[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skills:stats":3,"skills:sce6":8,"skills:pro6":55,"skills:creators9":100,"skills:all:stars:10820:":195},{"skillCount":4,"professionCount":5,"scenarioCount":6,"creatorCount":7},196852,11,7,1926,[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",1,39458,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",26531,[],{"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",13569,[],{"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",6289,[],{"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",6029,[],{"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",5531,[],[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",13274,[],{"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",7775,[],{"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",4988,[],{"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",3940,[],{"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",3446,[],{"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",3316,[],{"records":101,"total":7,"current":16,"size":194},[102,114,123,133,143,153,163,172,182],{"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",4160,229,828772,"2026-11-13 04:10:52",{"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.",362,32,437711,"2026-11-13 03:02:05",{"accountNo":124,"login":125,"displayName":126,"avatarUrl":127,"accountType":107,"accountTypeDesc":108,"profileDescription":18,"skillCount":128,"repoCount":129,"professionCount":6,"scenarioCount":130,"totalStars":131,"updatedAt":132},"01M6AC29YGCR4S2NTJFV0G8B81","openai","OpenAI","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F14957082?v=4",171,23,6,342695,"2026-11-13 04:03:41",{"accountNo":134,"login":135,"displayName":136,"avatarUrl":137,"accountType":107,"accountTypeDesc":108,"profileDescription":18,"skillCount":138,"repoCount":139,"professionCount":140,"scenarioCount":6,"totalStars":141,"updatedAt":142},"01M6AC83E6K4FDKK7HZ8WNMQKE","apache","The Apache Software Foundation","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F47359?v=4",420,75,8,298176,"2026-11-13 03:14:28",{"accountNo":144,"login":145,"displayName":146,"avatarUrl":147,"accountType":107,"accountTypeDesc":108,"profileDescription":148,"skillCount":149,"repoCount":150,"professionCount":5,"scenarioCount":6,"totalStars":151,"updatedAt":152},"01M698VYEFDM0VJBZDYYJ2T83S","google","Google","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1342004?v=4","Google ❤️ Open Source",535,48,242366,"2026-11-13 04:17:09",{"accountNo":154,"login":155,"displayName":156,"avatarUrl":157,"accountType":107,"accountTypeDesc":108,"profileDescription":158,"skillCount":159,"repoCount":160,"professionCount":5,"scenarioCount":6,"totalStars":161,"updatedAt":162},"01M6AC83DS133TPXFSP1DHNCC6","github","GitHub","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F9919?v=4","How people build software.",940,28,238441,"2026-11-13 03:57:35",{"accountNo":164,"login":165,"displayName":166,"avatarUrl":167,"accountType":107,"accountTypeDesc":108,"profileDescription":18,"skillCount":168,"repoCount":169,"professionCount":5,"scenarioCount":6,"totalStars":170,"updatedAt":171},"01M697XGDPXJAVQ8391V2BCYTE","anthropics","Anthropic","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F76263028?v=4",674,17,231826,"2026-11-13 04:18:26",{"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",86,2,10,188014,"2026-11-13 02:38:59",{"accountNo":183,"login":184,"displayName":185,"avatarUrl":186,"accountType":187,"accountTypeDesc":188,"profileDescription":189,"skillCount":190,"repoCount":191,"professionCount":130,"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. ",324,3,174614,"2026-11-13 04:05:07",9,{"records":196,"total":4,"current":579,"size":580},[197,233,254,276,302,329,351,372,391,412,431,455,480,500,515,531,547,564],{"skillNo":198,"name":199,"nameZh":18,"description":200,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":209,"categories":210,"keywords":211,"publishedAt":232},"01M6XGCBKX711PC4RGG57DAHYA","model-pruning","Reduce LLM size and accelerate inference using pruning techniques like Wanda and SparseGPT. Use when compressing models without retraining, achieving 50% sparsity with minimal accuracy loss, or enabling faster inference on hardware accelerators. Covers unstructured pruning, structured pruning, N:M sparsity, magnitude pruning, and one-shot methods.","fadrienne\u002FMyGitHub","fadrienne","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F197874571?v=4","01M6X64V8VWCRSMFZ4EPS2BWXG","Francesca Adrienne","01M6X64VD5MW4F9PHDWQKS7ZAF",0,"MIT","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fmodel-pruning\u002FSKILL.md",[],[212,213,214,215,216,217,218,219,220,221,222,223,224,225,226,227,228,229,230,231],"model","Model Pruning","training","Fast Inference","Transformers","methods","synthetic","Accelerate","Model Compression","Emerging Techniques","Wanda","SparseGPT","Sparsity","N:M Sparsity","One-Shot Pruning","Structured Pruning","Unstructured Pruning","unstructured","reduce","when","2026-11-10 10:12:30",{"skillNo":234,"name":235,"nameZh":18,"description":236,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":237,"categories":238,"keywords":239,"publishedAt":232},"01M6XGC6RBDCKW5GATWMWZWYG8","model-merging","Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task Arithmetic, linear merging, and production deployment strategies.","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fmodel-merging\u002FSKILL.md",[],[212,240,241,242,214,243,216,244,245,246,247,218,221,248,249,250,251,252,253,231],"Model Merging","performance","merge","capabilities","math","specialized","strategies","fusion","Mergekit","SLERP","TIES","DARE","expertise","variants",{"skillNo":255,"name":256,"nameZh":18,"description":257,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":258,"categories":259,"keywords":260,"publishedAt":232},"01M6XGC2AYQYGWS41A2ZKBDHXH","model-economics","Cost modeling and ROI analysis for specialized LLM development. Use when deciding whether to train a custom model, estimating total cost, or calculating break-even vs frontier APIs. Covers training costs, inference costs, and time-to-ROI projections.","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fmodel-economics\u002FSKILL.md",[],[212,261,262,263,264,214,265,266,245,267,268,269,218,270,271,272,273,274,231,275],"economics","model economics","roi","cost analysis","development","analysis","business case","time","modeling","projections","roi analysis","for","break","flywheel","Training Cost",{"skillNo":277,"name":278,"nameZh":18,"description":279,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":280,"categories":281,"keywords":285,"publishedAt":232},"01M6XGBVWQ0TD9ZA896TFHYA15","mlflow","Track ML experiments, manage model registry with versioning, deploy models to production, and reproduce experiments with MLflow - framework-agnostic ML lifecycle platform","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fmlflow\u002FSKILL.md",[282],{"categoryNo":283,"categoryName":284,"taxonomyType":11,"taxonomyTypeDesc":12},"01JBSEEDCAT000000000000011","运维与部署",[278,286,287,288,214,289,290,291,292,212,293,294,295,218,296,297,298,299,300,301],"tensorflow","experiments","pytorch","learn","registry","framework","boto3","versioning","experiment","scikit-learn","Experiment Tracking","Model Registry","ML Lifecycle","Model Versioning","sqlalchemy","track",{"skillNo":303,"name":304,"nameZh":18,"description":305,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":306,"categories":307,"keywords":312,"publishedAt":232},"01M6XGBPR2XX2MTBWKNJFDXJ75","ml-benchmark-evaluation","Rigorous methodology for evaluating ML models on established benchmarks. Covers proper train\u002Fval\u002Ftest splits, baseline verification from original papers, exact metric formula discrepancies, data-leak detection checklist, multi-seed robustness, and honest reporting templates. Use when claiming to beat published baselines, writing methods papers, or auditing existing results.","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fml-benchmark-evaluation\u002FSKILL.md",[308,311],{"categoryNo":309,"categoryName":310,"taxonomyType":11,"taxonomyTypeDesc":12},"01JBSEEDCAT000000000000008","测试与质量",{"categoryNo":73,"categoryName":74,"taxonomyType":58,"taxonomyTypeDesc":59},[313,314,315,316,317,318,214,319,320,321,322,323,324,325,326,218,327,272,328,231],"ml","benchmark","Evaluation","Benchmark Evaluation","ml benchmark evaluation","writing","verification","metrics","reporting","data","data-leak","formula","reproducibility","seed","papers","metric",{"skillNo":330,"name":331,"nameZh":18,"description":332,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":333,"categories":334,"keywords":335,"publishedAt":232},"01M6XGBJKPVYT981DRGNMA2TMM","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\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fmegatron-core\u002FSKILL.md",[],[336,214,337,338,339,212,340,341,342,343,246,344,218,345,346,347,348,349,350,231],"llms","megatron","training llms megatron","pipeline","context","ready","expert","efficiency","apex","H100","Megatron-Core","Model Parallelism","advanced","production-ready","sequence",{"skillNo":352,"name":353,"nameZh":18,"description":354,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":355,"categories":356,"keywords":357,"publishedAt":232},"01M6XGBENG5820MW1Y8NZKG4JB","mamba-architecture","State-space model with O(n) complexity vs Transformers' O(n²). 5× faster inference, million-token sequences, no KV cache. Selective SSM with hardware-aware design. Mamba-1 (d_state=16) and Mamba-2 (d_state=128, multi-head). Models 130M-2.8B on HuggingFace.","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fmamba\u002FSKILL.md",[],[358,359,360,361,313,214,362,363,216,212,340,364,365,366,218,367,368,369,370,371],"architecture","Mamba","mamba architecture","design","linear","cache","complexity","kv cache","token","long","state-space","Efficient Inference","Model Architecture","sequences",{"skillNo":373,"name":374,"nameZh":18,"description":375,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":376,"categories":377,"keywords":378,"publishedAt":232},"01M6XGB767RJXG27KMG9D944R8","evaluating-llms-harness","Evaluates LLMs across 60+ academic benchmarks (MMLU, HumanEval, GSM8K, TruthfulQA, HellaSwag). Use when benchmarking model quality, comparing models, reporting academic results, or tracking training progress. Industry standard used by EleutherAI, HuggingFace, and major labs. Supports HuggingFace, vLLM, APIs.","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Flm-evaluation-harness\u002FSKILL.md",[],[336,379,380,381,214,321,315,382,383,384,216,212,385,386,387,218,388,389,390,231],"harness","evaluating llms harness","benchmarking","HumanEval","MMLU","vLLM","eval","use","across","progress","LM Evaluation Harness","GSM8K",{"skillNo":392,"name":393,"nameZh":18,"description":394,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":395,"categories":396,"keywords":398,"publishedAt":232},"01M6XGB3MBRH5YMMCH4587EMEZ","llama-factory","Expert guidance for fine-tuning LLMs with LLaMA-Factory - WebUI no-code, 100+ models, 2\u002F3\u002F4\u002F5\u002F6\u002F8-bit QLoRA, multimodal support","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fllama-factory\u002FSKILL.md",[397],{"categoryNo":57,"categoryName":60,"taxonomyType":58,"taxonomyTypeDesc":59},[399,400,401,336,313,214,402,403,404,405,406,342,407,408,218,219,409,410,411,272],"Llama","factory","LLaMA Factory","code","Fine-Tuning","LoRA","QLoRA","HuggingFace","webui","bit","8-Bit","No-Code","with",{"skillNo":413,"name":414,"nameZh":18,"description":415,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":416,"categories":417,"keywords":418,"publishedAt":232},"01M6XGAZQSBD98NXPGSMYRJZQA","implementing-llms-litgpt","Implements and trains LLMs using Lightning AI's LitGPT with 20+ pretrained architectures (Llama, Gemma, Phi, Qwen, Mistral). Use when need clean model implementations, educational understanding of architectures, or production fine-tuning with LoRA\u002FQLoRA. Single-file implementations, no abstraction layers.","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Flitgpt\u002FSKILL.md",[],[336,419,420,421,358,214,216,422,423,212,424,425,218,426,370,427,428,429,430,231],"LitGPT","implementing llms litgpt","implementing","tuning","phi","single-file","clean","architectures","Lightning AI","LLM Implementation","layers","single",{"skillNo":432,"name":433,"nameZh":18,"description":434,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":435,"categories":436,"keywords":440,"publishedAt":232},"01M6XGATRMGB43MKVWW47HY3G9","knowledge-distillation","Compress large language models using knowledge distillation from teacher to student models. Use when deploying smaller models with retained performance, transferring GPT-4 capabilities to open-source models, or reducing inference costs. Covers temperature scaling, soft targets, reverse KLD, logit distillation, and MiniLLM training strategies.","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fknowledge-distillation\u002FSKILL.md",[437],{"categoryNo":438,"categoryName":439,"taxonomyType":11,"taxonomyTypeDesc":12},"01JBSEEDCAT000000000000021","资料检索与研究",[441,442,443,241,214,444,212,445,218,221,446,447,448,449,450,451,452,453,454,231],"knowledge","distillation","Knowledge Distillation","open","scaling","Teacher-Student","MiniLLM","Reverse KLD","Soft Targets","Temperature Scaling","Logit Distillation","Model Transfer","student","teacher",{"skillNo":456,"name":457,"nameZh":18,"description":458,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":459,"categories":460,"keywords":464,"publishedAt":232},"01M6XGAM0T7N39H78J270FYS1A","hugging-face-model-trainer","This skill should be used when users want to train or fine-tune language models using TRL (Transformer Reinforcement Learning) on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV scripts with PEP 723 format, dataset preparation and validation, hardware selection, cost estimation, Trackio monitoring, Hub authentication, and model persistence. Should be invoked for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs without local GPU setup.","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fhugging-face-model-trainer\u002FSKILL.md",[461],{"categoryNo":462,"categoryName":463,"taxonomyType":11,"taxonomyTypeDesc":12},"01JBSEEDCAT000000000000023","Skill与智能体开发",[212,465,466,467,468,469,470,214,471,472,473,474,475,476,217,477,478,219,272,479],"hugging face","hugging face model trainer","model trainer","face","cost estimation","infrastructure","deployment","authentication","monitoring","learning","uv","selection","hugging face jobs","local","Cloud GPU",{"skillNo":481,"name":482,"nameZh":18,"description":483,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":484,"categories":485,"keywords":489,"publishedAt":232},"01M6XGAEDHQ8GJ8YYRPV4TM317","hugging-face-evaluation","Add and manage evaluation results in Hugging Face model cards. Supports extracting eval tables from README content, importing scores from Artificial Analysis API, and running custom model evaluations with vLLM\u002Flighteval. Works with the model-index metadata format.","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fhugging-face-evaluation\u002FSKILL.md",[486],{"categoryNo":487,"categoryName":488,"taxonomyType":11,"taxonomyTypeDesc":12},"01JBSEEDCAT000000000000010","后端与接口",[315,465,490,468,381,214,320,266,491,492,212,385,493,494,495,218,496,497,498,499],"hugging face evaluation","api","markdown","README","manage","dotenv","add","extracting","tables","running",{"skillNo":501,"name":502,"nameZh":18,"description":503,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":504,"categories":505,"keywords":506,"publishedAt":232},"01M6XGAAHDNP02BN97WGW3VSCA","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\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fhqq\u002FSKILL.md",[],[507,508,509,336,510,511,214,512,322,216,212,386,343,408,218,513,220,272,514,231],"Quantization","HQQ","hqq quantization","memory","workflows","Optimization","Memory Efficiency","calibration",{"skillNo":516,"name":517,"nameZh":18,"description":518,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":519,"categories":520,"keywords":521,"publishedAt":232},"01M6XGA6J9EKKM9E2BZSXCY0GM","grpo-rl-training","Expert guidance for GRPO\u002FRL fine-tuning with TRL for reasoning and task-specific model training","https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fgrpo-rl-training\u002FSKILL.md",[],[214,522,523,524,313,403,525,526,527,216,422,474,212,342,269,528,218,529,530,272],"GRPO","rl","grpo rl training","Post-Training","RLHF","PPO","structured output","TRL","Reward Modeling",{"skillNo":532,"name":533,"nameZh":18,"description":534,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":535,"categories":536,"keywords":537,"publishedAt":232},"01M6XGA1TMKG6MSZF7PTBHWR4K","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\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fgptq\u002FSKILL.md",[],[538,336,510,214,512,539,215,540,216,422,541,408,218,542,543,544,545,546,272,231],"GPTQ","FP16","perplexity","auto","4-Bit","Memory Optimization","Consumer GPUs","Group-Wise Quantization","you",{"skillNo":548,"name":549,"nameZh":18,"description":550,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":551,"categories":552,"keywords":553,"publishedAt":232},"01M6XG9MWSNEDBTRZ39MDH4K00","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\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fflash-attention\u002FSKILL.md",[],[554,555,556,510,288,214,557,216,343,218,558,559,513,560,561,345,562,563,231,371],"flash","optimizing attention flash","attention","issues","Flash Attention","Attention Optimization","Speed Optimization","SDPA","native","window",{"skillNo":565,"name":566,"nameZh":18,"description":567,"descriptionZh":18,"repositoryFullName":201,"ownerLogin":202,"ownerAvatarUrl":203,"ownerAccountNo":204,"ownerDisplayName":205,"repositoryNo":206,"starsCount":207,"licenseSpdxId":208,"licenseFileUrl":18,"sourceHtmlUrl":568,"categories":569,"keywords":570,"publishedAt":232},"01M6XG9GZPRHGP1JV956MWBNS1","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\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fml-training\u002Fdeepspeed\u002FSKILL.md",[],[571,572,313,214,573,574,575,576,512,577,539,578,339,342,556,408,218,219,411,272],"DeepSpeed","distributed","Distributed Training","ZeRO","Pipeline Parallelism","Mixed Precision","Large-Scale Training","FP8",10820,18]