[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skills:stats":3,"skills:sce6":8,"skills:pro6":55,"skills:creators9":100,"skills:all:stars:9988:":195},{"skillCount":4,"professionCount":5,"scenarioCount":6,"creatorCount":7},482556,11,7,5953,[9,20,27,34,41,48],{"categoryNo":10,"taxonomyType":11,"taxonomyTypeDesc":12,"categoryName":13,"categoryNameEn":14,"description":15,"level":16,"skillCount":17,"parentNo":18,"parentName":18,"children":19},"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",1,39675,null,[],{"categoryNo":21,"taxonomyType":11,"taxonomyTypeDesc":12,"categoryName":22,"categoryNameEn":23,"description":24,"level":16,"skillCount":25,"parentNo":18,"parentName":18,"children":26},"01JBSEEDCAT000000000000006","软件开发","Software Development","编码、测试、部署等软件研发任务",26798,[],{"categoryNo":28,"taxonomyType":11,"taxonomyTypeDesc":12,"categoryName":29,"categoryNameEn":30,"description":31,"level":16,"skillCount":32,"parentNo":18,"parentName":18,"children":33},"01JBSEEDCAT000000000000014","设计创意","Design & Creative","视觉设计、品牌与多媒体创作",13847,[],{"categoryNo":35,"taxonomyType":11,"taxonomyTypeDesc":12,"categoryName":36,"categoryNameEn":37,"description":38,"level":16,"skillCount":39,"parentNo":18,"parentName":18,"children":40},"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":42,"taxonomyType":11,"taxonomyTypeDesc":12,"categoryName":43,"categoryNameEn":44,"description":45,"level":16,"skillCount":46,"parentNo":18,"parentName":18,"children":47},"01JBSEEDCAT000000000000017","内容创作","Content Creation","写作、文案与营销内容生产",6063,[],{"categoryNo":49,"taxonomyType":11,"taxonomyTypeDesc":12,"categoryName":50,"categoryNameEn":51,"description":52,"level":16,"skillCount":53,"parentNo":18,"parentName":18,"children":54},"01JBSEEDCAT000000000000020","研究与学习","Research & Learning","资料检索、知识整理与学习辅导",5574,[],[56,65,72,79,86,93],{"categoryNo":57,"taxonomyType":58,"taxonomyTypeDesc":59,"categoryName":60,"categoryNameEn":61,"description":62,"level":16,"skillCount":63,"parentNo":18,"parentName":18,"children":64},"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",13341,[],{"categoryNo":66,"taxonomyType":58,"taxonomyTypeDesc":59,"categoryName":67,"categoryNameEn":68,"description":69,"level":16,"skillCount":70,"parentNo":18,"parentName":18,"children":71},"01JBSEEDCAT000000000000032","设计师","Designer","视觉、交互与品牌设计人员",7902,[],{"categoryNo":73,"taxonomyType":58,"taxonomyTypeDesc":59,"categoryName":74,"categoryNameEn":75,"description":76,"level":16,"skillCount":77,"parentNo":18,"parentName":18,"children":78},"01JBSEEDCAT000000000000033","数据分析师","Data Analyst","数据分析、报表与商业智能人员",5005,[],{"categoryNo":80,"taxonomyType":58,"taxonomyTypeDesc":59,"categoryName":81,"categoryNameEn":82,"description":83,"level":16,"skillCount":84,"parentNo":18,"parentName":18,"children":85},"01JBSEEDCAT000000000000036","内容创作者","Content Creator","作者、编辑与自媒体创作者",3978,[],{"categoryNo":87,"taxonomyType":58,"taxonomyTypeDesc":59,"categoryName":88,"categoryNameEn":89,"description":90,"level":16,"skillCount":91,"parentNo":18,"parentName":18,"children":92},"01JBSEEDCAT000000000000035","市场营销","Marketer","市场推广、增长与品牌运营人员",3454,[],{"categoryNo":94,"taxonomyType":58,"taxonomyTypeDesc":59,"categoryName":95,"categoryNameEn":96,"description":97,"level":16,"skillCount":98,"parentNo":18,"parentName":18,"children":99},"01JBSEEDCAT000000000000037","律师与法务","Legal","合同、合规与法律事务人员",3337,[],{"records":101,"total":7,"current":16,"size":194},[102,114,123,132,142,152,162,172,182],{"accountNo":103,"login":104,"displayName":105,"avatarUrl":106,"accountType":107,"accountTypeDesc":108,"profileDescription":109,"skillCount":110,"repoCount":111,"professionCount":5,"scenarioCount":6,"totalStars":112,"updatedAt":113},"01M698VYDXZ6JW5WH3CRV650R8","microsoft","Microsoft","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F6154722?v=4","ORGANIZATION","组织","Open source projects and samples from Microsoft",4300,236,832929,"2026-11-15 04:18:35",{"accountNo":115,"login":116,"displayName":116,"avatarUrl":117,"accountType":107,"accountTypeDesc":108,"profileDescription":118,"skillCount":119,"repoCount":120,"professionCount":6,"scenarioCount":6,"totalStars":121,"updatedAt":122},"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":124,"login":125,"displayName":126,"avatarUrl":127,"accountType":107,"accountTypeDesc":108,"profileDescription":18,"skillCount":128,"repoCount":129,"professionCount":5,"scenarioCount":6,"totalStars":130,"updatedAt":131},"01M697XGDPXJAVQ8391V2BCYTE","anthropics","Anthropic","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F76263028?v=4",716,18,427323,"2026-11-15 05:52:49",{"accountNo":133,"login":134,"displayName":135,"avatarUrl":136,"accountType":107,"accountTypeDesc":108,"profileDescription":18,"skillCount":137,"repoCount":138,"professionCount":6,"scenarioCount":139,"totalStars":140,"updatedAt":141},"01M6AC29YGCR4S2NTJFV0G8B81","openai","OpenAI","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F14957082?v=4",177,23,6,344400,"2026-11-15 03:07:28",{"accountNo":143,"login":144,"displayName":145,"avatarUrl":146,"accountType":107,"accountTypeDesc":108,"profileDescription":18,"skillCount":147,"repoCount":148,"professionCount":149,"scenarioCount":6,"totalStars":150,"updatedAt":151},"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":153,"login":154,"displayName":155,"avatarUrl":156,"accountType":107,"accountTypeDesc":108,"profileDescription":157,"skillCount":158,"repoCount":159,"professionCount":5,"scenarioCount":6,"totalStars":160,"updatedAt":161},"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":163,"login":164,"displayName":165,"avatarUrl":166,"accountType":107,"accountTypeDesc":108,"profileDescription":167,"skillCount":168,"repoCount":169,"professionCount":5,"scenarioCount":6,"totalStars":170,"updatedAt":171},"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":173,"login":174,"displayName":175,"avatarUrl":176,"accountType":107,"accountTypeDesc":108,"profileDescription":18,"skillCount":177,"repoCount":178,"professionCount":179,"scenarioCount":6,"totalStars":180,"updatedAt":181},"01M6AC83EPDGBFN19MWYRREJD0","Significant-Gravitas","AutoGPT","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F130738209?v=4",87,2,10,188029,"2026-11-15 06:21:09",{"accountNo":183,"login":184,"displayName":185,"avatarUrl":186,"accountType":187,"accountTypeDesc":188,"profileDescription":189,"skillCount":190,"repoCount":191,"professionCount":139,"scenarioCount":6,"totalStars":192,"updatedAt":193},"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",9,{"records":196,"total":4,"current":331,"size":129},[197,212,219,226,233,240,247,254,261,268,275,282,289,296,303,310,317,324],{"skillNo":198,"name":199,"nameZh":18,"description":200,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":208,"categories":209,"keywords":210,"publishedAt":211},"01M73AC3WPSDJCH2BQNV6ZY4MG","ray-train","Distributed training orchestration across clusters. Scales PyTorch\u002FTensorFlow\u002FHuggingFace from laptop to 1000s of nodes. Built-in hyperparameter tuning with Ray Tune, fault tolerance, elastic scaling. Use when training massive models across multiple machines or running distributed hyperparameter sweeps.","qhjqhj00\u002Fresearch-skills-pool","qhjqhj00","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F29684092?v=4","01M72AE0ZGXVM5C57N512DGQG8","Tommy Chien","01M72AE107ESJMHT7QC5SXX144","MIT","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--08-distributed-training__ray-train\u002FSKILL.md",[],[],"2026-11-11 15:11:14",{"skillNo":213,"name":214,"nameZh":18,"description":215,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":216,"categories":217,"keywords":218,"publishedAt":211},"01M73AC0P4RRYEPKJADB1QR39H","pytorch-lightning","High-level PyTorch framework with Trainer class, automatic distributed training (DDP\u002FFSDP\u002FDeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--08-distributed-training__pytorch-lightning\u002FSKILL.md",[],[],{"skillNo":220,"name":221,"nameZh":18,"description":222,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":223,"categories":224,"keywords":225,"publishedAt":211},"01M73ABW1ZP5C1EQM1F3SZ0S8H","pytorch-fsdp2","Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision\u002Foffload config, and distributed checkpointing. Use when models exceed single-GPU memory or when you need DTensor-based sharding with DeviceMesh.","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--08-distributed-training__pytorch-fsdp2\u002FSKILL.md",[],[],{"skillNo":227,"name":228,"nameZh":18,"description":229,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":230,"categories":231,"keywords":232,"publishedAt":211},"01M73ABR2D7ERRRMQ480E23A0N","training-llms-megatron","Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100), or require tensor\u002Fpipeline\u002Fsequence\u002Fcontext\u002Fexpert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--08-distributed-training__megatron-core\u002FSKILL.md",[],[],{"skillNo":234,"name":235,"nameZh":18,"description":236,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":237,"categories":238,"keywords":239,"publishedAt":211},"01M73ABMMDCDAFRT1539QV3PMP","deepspeed","Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16\u002FBF16\u002FFP8, 1-bit Adam, sparse attention","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--08-distributed-training__deepspeed\u002FSKILL.md",[],[],{"skillNo":241,"name":242,"nameZh":18,"description":243,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":244,"categories":245,"keywords":246,"publishedAt":211},"01M73ABGN27DMT0RBTQRCFX9BF","huggingface-accelerate","Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed\u002FFSDP\u002FMegatron\u002FDDP. Automatic device placement, mixed precision (FP16\u002FBF16\u002FFP8). Interactive config, single launch command. HuggingFace ecosystem standard.","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--08-distributed-training__accelerate\u002FSKILL.md",[],[],{"skillNo":248,"name":249,"nameZh":18,"description":250,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":251,"categories":252,"keywords":253,"publishedAt":211},"01M73ABDMTBHZR04062PDN47JB","prompt-guard","Meta's 86M prompt injection and jailbreak detector. Filters malicious prompts and third-party data for LLM apps. 99%+ TPR, \u003C1% FPR. Fast (\u003C2ms GPU). Multilingual (8 languages). Deploy with HuggingFace or batch processing for RAG security.","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--07-safety-alignment__prompt-guard\u002FSKILL.md",[],[],{"skillNo":255,"name":256,"nameZh":18,"description":257,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":258,"categories":259,"keywords":260,"publishedAt":211},"01M73ABAMC7KYX9G7WRTX7JFF5","nemo-guardrails","NVIDIA's runtime safety framework for LLM applications. Features jailbreak detection, input\u002Foutput validation, fact-checking, hallucination detection, PII filtering, toxicity detection. Uses Colang 2.0 DSL for programmable rails. Production-ready, runs on T4 GPU.","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--07-safety-alignment__nemo-guardrails\u002FSKILL.md",[],[],{"skillNo":262,"name":263,"nameZh":18,"description":264,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":265,"categories":266,"keywords":267,"publishedAt":211},"01M73AB72PH4BCTYVXZ6G4VHR2","llamaguard","Meta's 7-8B specialized moderation model for LLM input\u002Foutput filtering. 6 safety categories - violence\u002Fhate, sexual content, weapons, substances, self-harm, criminal planning. 94-95% accuracy. Deploy with vLLM, HuggingFace, Sagemaker. Integrates with NeMo Guardrails.","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--07-safety-alignment__llamaguard\u002FSKILL.md",[],[],{"skillNo":269,"name":270,"nameZh":18,"description":271,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":272,"categories":273,"keywords":274,"publishedAt":211},"01M73AB3Y46G8Q0VZDTF33NCRP","constitutional-ai","Anthropic's method for training harmless AI through self-improvement. Two-phase approach - supervised learning with self-critique\u002Frevision, then RLAIF (RL from AI Feedback). Use for safety alignment, reducing harmful outputs without human labels. Powers Claude's safety system.","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--07-safety-alignment__constitutional-ai\u002FSKILL.md",[],[],{"skillNo":276,"name":277,"nameZh":18,"description":278,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":279,"categories":280,"keywords":281,"publishedAt":211},"01M73AB00MDVX3T64D1G7H9CTZ","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\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--06-post-training__verl\u002FSKILL.md",[],[],{"skillNo":283,"name":284,"nameZh":18,"description":285,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":286,"categories":287,"keywords":288,"publishedAt":211},"01M73AAWH3K09FDBTF2ZH29D8D","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\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--06-post-training__trl-fine-tuning\u002FSKILL.md",[],[],{"skillNo":290,"name":291,"nameZh":18,"description":292,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":293,"categories":294,"keywords":295,"publishedAt":211},"01M73AAR5MB2SH7P9TXZDWW06V","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\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--06-post-training__torchforge\u002FSKILL.md",[],[],{"skillNo":297,"name":298,"nameZh":18,"description":299,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":300,"categories":301,"keywords":302,"publishedAt":211},"01M73AAMCMW4TA47MBYMF09890","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\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--06-post-training__slime\u002FSKILL.md",[],[],{"skillNo":304,"name":305,"nameZh":18,"description":306,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":307,"categories":308,"keywords":309,"publishedAt":211},"01M73AAG4VYWBCFE20V31Q6REP","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\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--06-post-training__simpo\u002FSKILL.md",[],[],{"skillNo":311,"name":312,"nameZh":18,"description":313,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":314,"categories":315,"keywords":316,"publishedAt":211},"01M73AAD7CJ6WAZE92J28Q021M","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\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--06-post-training__openrlhf\u002FSKILL.md",[],[],{"skillNo":318,"name":319,"nameZh":18,"description":320,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":321,"categories":322,"keywords":323,"publishedAt":211},"01M73AA997MK27KV18GQ98G53Z","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\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--06-post-training__miles\u002FSKILL.md",[],[],{"skillNo":325,"name":326,"nameZh":18,"description":327,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":191,"licenseSpdxId":207,"licenseFileUrl":18,"sourceHtmlUrl":328,"categories":329,"keywords":330,"publishedAt":211},"01M73AA6FTFJ64704BRRW37PJX","grpo-rl-training","Expert guidance for GRPO\u002FRL fine-tuning with TRL for reasoning and task-specific model training","https:\u002F\u002Fgithub.com\u002Fqhjqhj00\u002Fresearch-skills-pool\u002Fblob\u002Fmaster\u002Fskills_collected\u002FOrchestra-Research--AI-research-SKILLs--06-post-training__grpo-rl-training\u002FSKILL.md",[],[],9988]