[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skills:pro6":3,"skills:sce6":50,"skills:creators9":95,"skills:stats":193,"skills:all:stars:450:":195},[4,15,22,29,36,43],{"categoryNo":5,"taxonomyType":6,"taxonomyTypeDesc":7,"categoryName":8,"categoryNameEn":9,"description":10,"level":11,"skillCount":12,"parentNo":13,"parentName":13,"children":14},"01JBSEEDCAT000000000000030","PROFESSION","适用职业","开发工程师","Software Engineer","从事软件编码、测试与运维的工程师",1,13341,null,[],{"categoryNo":16,"taxonomyType":6,"taxonomyTypeDesc":7,"categoryName":17,"categoryNameEn":18,"description":19,"level":11,"skillCount":20,"parentNo":13,"parentName":13,"children":21},"01JBSEEDCAT000000000000032","设计师","Designer","视觉、交互与品牌设计人员",7902,[],{"categoryNo":23,"taxonomyType":6,"taxonomyTypeDesc":7,"categoryName":24,"categoryNameEn":25,"description":26,"level":11,"skillCount":27,"parentNo":13,"parentName":13,"children":28},"01JBSEEDCAT000000000000033","数据分析师","Data Analyst","数据分析、报表与商业智能人员",5005,[],{"categoryNo":30,"taxonomyType":6,"taxonomyTypeDesc":7,"categoryName":31,"categoryNameEn":32,"description":33,"level":11,"skillCount":34,"parentNo":13,"parentName":13,"children":35},"01JBSEEDCAT000000000000036","内容创作者","Content Creator","作者、编辑与自媒体创作者",3978,[],{"categoryNo":37,"taxonomyType":6,"taxonomyTypeDesc":7,"categoryName":38,"categoryNameEn":39,"description":40,"level":11,"skillCount":41,"parentNo":13,"parentName":13,"children":42},"01JBSEEDCAT000000000000035","市场营销","Marketer","市场推广、增长与品牌运营人员",3454,[],{"categoryNo":44,"taxonomyType":6,"taxonomyTypeDesc":7,"categoryName":45,"categoryNameEn":46,"description":47,"level":11,"skillCount":48,"parentNo":13,"parentName":13,"children":49},"01JBSEEDCAT000000000000037","律师与法务","Legal","合同、合规与法律事务人员",3337,[],[51,60,67,74,81,88],{"categoryNo":52,"taxonomyType":53,"taxonomyTypeDesc":54,"categoryName":55,"categoryNameEn":56,"description":57,"level":11,"skillCount":58,"parentNo":13,"parentName":13,"children":59},"01JBSEEDCAT000000000000022","SCENARIO","使用场景","智能体与自动化","Agents & Automation","Skill、MCP、智能体与工作流自动化",39675,[],{"categoryNo":61,"taxonomyType":53,"taxonomyTypeDesc":54,"categoryName":62,"categoryNameEn":63,"description":64,"level":11,"skillCount":65,"parentNo":13,"parentName":13,"children":66},"01JBSEEDCAT000000000000006","软件开发","Software Development","编码、测试、部署等软件研发任务",26798,[],{"categoryNo":68,"taxonomyType":53,"taxonomyTypeDesc":54,"categoryName":69,"categoryNameEn":70,"description":71,"level":11,"skillCount":72,"parentNo":13,"parentName":13,"children":73},"01JBSEEDCAT000000000000014","设计创意","Design & Creative","视觉设计、品牌与多媒体创作",13847,[],{"categoryNo":75,"taxonomyType":53,"taxonomyTypeDesc":54,"categoryName":76,"categoryNameEn":77,"description":78,"level":11,"skillCount":79,"parentNo":13,"parentName":13,"children":80},"01JBSEEDCAT000000000000001","办公效率","Office Productivity","日常办公文档、表格、演示与沟通类任务",6403,[],{"categoryNo":82,"taxonomyType":53,"taxonomyTypeDesc":54,"categoryName":83,"categoryNameEn":84,"description":85,"level":11,"skillCount":86,"parentNo":13,"parentName":13,"children":87},"01JBSEEDCAT000000000000017","内容创作","Content Creation","写作、文案与营销内容生产",6063,[],{"categoryNo":89,"taxonomyType":53,"taxonomyTypeDesc":54,"categoryName":90,"categoryNameEn":91,"description":92,"level":11,"skillCount":93,"parentNo":13,"parentName":13,"children":94},"01JBSEEDCAT000000000000020","研究与学习","Research & Learning","资料检索、知识整理与学习辅导",5574,[],{"records":96,"total":191,"current":11,"size":192},[97,111,120,129,139,149,159,169,179],{"accountNo":98,"login":99,"displayName":100,"avatarUrl":101,"accountType":102,"accountTypeDesc":103,"profileDescription":104,"skillCount":105,"repoCount":106,"professionCount":107,"scenarioCount":108,"totalStars":109,"updatedAt":110},"01M698VYDXZ6JW5WH3CRV650R8","microsoft","Microsoft","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F6154722?v=4","ORGANIZATION","组织","Open source projects and samples from Microsoft",4300,236,11,7,832929,"2026-11-15 04:18:35",{"accountNo":112,"login":113,"displayName":113,"avatarUrl":114,"accountType":102,"accountTypeDesc":103,"profileDescription":115,"skillCount":116,"repoCount":117,"professionCount":108,"scenarioCount":108,"totalStars":118,"updatedAt":119},"01M698P3FSAZ6B4R928ZF19KPT","openclaw","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F252820863?v=4","Your personal, open source AI assistant.",369,32,437811,"2026-11-15 03:04:02",{"accountNo":121,"login":122,"displayName":123,"avatarUrl":124,"accountType":102,"accountTypeDesc":103,"profileDescription":13,"skillCount":125,"repoCount":126,"professionCount":107,"scenarioCount":108,"totalStars":127,"updatedAt":128},"01M697XGDPXJAVQ8391V2BCYTE","anthropics","Anthropic","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F76263028?v=4",716,18,427323,"2026-11-15 05:52:49",{"accountNo":130,"login":131,"displayName":132,"avatarUrl":133,"accountType":102,"accountTypeDesc":103,"profileDescription":13,"skillCount":134,"repoCount":135,"professionCount":108,"scenarioCount":136,"totalStars":137,"updatedAt":138},"01M6AC29YGCR4S2NTJFV0G8B81","openai","OpenAI","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F14957082?v=4",177,23,6,344400,"2026-11-15 03:07:28",{"accountNo":140,"login":141,"displayName":142,"avatarUrl":143,"accountType":102,"accountTypeDesc":103,"profileDescription":13,"skillCount":144,"repoCount":145,"professionCount":146,"scenarioCount":108,"totalStars":147,"updatedAt":148},"01M6AC83E6K4FDKK7HZ8WNMQKE","apache","The Apache Software Foundation","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F47359?v=4",440,76,8,298252,"2026-11-15 04:11:16",{"accountNo":150,"login":151,"displayName":152,"avatarUrl":153,"accountType":102,"accountTypeDesc":103,"profileDescription":154,"skillCount":155,"repoCount":156,"professionCount":107,"scenarioCount":108,"totalStars":157,"updatedAt":158},"01M6AC83DS133TPXFSP1DHNCC6","github","GitHub","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F9919?v=4","How people build software.",1022,30,278314,"2026-11-15 06:45:15",{"accountNo":160,"login":161,"displayName":162,"avatarUrl":163,"accountType":102,"accountTypeDesc":103,"profileDescription":164,"skillCount":165,"repoCount":166,"professionCount":107,"scenarioCount":108,"totalStars":167,"updatedAt":168},"01M698VYEFDM0VJBZDYYJ2T83S","google","Google","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1342004?v=4","Google ❤️ Open Source",542,50,243804,"2026-11-15 03:00:30",{"accountNo":170,"login":171,"displayName":172,"avatarUrl":173,"accountType":102,"accountTypeDesc":103,"profileDescription":13,"skillCount":174,"repoCount":175,"professionCount":176,"scenarioCount":108,"totalStars":177,"updatedAt":178},"01M6AC83EPDGBFN19MWYRREJD0","Significant-Gravitas","AutoGPT","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F130738209?v=4",87,2,10,188029,"2026-11-15 06:21:09",{"accountNo":180,"login":181,"displayName":182,"avatarUrl":183,"accountType":184,"accountTypeDesc":185,"profileDescription":186,"skillCount":187,"repoCount":188,"professionCount":136,"scenarioCount":108,"totalStars":189,"updatedAt":190},"01M6DJCCDDG53MTYYRTXAXFG17","garrytan","Garry Tan","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F19957?v=4","USER","个人作者","Writes software, dreams. ",340,3,175019,"2026-11-15 03:57:03",5953,9,{"skillCount":194,"professionCount":107,"scenarioCount":108,"creatorCount":191},482556,{"records":196,"total":194,"current":581,"size":126},[197,240,280,303,327,351,377,396,416,432,449,465,480,493,510,527,546,566],{"skillNo":198,"name":199,"nameZh":13,"description":200,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":209,"categories":210,"keywords":217,"publishedAt":239},"01M6FCCMPVDKCB2PWY97G8T056","chroma","Open-source embedding database for AI applications. Store embeddings and metadata, perform vector and full-text search, filter by metadata. Simple 4-function API. Scales from notebooks to production clusters. Use for semantic search, RAG applications, or document retrieval. Best for local development and open-source projects.","Dicklesworthstone\u002Fpi_agent_rust","Dicklesworthstone","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F35050222?v=4","01M6CG26AR942CSVMN1M5E3DSE","Jeff Emanuel","01M6CG26EPP71PSHW1VDS71121",1992,"MIT","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Frag-chroma\u002FSKILL.md",[211,214],{"categoryNo":212,"categoryName":213,"taxonomyType":53,"taxonomyTypeDesc":54},"01JBSEEDCAT000000000000002","文档处理",{"categoryNo":215,"categoryName":216,"taxonomyType":53,"taxonomyTypeDesc":54},"01JBSEEDCAT000000000000010","后端与接口",[199,218,219,220,221,222,223,224,225,226,227,228,229,230,231,232,233,234,235,236,237,238],"rag","semantic search","database","search","vector","embeddings","open source","Transformers","api","open","projects","text","vector database","self-hosted","self","local","full","filter","embedding","source","Clusters","2026-11-10 10:12:30",{"skillNo":241,"name":242,"nameZh":13,"description":243,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":244,"categories":245,"keywords":250,"publishedAt":239},"01M6FCC3A8QTCMKJ3BGJVKK6VA","outlines","Guarantee valid JSON\u002FXML\u002Fcode structure during generation, use Pydantic models for type-safe outputs, support local models (Transformers, vLLM), and maximize inference speed with Outlines - dottxt.ai's structured generation library","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fprompt-engineering-outlines\u002FSKILL.md",[246,249],{"categoryNo":247,"categoryName":248,"taxonomyType":53,"taxonomyTypeDesc":54},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":5,"categoryName":8,"taxonomyType":6,"taxonomyTypeDesc":7},[251,252,253,254,255,256,257,258,259,260,261,262,263,264,265,225,266,267,268,269,270,271,272,273,274,233,275,276,277,278,279],"Outlines","safe","inference","valid","research","support","json schema","code","library","engineering","Structured Generation","vLLM","Pydantic","Local Models","Grammar-Based Generation","Type Safety","Safety","json","grammar","prompt","use","schema","models","structure","speed","type","for","code structure","outputs",{"skillNo":281,"name":282,"nameZh":13,"description":283,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":284,"categories":285,"keywords":288,"publishedAt":239},"01M6FCBYSNPR340913DPSBJABN","instructor","Extract structured data from LLM responses with Pydantic validation, retry failed extractions automatically, parse complex JSON with type safety, and stream partial results with Instructor - battle-tested structured output library","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fprompt-engineering-instructor\u002FSKILL.md",[286,287],{"categoryNo":247,"categoryName":248,"taxonomyType":53,"taxonomyTypeDesc":54},{"categoryNo":23,"categoryName":24,"taxonomyType":6,"taxonomyTypeDesc":7},[282,289,255,290,131,291,292,259,260,293,263,266,267,294,268,270,295,276,296,297,298,299,300,301,302],"structured data","validation","streaming","extraction","data","parsing","results","structured output","Data Extraction","output","with","retry","stream","parse",{"skillNo":304,"name":305,"nameZh":13,"description":306,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":307,"categories":308,"keywords":313,"publishedAt":239},"01M6FCBT1RXNTCTQE7JVMG192P","guidance","Control LLM output with regex and grammars, guarantee valid JSON\u002FXML\u002Fcode generation, enforce structured formats, and build multi-step workflows with Guidance - Microsoft Research's constrained generation framework","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fprompt-engineering-guidance\u002FSKILL.md",[309,312],{"categoryNo":310,"categoryName":311,"taxonomyType":53,"taxonomyTypeDesc":54},"01JBSEEDCAT000000000000021","资料检索与研究",{"categoryNo":247,"categoryName":248,"taxonomyType":53,"taxonomyTypeDesc":54},[305,254,314,258,315,316,225,269,270,317,318,319,296,320,321,322,323,324,298,299,325,326],"workflows","framework","Code Generation","enforcement","enforce","regex","Constrained Generation","JSON Validation","Microsoft Research","Format Enforcement","Multi-Step Workflows","formats","orchestra",{"skillNo":328,"name":329,"nameZh":13,"description":330,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":331,"categories":332,"keywords":335,"publishedAt":239},"01M6FCBN31KQTMZGGQR2HHD8PG","dspy","Build complex AI systems with declarative programming, optimize prompts automatically, create modular RAG systems and agents with DSPy - Stanford NLP's framework for systematic LM programming","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fprompt-engineering-dspy\u002FSKILL.md",[333,334],{"categoryNo":247,"categoryName":248,"taxonomyType":53,"taxonomyTypeDesc":54},{"categoryNo":5,"categoryName":8,"taxonomyType":6,"taxonomyTypeDesc":7},[329,336,255,337,218,338,131,315,339,340,270,341,342,343,344,345,346,347,348,299,326,277,349,350],"nlp","modular","create","Optimization","programming","optimize","systematic","Declarative Programming","Prompt Optimization","LM Programming","Stanford NLP","Automatic Optimization","Modular AI","automatic","declarative",{"skillNo":352,"name":353,"nameZh":13,"description":354,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":355,"categories":356,"keywords":357,"publishedAt":239},"01M6FCB989YT362W8HEQ6CWCG0","verl-rl-training","Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fpost-training-verl\u002FSKILL.md",[],[358,359,360,361,362,363,364,365,366,367,368,369,262,225,370,371,271,372,373,299,374,326,375,277,376],"training","rl","verl","verl rl training","llms","distributed","infrastructure","Distributed Training","Post-Training","RLHF","PPO","Ray","learning","engine","algorithms","volcano","at","implementing","when",{"skillNo":378,"name":379,"nameZh":13,"description":380,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":381,"categories":382,"keywords":383,"publishedAt":239},"01M6FCB4K5B9K8FTSZE472T23F","fine-tuning-with-trl","Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO\u002FGRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fpost-training-trl-fine-tuning\u002FSKILL.md",[],[384,385,386,387,299,362,358,339,366,367,225,370,388,389,390,391,392,393,394,395,326,277,376],"tuning","fine tuning","fine tuning with trl","TRL","preferences","model","feedback","alignment","align","human","Accelerate","Instruction Tuning",{"skillNo":397,"name":398,"nameZh":13,"description":399,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":400,"categories":401,"keywords":402,"publishedAt":239},"01M6FCAZM3CF3E64PG8VWC5VZH","torchforge-rl-training","Provides guidance for PyTorch-native agentic RL using torchforge, Meta's library separating infra from algorithms. Use when you want clean RL abstractions, easy algorithm experimentation, or scalable training with Monarch and TorchTitan.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fpost-training-torchforge\u002FSKILL.md",[],[358,359,403,404,405,406,259,262,407,370,271,408,409,410,372,411,412,413,414,415,326,277,376],"torchforge","torchforge rl training","agentic","pytorch","Meta","agentic rl","clean","infra","experimentation","Scalable","torchtitan","native","you",{"skillNo":417,"name":418,"nameZh":13,"description":419,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":420,"categories":421,"keywords":422,"publishedAt":239},"01M6FCAW77KT230GR4TTJED9RC","slime-rl-training","Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM models, implementing custom data generation workflows, or needing tight Megatron-LM integration for RL scaling.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fpost-training-slime\u002FSKILL.md",[],[358,359,423,424,314,315,425,426,293,366,369,225,370,271,427,428,429,430,431,299,326,375,277,376],"slime rl training","slime","GRPO","SGLang","router","scaling","glm","megatron","Megatron-LM",{"skillNo":433,"name":434,"nameZh":13,"description":435,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":436,"categories":437,"keywords":438,"publishedAt":239},"01M6FCAR19CW04CWFE7VQ9V2ZC","simpo-training","Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO\u002FPPO.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fpost-training-simpo\u002FSKILL.md",[],[358,439,440,441,339,442,366,368,225,389,271,391,443,394,387,444,445,446,447,448,326,277,376],"simpo training","simpo","performance","DPO","free","Efficient Training","Preference Optimization","DPO Alternative","Reference-Free","LLM Alignment",{"skillNo":450,"name":451,"nameZh":13,"description":452,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":453,"categories":454,"keywords":455,"publishedAt":239},"01M6FCAJR7JQSS9YAD3R9P8YEG","openrlhf-training","High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fpost-training-openrlhf\u002FSKILL.md",[],[358,456,457,441,458,315,459,365,460,366,367,368,461,369,262,462,463,225,271,464,326,277],"OpenRLHF","openrlhf training","architecture","DeepSpeed","ZeRO","RLOO","Large Models","ZeRO-3","high-performance",{"skillNo":466,"name":467,"nameZh":13,"description":468,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":469,"categories":470,"keywords":471,"publishedAt":239},"01M6FCAERJEBW7BQ9EY3X5ECCT","miles-rl-training","Provides guidance for enterprise-grade RL training using miles, a production-ready fork of slime. Use when training large MoE models with FP8\u002FINT4, needing train-inference alignment, or requiring speculative RL for maximum throughput.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fpost-training-miles\u002FSKILL.md",[],[358,359,472,473,474,475,426,369,225,370,271,427,476,391,477,424,430,431,478,299,479,326,277,376],"miles rl training","miles","FP8","Enterprise","ready","moe","throughput","production-ready",{"skillNo":481,"name":482,"nameZh":13,"description":483,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":484,"categories":485,"keywords":486,"publishedAt":239},"01M6FCABK7CJ4HSREQRXFF7ZWZ","grpo-rl-training","Expert guidance for GRPO\u002FRL fine-tuning with TRL for reasoning and task-specific model training","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Fpost-training-grpo-rl-training\u002FSKILL.md",[],[358,425,359,487,488,442,366,367,368,225,384,370,389,489,490,491,296,387,492,298,326,277],"grpo rl training","Fine-Tuning","expert","task","modeling","Reward Modeling",{"skillNo":494,"name":495,"nameZh":13,"description":496,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":497,"categories":498,"keywords":499,"publishedAt":239},"01M6FCA1WB4KT5S5C3RQ8BB5ZP","hqq-quantization","Half-Quadratic Quantization for LLMs without calibration data. Use when quantizing models to 4\u002F3\u002F2-bit precision without needing calibration datasets, for fast quantization workflows, or when deploying with vLLM or HuggingFace Transformers.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Foptimization-hqq\u002FSKILL.md",[],[500,501,502,362,503,255,314,339,504,293,262,225,389,271,273,505,506,507,508,299,326,277,509,376],"Quantization","HQQ","hqq quantization","memory","HuggingFace","efficiency","bit","Memory Efficiency","Model Compression","calibration",{"skillNo":511,"name":512,"nameZh":13,"description":513,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":514,"categories":515,"keywords":516,"publishedAt":239},"01M6FC9Y5WR8AQ1D7MTE0PCKMV","gptq","Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with \u003C2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Foptimization-gptq\u002FSKILL.md",[],[517,362,503,358,339,518,519,520,366,225,384,271,521,522,506,523,524,525,526,415,326,277,376],"GPTQ","FP16","Fast Inference","perplexity","auto","minimal","4-Bit","Memory Optimization","Consumer GPUs","Group-Wise Quantization",{"skillNo":528,"name":529,"nameZh":13,"description":530,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":531,"categories":532,"keywords":533,"publishedAt":239},"01M6FC9TPBC7V9F8WH50S77NA0","gguf-quantization","GGUF format and llama.cpp quantization for efficient CPU\u002FGPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Foptimization-gguf\u002FSKILL.md",[],[534,500,535,536,255,537,538,339,539,389,271,273,540,541,506,542,543,544,545,508,326,277,376],"gguf quantization","GGUF","python","cpp","gpu","apple silicon","requirements","apple","silicon","CPU","llama.cpp","CPU Inference",{"skillNo":547,"name":548,"nameZh":13,"description":549,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":550,"categories":551,"keywords":552,"publishedAt":239},"01M6FC9Q3DTTZEDZB8K538DEYT","optimizing-attention-flash","Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training\u002Frunning transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Foptimization-flash-attention\u002FSKILL.md",[],[553,554,555,503,406,358,556,259,339,225,557,505,558,559,560,507,561,562,563,414,564,376,565],"flash","optimizing attention flash","attention","issues","context","tokens","Flash Attention","Attention Optimization","Speed Optimization","SDPA","H100","window","sequences",{"skillNo":567,"name":568,"nameZh":13,"description":569,"descriptionZh":13,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":13,"sourceHtmlUrl":570,"categories":571,"keywords":572,"publishedAt":239},"01M6FC9KF1EQ2CG0E9509AXFB1","quantizing-models-bitsandbytes","Quantizes LLMs to 8-bit or 4-bit for 50-75% memory reduction with minimal accuracy loss. Use when GPU memory is limited, need to fit larger models, or want faster inference. Supports INT8, NF4, FP4 formats, QLoRA training, and 8-bit optimizers. Works with HuggingFace Transformers.","https:\u002F\u002Fgithub.com\u002FDicklesworthstone\u002Fpi_agent_rust\u002Fblob\u002Fmain\u002Ftests\u002Fext_conformance\u002Fartifacts\u002Ftemplates-davila7\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fai-research\u002Foptimization-bitsandbytes\u002FSKILL.md",[],[273,573,574,362,503,358,339,575,225,271,522,506,576,394,523,524,577,578,579,580,325,326,277,376],"Bitsandbytes","quantizing models bitsandbytes","QLoRA","fit","8-Bit","NF4","INT8","Efficient Inference",450]