{"id":9165,"date":"2026-06-16T06:04:12","date_gmt":"2026-06-15T22:04:12","guid":{"rendered":"https:\/\/infernews.com\/blog\/official-implementations-of-quot-rhymeflow-training-free-acceleration-for-video\/"},"modified":"2026-06-16T06:11:37","modified_gmt":"2026-06-15T22:11:37","slug":"official-implementations-of-quot-rhymeflow-training-free-acceleration-for-video","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/official-implementations-of-quot-rhymeflow-training-free-acceleration-for-video\/","title":{"rendered":"RhymeFlow\uff1a\u52a0\u5feb\u5f71\u7247\u751f\u6210\u7684\u65b0\u8def\u7dda"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/pasted-77bdbacc31fd.jpg\" alt=\"Repository image for Simon-Dcs\/RhymeFlow\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u6642\u4e0d\u5c11\u5f71\u7247\u751f\u6210\u52a0\u901f\u65b9\u6cd5\uff0c\u4e3b\u8981\u4ecd\u6cbf\u7528\u6a19\u6e96 diffusion pipeline\uff1a\u6bcf\u4e00\u5e40\u90fd\u8981\u5728\u6240\u6709 diffusion timesteps \u5b8c\u6574\u505a\u4e00\u6b21 dense denoising\uff0c\u518d\u914d\u5408 sparse attention \u6216 KV-caching \u6e1b\u5c11\u55ae\u6b65\u8a08\u7b97\u3002RhymeFlow \u6307\u51fa\uff0c\u9019\u7a2e\u56fa\u5b9a\u7bc4\u5f0f\u5ffd\u7565\u4e86\u76f8\u9130\u5f71\u683c\u5167\u5bb9\u8207\u52d5\u4f5c\u9ad8\u5ea6\u76f8\u95dc\uff0c\u4ee4\u81ea\u7136\u5f71\u7247\u88e1\u5927\u91cf\u4e2d\u9593\u6b65\u9a5f\u5176\u5be6\u5c6c\u65bc\u91cd\u8907\u904b\u7b97\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9019\u662f\u4e00\u500b<strong>\u514d\u8a13\u7df4\u7684\u5f71\u7247\u751f\u6210\u52a0\u901f\u6846\u67b6<\/strong>\uff0c\u6838\u5fc3\u76ee\u6a19\u662f\u66ff DiT\uff08Diffusion Transformers\uff09\u5f71\u7247\u6a21\u578b\u6e1b\u5c11\u63a8\u7406\u5ef6\u9072\u8207\u904b\u7b97\u6210\u672c\u3002\u5b83\u5c07\u4e0d\u540c\u5f71\u683c\u7684 denoising trajectory \u62c6\u958b\u8655\u7406\uff1a\u5148\u627e\u51fa\u4e3b\u5c0e\u8a9e\u610f\u8b8a\u5316\u7684 keyframes\uff0c\u8b93 keyframes \u4fdd\u6301\u9010\u6b65\u53bb\u566a\uff0c\u975e keyframes \u5247\u9010\u6b65\u8df3\u904e\u90e8\u5206\u6b65\u9a5f\uff0c\u518d\u7528 latent trajectory projection \u88dc\u56de\u6642\u9593\u4e00\u81f4\u6027\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9019\u500b\u505a\u6cd5\u7684\u5275\u65b0\uff0c\u4e0d\u5728\u65bc\u55ae\u7d14\u628a attention \u518d\u7a00\u758f\u5316\uff0c\u800c\u662f\u76f4\u63a5\u6311\u6230\u300c\u6240\u6709\u5f71\u683c\u90fd\u8981\u540c\u6b65\u3001\u5bc6\u96c6\u53bb\u566a\u300d\u7684\u820a\u5047\u8a2d\u3002\u8ad6\u6587\u63cf\u8ff0\uff0cRhymeFlow \u5728\u73fe\u6709 DiT-based video generation models \u4e0a\uff0c\u80fd\u540c\u6642\u53d6\u5f97\u66f4\u9ad8 inference speed \u8207\u66f4\u597d visual quality\uff1b\u4e0d\u904e GitHub \u76ee\u524d\u516c\u958b\u91cd\u9ede\u653e\u5728 Wan 2.1 adaptation\uff0cHunyuanVideo adaptation \u4ecd\u5728\u6e96\u5099\u4e2d\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5982\u679c\u4f60\u60f3\u8a66\uff0c\u8f03\u5408\u7406\u7684\u5207\u5165\u9ede\u662f\u628a\u5b83\u7576\u6210 Wan 2.1 \u7684\u52a0\u901f\u5be6\u9a57\u6846\u67b6\uff0c\u6bd4\u8f03 dense\u3001svg\u3001sap\u3001rhyme\u3001rhyme_sap \u5e7e\u7a2e\u65b9\u6cd5\u8f38\u51fa\u6642\u9593\u8207\u756b\u9762\u5dee\u7570\u3002\u74b0\u5883\u8981\u6c42\u504f\u9ad8\uff0c\u6587\u4ef6\u5217\u51fa CUDA 12.4 \/ 12.8 \u8207 PyTorch 2.5.1 \/ 2.6.0\uff0c\u4ea6\u727d\u6d89 FlashInfer\u3001flash-attn \u548c\u81ea\u8a02 kernels\uff0c\u8f03\u9069\u5408\u5df2\u6709 GPU \u8207 PyTorch \u7d93\u9a57\u7684\u4eba\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u9805\u76ee\u985e\u578b<\/strong>\uff1atraining-free \u5f71\u7247\u751f\u6210\u52a0\u901f\u6846\u67b6\uff0c\u8655\u7406 DiT \u5f71\u7247\u6a21\u578b\u63a8\u7406\u592a\u6162\u7684\u554f\u984c<\/li>\n\n\n\n<li><strong>\u65b9\u6cd5\u91cd\u9ede<\/strong>\uff1akeyframes \u505a dense denoising\uff0c\u975e keyframes \u8df3\u6b65\u8655\u7406\uff0c\u518d\u7528 latent trajectory projection \u7dad\u6301\u6642\u5e8f\u4e00\u81f4<\/li>\n\n\n\n<li><strong>\u53ef\u6bd4\u8f03\u65b9\u6cd5<\/strong>\uff1adense\u3001svg\u3001sap\u3001rhyme\u3001rhyme_sap<\/li>\n\n\n\n<li><strong>\u76f8\u95dc\u6a21\u578b<\/strong>\uff1aWan 2.1 \u5df2\u6709 adaptation\uff0cHunyuanVideo adaptation \u5c1a\u672a\u5b8c\u6574\u91cb\u51fa<\/li>\n\n\n\n<li><strong>\u9069\u5408\u5834\u666f<\/strong>\uff1a\u7814\u7a76\u5f71\u7247\u751f\u6210\u63a8\u7406\u512a\u5316\u3001\u6bd4\u8f03\u4e0d\u540c\u52a0\u901f\u7b56\u7565\u3001\u6e2c\u8a66\u901f\u5ea6\u8207\u756b\u8cea\u53d6\u6368<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u6574\u9ad4\u4f86\u770b\uff0cRhymeFlow \u7684\u50f9\u503c\u5f88\u660e\u78ba\uff1a\u5b83\u4e0d\u662f\u6539\u6a21\u578b\u6b0a\u91cd\uff0c\u4e5f\u4e0d\u662f\u91cd\u65b0\u8a13\u7df4\uff0c\u800c\u662f\u91cd\u6392 denoising flow scheduling\uff0c\u5f9e\u6d41\u7a0b\u5c64\u9762\u7bc0\u7701\u8a08\u7b97\u3002\u5c0d\u7814\u7a76\u8005\u8207\u9032\u968e\u958b\u767c\u8005\u800c\u8a00\uff0c\u9019\u985e\u601d\u8def\u6bd4\u55ae\u7d14\u5806\u786c\u4ef6\u66f4\u6709\u53c3\u8003\u50f9\u503c\uff1b\u5c0d\u4e00\u822c\u5275\u4f5c\u8005\u4f86\u8aaa\uff0c\u73fe\u968e\u6bb5\u9580\u6abb\u4ecd\u5728\u90e8\u7f72\u8207 GPU \u74b0\u5883\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/Simon-Dcs\/RhymeFlow\" rel=\"noopener noreferrer\">https:\/\/github.com\/Simon-Dcs\/RhymeFlow<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Paper\uff1a<\/strong> <a href=\"https:\/\/arxiv.org\/pdf\/2606.06309\" rel=\"noopener noreferrer\">https:\/\/arxiv.org\/pdf\/2606.06309<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>RhymeFlow \u91dd\u5c0d\u5f71\u7247\u751f\u6210\u592a\u6162\u7684\u8001\u554f\u984c\uff0c\u63d0\u51fa\u514d\u8a13\u7df4\u52a0\u901f\u65b9\u6cd5\u3002\u5b83\u4e0d\u53ea\u8ffd\u6c42\u66f4\u5feb\uff0c\u4ea6\u5617\u8a66\u4fdd\u4f4f\u756b\u9762\u9023\u8cab\u5ea6\u3002<\/p>\n","protected":false},"author":8,"featured_media":9164,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,172,179,31,32,162,152,127,160,141,196,197],"tags":[],"class_list":["post-9165","post","type-post","status-publish","format-standard","hentry","category-133","category-172","category-nvidia","category-video","category-contentcreator","category-ai-productions","category-python","category-127","category-160","category-141","category-196","category-framework"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9165","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=9165"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9165\/revisions"}],"predecessor-version":[{"id":9167,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9165\/revisions\/9167"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9164"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9165"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9165"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9165"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}