8bc6\u522b\u3002\u9879\u76ee\u901a\u8fc7\u5206\u6790\u89c6\u9891\u548c\u97f3\u9891\u6570\u636e\uff0c\u8bc6\u522b\u4eba\u7269\u7684\u60c5\u7eea\uff0c\u5982\u6124\u6012\u3001\u5feb\u4e50\u6216\u60ca\u8bb6\u7b49\u3002\u5b83\u4ee5 Hu…<\/p>\n","protected":false},"author":1,"featured_media":62026,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[230],"class_list":["post-28373","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-tool","tag-aikaiyuanxiangmu"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts\/28373","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/comments?post=28373"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts\/28373\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/media\/62026"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/media?parent=28373"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/categories?post=28373"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/tags?post=28373"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}