[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-10652":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":16,"forks30d":16,"starsTrendScore":16,"compositeScore":19,"rankGlobal":10,"rankLanguage":10,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":21,"topics":24,"createdAt":10,"pushedAt":10,"updatedAt":31,"readmeContent":32,"aiSummary":33,"trendingCount":16,"starSnapshotCount":16,"syncStatus":17,"lastSyncTime":34,"discoverSource":35},10652,"ChatRWKV","BlinkDL\u002FChatRWKV","BlinkDL","ChatRWKV is like ChatGPT but powered by RWKV (100% RNN) language model, and open source.","",null,"Python",9492,686,91,35,0,2,6,39.51,"Apache License 2.0",false,"main",true,[25,26,27,28,29,30],"chatbot","chatgpt","language-model","pytorch","rnn","rwkv","2026-06-12 02:02:24","# ChatRWKV (pronounced as \"RwaKuv\" (rʌkuv in IPA), from 4 major params: R W K V)\n\n## RWKV homepage: https:\u002F\u002Fwww.rwkv.com\n\nRWKV-7 code: https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FRWKV-LM\u002Ftree\u002Fmain\u002FRWKV-v7\n\nChatRWKV is like ChatGPT but powered by my RWKV (100% RNN) language model, which is the only RNN (as of now) that can match transformers in quality and scaling, while being faster and saves VRAM.\n\nOur latest version is **RWKV-7** https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.14456 (Preview models: https:\u002F\u002Fhuggingface.co\u002FBlinkDL\u002Ftemp )\n\nGradio Demo 1: https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FBlinkDL\u002FRWKV-Gradio-1\n\nGradio Demo 2: https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FBlinkDL\u002FRWKV-Gradio-2\n\n**RWKV-LM main repo**: https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FRWKV-LM (explanation, fine-tuning, training, etc.)\n\nChat Demo for developers: https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Fblob\u002Fmain\u002FAPI_DEMO_CHAT.py\n\n**Efficient inference project**: https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FAlbatross\n\n**RWKV APP**: https:\u002F\u002Fgithub.com\u002FRWKV-APP\u002FRWKV_APP (local inference for Android \u002F iOS)\n\n## RWKV Discord: https:\u002F\u002Fdiscord.gg\u002FbDSBUMeFpc (7k+ members)\n\n**Twitter**: https:\u002F\u002Ftwitter.com\u002FBlinkDL_AI\n\n**Homepage**: https:\u002F\u002Fwww.rwkv.com\u002F\n\n**Raw cutting-edge RWKV weights:** https:\u002F\u002Fhuggingface.co\u002FBlinkDL\n\n**GGUF:** https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Fshoumenchougou\u002Frwkv7-gxx-gguf\n\n**HF-compatible RWKV weights:** https:\u002F\u002Fhuggingface.co\u002FRWKV\n\nUse v2\u002Fconvert_model.py to convert a model for a strategy, for faster loading & saves CPU RAM.\n\nNote RWKV_CUDA_ON will build a CUDA kernel (much faster & saves VRAM). Here is how to build it (\"pip install ninja\" first):\n```\n# How to build in Linux: set these and run v2\u002Fchat.py\nexport PATH=\u002Fusr\u002Flocal\u002Fcuda\u002Fbin:$PATH\nexport LD_LIBRARY_PATH=\u002Fusr\u002Flocal\u002Fcuda\u002Flib64:$LD_LIBRARY_PATH\n# How to build in win:\nInstall VS2022 build tools (https:\u002F\u002Faka.ms\u002Fvs\u002F17\u002Frelease\u002Fvs_BuildTools.exe select Desktop C++). Reinstall CUDA 11.7 (install VC++ extensions). Run v2\u002Fchat.py in \"x64 native tools command prompt\". \n```\n**RWKV pip package**: https:\u002F\u002Fpypi.org\u002Fproject\u002Frwkv\u002F **(please always check for latest version and upgrade)**\n\nhttps:\u002F\u002Fgithub.com\u002Fcgisky1980\u002Fai00_rwkv_server Fastest GPU inference API with vulkan (good for nvidia\u002Famd\u002Fintel)\n\nhttps:\u002F\u002Fgithub.com\u002Fcryscan\u002Fweb-rwkv backend for ai00_rwkv_server\n\nhttps:\u002F\u002Fgithub.com\u002FsaharNooby\u002Frwkv.cpp Fast CPU\u002FcuBLAS\u002FCLBlast inference: int4\u002Fint8\u002Ffp16\u002Ffp32\n\nhttps:\u002F\u002Fgithub.com\u002FJL-er\u002FRWKV-PEFT lora\u002Fpissa\u002FQlora\u002FQpissa\u002Fstate tuning\n\nhttps:\u002F\u002Fgithub.com\u002FRWKV\u002FRWKV-infctx-trainer Infctx trainer\n\n**World demo script:** https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Fblob\u002Fmain\u002FAPI_DEMO_WORLD.py\n\n**Raven Q&A demo script:** https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Fblob\u002Fmain\u002Fv2\u002Fbenchmark_more.py\n\n![ChatRWKV-strategy](ChatRWKV-strategy.png)\n\n**RWKV in 150 lines** (model, inference, text generation): https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Fblob\u002Fmain\u002FRWKV_in_150_lines.py\n\n**🔥 RWKV v5 in 250 lines 🔥** (with tokenizer too): https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Fblob\u002Fmain\u002FRWKV_v5_demo.py\n\n**🔥 Building your own RWKV inference engine 🔥**: begin with https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Fblob\u002Fmain\u002Fsrc\u002Fmodel_run.py which is easier to understand (used by https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Fblob\u002Fmain\u002Fchat.py).\n\n**RWKV preprint** https:\u002F\u002Farxiv.org\u002Fabs\u002F2305.13048\n\n![RWKV-paper](RWKV-paper.png)\n\nRWKV v6 illustrated:\n\n![RWKV-v6](rwkv-x060.png)\n\n**Cool Community RWKV Projects**:\n\nhttps:\u002F\u002Fgithub.com\u002FsaharNooby\u002Frwkv.cpp fast i4 i8 fp16 fp32 CPU inference using [ggml](https:\u002F\u002Fgithub.com\u002Fggerganov\u002Fggml)\n\nhttps:\u002F\u002Fgithub.com\u002Fharrisonvanderbyl\u002Frwkv-cpp-cuda fast windows\u002Flinux & cuda\u002Frocm\u002Fvulkan GPU inference (no need for python & pytorch)\n\nhttps:\u002F\u002Fgithub.com\u002FBlealtan\u002FRWKV-LM-LoRA LoRA fine-tuning\n\nhttps:\u002F\u002Fgithub.com\u002FjosStorer\u002FRWKV-Runner cool GUI\n\nMore RWKV projects: https:\u002F\u002Fgithub.com\u002Fsearch?o=desc&q=rwkv&s=updated&type=Repositories\n\nChatRWKV v2: with \"stream\" and \"split\" strategies, and INT8. 3G VRAM is enough to run RWKV 14B :) https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Ftree\u002Fmain\u002Fv2\n```python\nos.environ[\"RWKV_JIT_ON\"] = '1'\nos.environ[\"RWKV_CUDA_ON\"] = '0' # if '1' then use CUDA kernel for seq mode (much faster)\nfrom rwkv.model import RWKV                         # pip install rwkv\nmodel = RWKV(model='\u002Ffsx\u002FBlinkDL\u002FHF-MODEL\u002Frwkv-4-pile-1b5\u002FRWKV-4-Pile-1B5-20220903-8040', strategy='cuda fp16')\n\nout, state = model.forward([187, 510, 1563, 310, 247], None)   # use 20B_tokenizer.json\nprint(out.detach().cpu().numpy())                   # get logits\nout, state = model.forward([187, 510], None)\nout, state = model.forward([1563], state)           # RNN has state (use deepcopy if you want to clone it)\nout, state = model.forward([310, 247], state)\nprint(out.detach().cpu().numpy())                   # same result as above\n```\n![RWKV-eval](RWKV-eval.png)\n\nHere is https:\u002F\u002Fhuggingface.co\u002FBlinkDL\u002Frwkv-4-raven\u002Fblob\u002Fmain\u002FRWKV-4-Raven-14B-v7-Eng-20230404-ctx4096.pth in action:\n![ChatRWKV](ChatRWKV.png)\n\nWhen you build a RWKV chatbot, always check the text corresponding to the state, in order to prevent bugs.\n\n1. Never call raw forward() directly. Instead, put it in a function that will record the text corresponding to the state.\n\n**(For v4-raven models, use Bob\u002FAlice. For v4\u002Fv5\u002Fv6-world models, use User\u002FAssistant)**\n\n2. The best chat format (check whether your text is of this format):\n```Bob: xxxxxxxxxxxxxxxxxx\\n\\nAlice: xxxxxxxxxxxxx\\n\\nBob: xxxxxxxxxxxxxxxx\\n\\nAlice:```\n\n* There should not be any space after the final \"Alice:\". The generation result will have a space in the beginning, and you can simply strip it.\n* You can use \\n in xxxxx, but avoid \\n\\n. So simply do ```xxxxx = xxxxx.strip().replace('\\r\\n','\\n').replace('\\n\\n','\\n')```\n\nIf you are building your own RWKV inference engine, begin with https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Fblob\u002Fmain\u002Fsrc\u002Fmodel_run.py which is easier to understand (used by https:\u002F\u002Fgithub.com\u002FBlinkDL\u002FChatRWKV\u002Fblob\u002Fmain\u002Fchat.py)\n\nThe lastest \"Raven\"-series Alpaca-style-tuned RWKV 14B & 7B models are very good (almost ChatGPT-like, good at multiround chat too). Download: https:\u002F\u002Fhuggingface.co\u002FBlinkDL\u002Frwkv-4-raven\n\nPrevious old model results:\n![ChatRWKV](misc\u002Fsample-1.png)\n![ChatRWKV](misc\u002Fsample-2.png)\n![ChatRWKV](misc\u002Fsample-3.png)\n![ChatRWKV](misc\u002Fsample-4.png)\n![ChatRWKV](misc\u002Fsample-5.png)\n![ChatRWKV](misc\u002Fsample-6.png)\n![ChatRWKV](misc\u002Fsample-7.png)\n\n## 中文模型\n\nQQ群 553456870（加入时请简单自我介绍）。有研发能力的朋友加群 325154699。\n\n中文使用教程：https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F618011122 https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F616351661\n\n推荐UI：https:\u002F\u002Fgithub.com\u002Fl15y\u002Fwenda\n\n## Star History\n\n[![Star History Chart](https:\u002F\u002Fapi.star-history.com\u002Fsvg?repos=BlinkDL\u002FChatRWKV&type=Date)](https:\u002F\u002Fstar-history.com\u002F#BlinkDL\u002FChatRWKV&Date)\n","ChatRWKV 是一个基于 RWKV（100% RNN）语言模型的聊天机器人项目，类似于 ChatGPT 且开源。其核心功能包括高质量的语言生成和对话能力，技术特点在于使用了完全基于 RNN 的架构，这使得它在保持与 transformer 模型相当的质量和扩展性的同时，运行速度更快且更节省显存。ChatRWKV 支持多种部署方式，包括通过 Gradio 构建的在线演示、本地推理引擎以及适用于 Android\u002FiOS 的应用程序，适合需要高性能文本生成或对话系统的场景，特别是在资源受限的环境中。","2026-06-11 03:29:33","top_topic"]