[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"extension-skill-huggingface-train-sentence-transformers-zh-CN":3,"guides-for-huggingface-train-sentence-transformers":747,"similar-k17eq44byzzy319j4x3k26y4vx86n4a3-zh-CN":748},{"_creationTime":4,"_id":5,"children":6,"community":7,"display":9,"evaluation":15,"identity":254,"isFallback":251,"parentExtension":260,"providers":294,"relations":298,"repo":299,"tags":745,"workflow":746},1778690773482.4895,"k17eq44byzzy319j4x3k26y4vx86n4a3",[],{"reviewCount":8},0,{"description":10,"installMethods":11,"name":13,"sourceUrl":14},"Train or fine-tune sentence-transformers models across `SentenceTransformer` (bi-encoder; dense or static embedding model; for retrieval, similarity, clustering, classification, paraphrase mining, dedup, multimodal), `CrossEncoder` (reranker; pair scoring for two-stage retrieval / pair classification), and `SparseEncoder` (SPLADE, sparse embedding model; for learned-sparse retrieval). Covers loss selection, hard-negative mining, evaluators, distillation, LoRA, Matryoshka, and Hugging Face Hub publishing. Use for any sentence-transformers training task.",{"claudeCode":12},"huggingface/skills","Train Sentence-Transformers","https://github.com/huggingface/skills",{"_creationTime":16,"_id":17,"extensionId":5,"locale":18,"result":19,"trustSignals":235,"workflow":252},1778691446975.0137,"kn7es3x85jsrs3aacyqpsz4bg586n53q","en",{"checks":20,"evaluatedAt":190,"extensionSummary":191,"features":192,"nonGoals":198,"practices":203,"prerequisites":209,"promptVersionExtension":215,"promptVersionScoring":216,"purpose":217,"rationale":218,"score":219,"summary":220,"tags":221,"targetMarket":228,"tier":229,"useCases":230},[21,26,29,32,36,39,43,46,50,54,58,61,64,68,72,76,77,80,83,86,89,93,97,101,104,108,112,116,120,123,127,130,133,136,139,142,145,149,152,156,160,163,166,169,172,175,178,181,184,187],{"category":22,"check":23,"severity":24,"summary":25},"Practical Utility","Problem relevance","pass","The description clearly states the extension trains sentence-transformers models across three architectures (bi-encoder, cross-encoder, sparse-encoder) for various NLP tasks.",{"category":22,"check":27,"severity":24,"summary":28},"Unique selling proposition","The skill offers a comprehensive solution for training sentence-transformers models, covering aspects like loss selection, hard-negative mining, and Hugging Face Hub publishing, which goes beyond basic model loading.",{"category":22,"check":30,"severity":24,"summary":31},"Production readiness","The skill provides production-ready templates, detailed reference documentation for various training configurations, and clear error handling guidance, enabling immediate use for training tasks.",{"category":33,"check":34,"severity":24,"summary":35},"Scope","Single responsibility principle","The skill is focused on training sentence-transformers models, with all included scripts and references directly supporting this primary function.",{"category":33,"check":37,"severity":24,"summary":38},"Description quality","The displayed description accurately reflects the skill's capabilities, covering the different model types, training aspects, and intended use cases.",{"category":40,"check":41,"severity":24,"summary":42},"Invocation","Precise Purpose","The SKILL.md clearly outlines the model types (SentenceTransformer, CrossEncoder, SparseEncoder), tasks (retrieval, classification, reranking), and the purpose of training these models.",{"category":40,"check":44,"severity":24,"summary":45},"Concise Frontmatter","The SKILL.md frontmatter is concise and effectively summarizes the core capability for routing.",{"category":47,"check":48,"severity":24,"summary":49},"Documentation","Concise Body","The SKILL.md is well-structured, outlining the process and delegating detailed procedural content to separate reference files for progressive disclosure.",{"category":51,"check":52,"severity":24,"summary":53},"Context","Progressive Disclosure","Detailed information, procedures, and examples are appropriately organized into separate reference files (`references/`, `scripts/`) linked from the main SKILL.md.",{"category":51,"check":55,"severity":56,"summary":57},"Forked exploration","not_applicable","The skill is focused on providing training scripts and guidance, not deep exploration or code review, so a forked context is not applicable.",{"category":22,"check":59,"severity":24,"summary":60},"Usage examples","The script examples cover core functionalities like distillation, LoRA, and static embeddings, providing runnable code as starting points.",{"category":22,"check":62,"severity":24,"summary":63},"Edge cases","The troubleshooting reference documentation comprehensively addresses common failure modes like OOM, NaN loss, hanging training, and Hub push failures with recovery steps.",{"category":65,"check":66,"severity":56,"summary":67},"Code Execution","Tool Fallback","The skill does not appear to rely on optional external tools or MCP servers that would require fallbacks.",{"category":69,"check":70,"severity":56,"summary":71},"Safety","Halt on unexpected state","The skill focuses on training scripts and does not appear to interact with user workflows in a way that would require pre-state validation or rollback procedures.",{"category":73,"check":74,"severity":24,"summary":75},"Portability","Cross-skill coupling","The skill is self-contained and does not implicitly rely on other skills being loaded concurrently; it provides clear instructions for its own use.",{"category":40,"check":44,"severity":24,"summary":45},{"category":47,"check":78,"severity":24,"summary":79},"README","The README.md provides a good overview of Hugging Face skills, installation instructions for various agents, and a list of available skills.",{"category":33,"check":81,"severity":56,"summary":82},"Tool surface size","This is a skill that provides scripts and guidance, not a collection of tools to be enumerated.",{"category":40,"check":84,"severity":56,"summary":85},"Overlapping near-synonym tools","The skill does not expose multiple tools with overlapping functionality.",{"category":47,"check":87,"severity":24,"summary":88},"Phantom features","All advertised features, including various training techniques and examples, correspond to implemented scripts and reference documentation.",{"category":90,"check":91,"severity":24,"summary":92},"Install","Installation instruction","The README provides clear installation instructions for Claude Code, Codex, Gemini CLI, and Cursor, along with copy-pasteable examples.",{"category":94,"check":95,"severity":24,"summary":96},"Errors","Actionable error messages","The troubleshooting guide extensively details common errors, their causes, and remediation steps, providing actionable advice.",{"category":98,"check":99,"severity":24,"summary":100},"Execution","Pinned dependencies","The script headers (`# /// script ... dependencies = [...]`) clearly declare necessary Python dependencies, ensuring reproducible environments.",{"category":33,"check":102,"severity":56,"summary":103},"Dry-run preview","The skill is focused on model training, which is inherently a state-changing operation but does not have a distinct 'dry-run' concept.",{"category":105,"check":106,"severity":56,"summary":107},"Protocol","Idempotent retry & timeouts","The skill does not involve remote calls or state-changing operations that would require idempotency or timeouts.",{"category":109,"check":110,"severity":24,"summary":111},"Compliance","Telemetry opt-in","Telemetry via Trackio is documented as opt-in by default (report_to=\"none\" if SMOKE_TEST=1), with clear instructions on how to enable it.",{"category":113,"check":114,"severity":24,"summary":115},"License","License usability","The extension is distributed under the Apache-2.0 license, which is a permissive OSS license.",{"category":117,"check":118,"severity":24,"summary":119},"Maintenance","Commit recency","The repository shows recent commits within the last 3 months, indicating active maintenance.",{"category":117,"check":121,"severity":24,"summary":122},"Dependency Management","Dependencies are managed through script headers and specified requirements, allowing for suitable updates.",{"category":124,"check":125,"severity":24,"summary":126},"Security","Secret Management","Secrets like HF_TOKEN are handled via environment variables or job submission secrets, and not hardcoded.",{"category":124,"check":128,"severity":24,"summary":129},"Injection","The skill uses pre-defined scripts and datasets, and does not appear to load or execute untrusted third-party code dynamically.",{"category":124,"check":131,"severity":24,"summary":132},"Transitive Supply-Chain Grenades","The scripts rely on pinned dependencies specified in headers and do not fetch or execute external code at runtime.",{"category":124,"check":134,"severity":24,"summary":135},"Sandbox Isolation","The training scripts operate within their own execution context and do not manipulate files outside of their designated output directories.",{"category":124,"check":137,"severity":24,"summary":138},"Sandbox escape primitives","No detached process spawns or deny-retry loops were detected in the provided scripts.",{"category":124,"check":140,"severity":24,"summary":141},"Data Exfiltration","The skill does not reference confidential data and outbound calls are primarily for model hub interaction or optional telemetry, which are documented.",{"category":124,"check":143,"severity":24,"summary":144},"Hidden Text Tricks","The bundled content (SKILL.md, scripts, references) appears free of hidden-steering tricks or malicious Unicode characters.",{"category":146,"check":147,"severity":24,"summary":148},"Hooks","Opaque code execution","The provided scripts are plain Python and do not contain obfuscated code, base64 payloads, or runtime code fetching.",{"category":73,"check":150,"severity":24,"summary":151},"Structural Assumption","The scripts handle dataset loading and output paths robustly, and do not rely on specific user project structures.",{"category":153,"check":154,"severity":24,"summary":155},"Trust","Issues Attention","There are 4 open and 6 closed issues in the last 90 days, indicating good maintainer responsiveness (closure rate > 50%).",{"category":157,"check":158,"severity":24,"summary":159},"Versioning","Release Management","The SKILL.md frontmatter includes a version declaration, and the repository structure implies versioning through commits.",{"category":65,"check":161,"severity":24,"summary":162},"Validation","Input arguments and dataset handling in the scripts appear to follow standard Python practices and dataset library validation.",{"category":124,"check":164,"severity":56,"summary":165},"Unguarded Destructive Operations","The skill focuses on model training, which does not involve destructive operations like file deletion or system modification.",{"category":65,"check":167,"severity":24,"summary":168},"Error Handling","The scripts include error handling for common issues like data loading failures and Hub push errors, with informative logging.",{"category":65,"check":170,"severity":24,"summary":171},"Logging","The scripts tee output to logs, use trackio for monitoring, and provide clear VERDICT lines, facilitating auditability.",{"category":109,"check":173,"severity":56,"summary":174},"GDPR","The skill does not operate on personal data; it focuses on model training with public datasets.",{"category":109,"check":176,"severity":24,"summary":177},"Target market","The extension is designed for general use and does not exhibit any regional or jurisdictional limitations; targetMarket defaults to 'global'.",{"category":73,"check":179,"severity":24,"summary":180},"Runtime stability","The scripts declare Python version requirements and dependencies, and aim for cross-platform compatibility via standard libraries and `accelerate`.",{"category":33,"check":182,"severity":24,"summary":183},"Minimal I/O surface","The training scripts primarily interact with datasets and model checkpoints, with well-defined inputs and outputs for the training process.",{"category":33,"check":185,"severity":56,"summary":186},"Scoped tools","This is a skill providing training scripts, not a collection of distinct tools.",{"category":33,"check":188,"severity":56,"summary":189},"Tool naming","N/A as this skill does not expose discrete tools.",1778691446780,"This skill provides example training scripts and detailed reference documentation for training sentence-transformers models. It covers various architectures like bi-encoders, cross-encoders, and sparse encoders (SPLADE), along with techniques such as loss selection, hard-negative mining, distillation, LoRA, and Hugging Face Hub publishing.",[193,194,195,196,197],"Train bi-encoder, cross-encoder, and sparse-encoder models","Supports various training techniques (losses, mining, distillation, LoRA)","Includes runnable Python scripts for common scenarios","Provides detailed reference documentation for configuration and troubleshooting","Facilitates Hugging Face Hub publishing",[199,200,201,202],"Providing pre-trained models directly","Automating hyperparameter search (though references discuss it)","Executing training jobs without user intervention","Handling dataset creation from raw text",[204,205,206,207,208],"Model training","Contrastive learning","Distillation","Transfer learning","Model evaluation",[210,211,212,213,214],"pip install \"sentence-transformers[train]>=5.0\"","pip install datasets>=2.19.0","pip install accelerate>=0.26.0","pip install trackio","GPU strongly recommended","3.0.0","4.4.0","To enable users to train or fine-tune sentence-transformers models for diverse NLP tasks by providing example scripts, best practices, and comprehensive documentation.","The skill provides comprehensive, well-documented, and production-ready training scripts for various sentence-transformers architectures, with strong adherence to best practices and clear error handling.",98,"A robust skill for training sentence-transformers models, covering multiple architectures and advanced techniques.",[222,223,224,225,226,227],"sentence-transformers","nlp","machine-learning","deep-learning","fine-tuning","embeddings","global","verified",[231,232,233,234],"Training a custom embedding model for retrieval tasks","Fine-tuning a large language model for reranking using LoRA","Adapting a pre-trained model to a specific domain using distillation","Experimenting with different training strategies and hyperparameters",{"codeQuality":236,"collectedAt":238,"documentation":239,"maintenance":242,"security":248,"testCoverage":250},{"hasLockfile":237},false,1778691436469,{"descriptionLength":240,"readmeSize":241},558,9821,{"closedIssues90d":243,"forks":244,"hasChangelog":237,"openIssues90d":245,"pushedAt":246,"stars":247},6,663,4,1778593131000,10482,{"hasNpmPackage":237,"license":249,"smitheryVerified":237},"Apache-2.0",{"hasCi":251,"hasTests":237},true,{"updatedAt":253},1778691446975,{"basePath":255,"githubOwner":256,"githubRepo":257,"locale":18,"slug":258,"type":259},"skills/train-sentence-transformers","huggingface","skills","train-sentence-transformers","skill",{"_creationTime":261,"_id":262,"community":263,"display":264,"identity":269,"parentExtension":272,"providers":273,"relations":288,"tags":290,"workflow":291},1778690773482.486,"k175g1spb5757qt4tnj9cktcn986mshy",{"reviewCount":8},{"description":265,"installMethods":266,"name":268,"sourceUrl":14},"Agent Skills for AI/ML tasks including dataset creation, model training, evaluation, and research paper publishing on Hugging Face Hub",{"claudeCode":267},"huggingface-skills","Hugging Face Skills",{"basePath":270,"githubOwner":256,"githubRepo":257,"locale":18,"slug":257,"type":271},"","plugin",null,{"evaluate":274,"extract":283},{"promptVersionExtension":215,"promptVersionScoring":216,"score":219,"tags":275,"targetMarket":228,"tier":229},[256,276,277,278,279,280,281,282],"ai","ml","datasets","models","training","cli","python",{"commitSha":284,"license":249,"plugin":285},"HEAD",{"mcpCount":8,"provider":286,"skillCount":287},"classify",14,{"repoId":289},"kd72xwt5xnc0ktc4p7smzfcp3986m959",[276,281,278,256,277,279,282,280],{"evaluatedAt":292,"extractAt":293,"updatedAt":292},1778691185872,1778690773482,{"evaluate":295,"extract":297},{"promptVersionExtension":215,"promptVersionScoring":216,"score":219,"tags":296,"targetMarket":228,"tier":229},[222,223,224,225,226,227],{"commitSha":284,"license":249},{"parentExtensionId":262,"repoId":289},{"_creationTime":300,"_id":289,"identity":301,"providers":302,"workflow":741},1778689536128.5474,{"githubOwner":256,"githubRepo":257,"sourceUrl":14},{"classify":303,"discover":734,"github":737},{"commitSha":284,"extensions":304},[305,319,328,336,344,352,360,368,376,384,392,400,408,416,424,430,473,481,487,493,510,516,523,565,576,595,601,621,633,657,714],{"basePath":270,"description":265,"displayName":267,"installMethods":306,"rationale":307,"selectedPaths":308,"source":317,"sourceLanguage":18,"type":318},{"claudeCode":12},"marketplace.json at .claude-plugin/marketplace.json",[309,312,314],{"path":310,"priority":311},".claude-plugin/marketplace.json","mandatory",{"path":313,"priority":311},"README.md",{"path":315,"priority":316},"LICENSE","high","rule","marketplace",{"basePath":320,"description":321,"displayName":322,"installMethods":323,"rationale":324,"selectedPaths":325,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-llm-trainer","Train or fine-tune language models using TRL on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes hardware selection, cost estimation, Trackio monitoring, and Hub persistence.","huggingface-llm-trainer",{"claudeCode":322},"inline plugin source from marketplace.json at skills/huggingface-llm-trainer",[326],{"path":327,"priority":316},"SKILL.md",{"basePath":329,"description":330,"displayName":331,"installMethods":332,"rationale":333,"selectedPaths":334,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-local-models","Use to select models to run locally with llama.cpp and GGUF on CPU, Mac Metal, CUDA, or ROCm. Covers finding GGUFs, quant selection, running servers, exact GGUF file lookup, conversion, and OpenAI-compatible local serving.","huggingface-local-models",{"claudeCode":331},"inline plugin source from marketplace.json at skills/huggingface-local-models",[335],{"path":327,"priority":316},{"basePath":337,"description":338,"displayName":339,"installMethods":340,"rationale":341,"selectedPaths":342,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-paper-publisher","Publish and manage research papers on Hugging Face Hub. Supports creating paper pages, linking papers to models/datasets, claiming authorship, and generating professional markdown-based research articles.","huggingface-paper-publisher",{"claudeCode":339},"inline plugin source from marketplace.json at skills/huggingface-paper-publisher",[343],{"path":327,"priority":316},{"basePath":345,"description":346,"displayName":347,"installMethods":348,"rationale":349,"selectedPaths":350,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-papers","Look up and read Hugging Face paper pages in markdown, and use the papers API for structured metadata like authors, linked models, datasets, Spaces, and media URLs when needed.","huggingface-papers",{"claudeCode":347},"inline plugin source from marketplace.json at skills/huggingface-papers",[351],{"path":327,"priority":316},{"basePath":353,"description":354,"displayName":355,"installMethods":356,"rationale":357,"selectedPaths":358,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-community-evals","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 evaluations with vLLM/lighteval.","huggingface-community-evals",{"claudeCode":355},"inline plugin source from marketplace.json at skills/huggingface-community-evals",[359],{"path":327,"priority":316},{"basePath":361,"description":362,"displayName":363,"installMethods":364,"rationale":365,"selectedPaths":366,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-best","Find the best AI model for any task by querying Hugging Face leaderboards and benchmarks. Recommends top models based on task type, hardware constraints, and benchmark scores.","huggingface-best",{"claudeCode":363},"inline plugin source from marketplace.json at skills/huggingface-best",[367],{"path":327,"priority":316},{"basePath":369,"description":370,"displayName":371,"installMethods":372,"rationale":373,"selectedPaths":374,"source":317,"sourceLanguage":18,"type":271},"skills/hf-cli","Execute Hugging Face Hub operations using the hf CLI. Download models/datasets, upload files, manage repos, and run cloud compute jobs.","hf-cli",{"claudeCode":371},"inline plugin source from marketplace.json at skills/hf-cli",[375],{"path":327,"priority":316},{"basePath":377,"description":378,"displayName":379,"installMethods":380,"rationale":381,"selectedPaths":382,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-trackio","Track and visualize ML training experiments with Trackio. Log metrics via Python API and retrieve them via CLI. Supports real-time dashboards synced to HF Spaces.","huggingface-trackio",{"claudeCode":379},"inline plugin source from marketplace.json at skills/huggingface-trackio",[383],{"path":327,"priority":316},{"basePath":385,"description":386,"displayName":387,"installMethods":388,"rationale":389,"selectedPaths":390,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-datasets","Explore, query, and extract data from any Hugging Face dataset using the Dataset Viewer REST API and npx tooling. Zero Python dependencies — covers split/config discovery, row pagination, text search, filtering, SQL via parquetlens, and dataset upload via CLI.","huggingface-datasets",{"claudeCode":387},"inline plugin source from marketplace.json at skills/huggingface-datasets",[391],{"path":327,"priority":316},{"basePath":393,"description":394,"displayName":395,"installMethods":396,"rationale":397,"selectedPaths":398,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-tool-builder","Build reusable scripts for Hugging Face Hub and API workflows. Useful for chaining API calls, enriching Hub metadata, or automating repeated tasks.","huggingface-tool-builder",{"claudeCode":395},"inline plugin source from marketplace.json at skills/huggingface-tool-builder",[399],{"path":327,"priority":316},{"basePath":401,"description":402,"displayName":403,"installMethods":404,"rationale":405,"selectedPaths":406,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-gradio","Build Gradio web UIs and demos in Python. Use when creating or editing Gradio apps, components, event listeners, layouts, or chatbots.","huggingface-gradio",{"claudeCode":403},"inline plugin source from marketplace.json at skills/huggingface-gradio",[407],{"path":327,"priority":316},{"basePath":409,"description":410,"displayName":411,"installMethods":412,"rationale":413,"selectedPaths":414,"source":317,"sourceLanguage":18,"type":271},"skills/transformers-js","Run state-of-the-art machine learning models directly in JavaScript/TypeScript for NLP, computer vision, audio processing, and multimodal tasks. Works in Node.js and browsers with WebGPU/WASM using Hugging Face models.","transformers-js",{"claudeCode":411},"inline plugin source from marketplace.json at skills/transformers-js",[415],{"path":327,"priority":316},{"basePath":417,"description":418,"displayName":419,"installMethods":420,"rationale":421,"selectedPaths":422,"source":317,"sourceLanguage":18,"type":271},"skills/huggingface-vision-trainer","Train and fine-tune object detection models (RTDETRv2, YOLOS, DETR and others) and image classification models (timm and transformers models — MobileNetV3, MobileViT, ResNet, ViT/DINOv3) using Transformers Trainer API on Hugging Face Jobs infrastructure or locally. Includes COCO dataset format support, Albumentations augmentation, mAP/mAR metrics, trackio tracking, hardware selection, and Hub persistence.","huggingface-vision-trainer",{"claudeCode":419},"inline plugin source from marketplace.json at skills/huggingface-vision-trainer",[423],{"path":327,"priority":316},{"basePath":255,"description":425,"displayName":258,"installMethods":426,"rationale":427,"selectedPaths":428,"source":317,"sourceLanguage":18,"type":271},"Train or fine-tune sentence-transformers models across all three architectures: SentenceTransformer (bi-encoder embeddings), CrossEncoder (rerankers), and SparseEncoder (SPLADE). Covers loss selection, hard-negative mining, evaluators, distillation, LoRA, Matryoshka, and Hugging Face Hub publishing.",{"claudeCode":258},"inline plugin source from marketplace.json at skills/train-sentence-transformers",[429],{"path":327,"priority":316},{"basePath":270,"description":265,"displayName":267,"installMethods":431,"license":249,"rationale":432,"selectedPaths":433,"source":317,"sourceLanguage":18,"type":271},{"claudeCode":267},"plugin manifest at .claude-plugin/plugin.json",[434,436,437,438,441,443,445,447,449,451,453,455,457,459,461,463,465,467,469,471],{"path":435,"priority":311},".claude-plugin/plugin.json",{"path":313,"priority":311},{"path":315,"priority":316},{"path":439,"priority":440},"skills/hf-cli/SKILL.md","medium",{"path":442,"priority":440},"skills/huggingface-best/SKILL.md",{"path":444,"priority":440},"skills/huggingface-community-evals/SKILL.md",{"path":446,"priority":440},"skills/huggingface-datasets/SKILL.md",{"path":448,"priority":440},"skills/huggingface-gradio/SKILL.md",{"path":450,"priority":440},"skills/huggingface-llm-trainer/SKILL.md",{"path":452,"priority":440},"skills/huggingface-local-models/SKILL.md",{"path":454,"priority":440},"skills/huggingface-paper-publisher/SKILL.md",{"path":456,"priority":440},"skills/huggingface-papers/SKILL.md",{"path":458,"priority":440},"skills/huggingface-tool-builder/SKILL.md",{"path":460,"priority":440},"skills/huggingface-trackio/SKILL.md",{"path":462,"priority":440},"skills/huggingface-vision-trainer/SKILL.md",{"path":464,"priority":440},"skills/train-sentence-transformers/SKILL.md",{"path":466,"priority":440},"skills/transformers-js/SKILL.md",{"path":468,"priority":311},".mcp.json",{"path":470,"priority":316},"agents/AGENTS.md",{"path":472,"priority":316},".cursor-plugin/plugin.json",{"basePath":474,"description":475,"displayName":476,"installMethods":477,"rationale":478,"selectedPaths":479,"source":317,"sourceLanguage":18,"type":259},"hf-mcp/skills/hf-mcp","Use Hugging Face Hub via MCP server tools. Search models, datasets, Spaces, papers. Get repo details, fetch documentation, run compute jobs, and use Gradio Spaces as AI tools. Available when connected to the HF MCP server.","hf-mcp",{"claudeCode":12},"SKILL.md frontmatter at hf-mcp/skills/hf-mcp/SKILL.md",[480],{"path":327,"priority":311},{"basePath":369,"description":482,"displayName":371,"installMethods":483,"rationale":484,"selectedPaths":485,"source":317,"sourceLanguage":18,"type":259},"Hugging Face Hub CLI (`hf`) for downloading, uploading, and managing models, datasets, spaces, buckets, repos, papers, jobs, and more on the Hugging Face Hub. Use when: handling authentication; managing local cache; managing Hugging Face Buckets; running or scheduling jobs on Hugging Face infrastructure; managing Hugging Face repos; discussions and pull requests; browsing models, datasets and spaces; reading, searching, or browsing academic papers; managing collections; querying datasets; configuring spaces; setting up webhooks; or deploying and managing HF Inference Endpoints. Make sure to use this skill whenever the user mentions 'hf', 'huggingface', 'Hugging Face', 'huggingface-cli', or 'hugging face cli', or wants to do anything related to the Hugging Face ecosystem and to AI and ML in general. Also use for cloud storage needs like training checkpoints, data pipelines, or agent traces. Use even if the user doesn't explicitly ask for a CLI command. Replaces the deprecated `huggingface-cli`.",{"claudeCode":12},"SKILL.md frontmatter at skills/hf-cli/SKILL.md",[486],{"path":327,"priority":311},{"basePath":361,"description":488,"displayName":363,"installMethods":489,"rationale":490,"selectedPaths":491,"source":317,"sourceLanguage":18,"type":259},"Use when the user asks about finding the best, top, or recommended model for a task, wants to know what AI model to use, or wants to compare models by benchmark scores. Triggers on: \"best model for X\", \"what model should I use for\", \"top models for [task]\", \"which model runs on my laptop/machine/device\", \"recommend a model for\", \"what LLM should I use for\", \"compare models for\", \"what's state of the art for\", or any question about choosing an AI model for a specific use case. Always use this skill when the user wants model recommendations or comparisons, even if they don't explicitly mention HuggingFace or benchmarks.\n",{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-best/SKILL.md",[492],{"path":327,"priority":311},{"basePath":353,"description":494,"displayName":355,"installMethods":495,"rationale":496,"selectedPaths":497,"source":317,"sourceLanguage":18,"type":259},"Run evaluations for Hugging Face Hub models using inspect-ai and lighteval on local hardware. Use for backend selection, local GPU evals, and choosing between vLLM / Transformers / accelerate. Not for HF Jobs orchestration, model-card PRs, .eval_results publication, or community-evals automation.",{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-community-evals/SKILL.md",[498,499,502,504,506,508],{"path":327,"priority":311},{"path":500,"priority":501},"examples/.env.example","low",{"path":503,"priority":501},"examples/USAGE_EXAMPLES.md",{"path":505,"priority":501},"scripts/inspect_eval_uv.py",{"path":507,"priority":501},"scripts/inspect_vllm_uv.py",{"path":509,"priority":501},"scripts/lighteval_vllm_uv.py",{"basePath":385,"description":511,"displayName":387,"installMethods":512,"rationale":513,"selectedPaths":514,"source":317,"sourceLanguage":18,"type":259},"Use this skill for Hugging Face Dataset Viewer API workflows that fetch subset/split metadata, paginate rows, search text, apply filters, download parquet URLs, and read size or statistics.\r",{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-datasets/SKILL.md",[515],{"path":327,"priority":311},{"basePath":401,"description":402,"displayName":403,"installMethods":517,"rationale":518,"selectedPaths":519,"source":317,"sourceLanguage":18,"type":259},{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-gradio/SKILL.md",[520,521],{"path":327,"priority":311},{"path":522,"priority":440},"examples.md",{"basePath":320,"description":524,"displayName":322,"installMethods":525,"rationale":526,"selectedPaths":527,"source":317,"sourceLanguage":18,"type":259},"Train or fine-tune language and vision models using TRL (Transformer Reinforcement Learning) or Unsloth with 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, model selection/leaderboards and model persistence. Use for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs without local GPU setup.",{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-llm-trainer/SKILL.md",[528,529,531,533,535,537,539,541,543,545,547,549,551,553,555,557,559,561,563],{"path":327,"priority":311},{"path":530,"priority":440},"references/gguf_conversion.md",{"path":532,"priority":440},"references/hardware_guide.md",{"path":534,"priority":440},"references/hub_saving.md",{"path":536,"priority":440},"references/local_training_macos.md",{"path":538,"priority":440},"references/reliability_principles.md",{"path":540,"priority":440},"references/trackio_guide.md",{"path":542,"priority":440},"references/training_methods.md",{"path":544,"priority":440},"references/training_patterns.md",{"path":546,"priority":440},"references/troubleshooting.md",{"path":548,"priority":440},"references/unsloth.md",{"path":550,"priority":501},"scripts/convert_to_gguf.py",{"path":552,"priority":501},"scripts/dataset_inspector.py",{"path":554,"priority":501},"scripts/estimate_cost.py",{"path":556,"priority":501},"scripts/hf_benchmarks.py",{"path":558,"priority":501},"scripts/train_dpo_example.py",{"path":560,"priority":501},"scripts/train_grpo_example.py",{"path":562,"priority":501},"scripts/train_sft_example.py",{"path":564,"priority":501},"scripts/unsloth_sft_example.py",{"basePath":329,"description":330,"displayName":331,"installMethods":566,"rationale":567,"selectedPaths":568,"source":317,"sourceLanguage":18,"type":259},{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-local-models/SKILL.md",[569,570,572,574],{"path":327,"priority":311},{"path":571,"priority":440},"references/hardware.md",{"path":573,"priority":440},"references/hub-discovery.md",{"path":575,"priority":440},"references/quantization.md",{"basePath":337,"description":338,"displayName":339,"installMethods":577,"rationale":578,"selectedPaths":579,"source":317,"sourceLanguage":18,"type":259},{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-paper-publisher/SKILL.md",[580,581,583,585,587,589,591,593],{"path":327,"priority":311},{"path":582,"priority":501},"examples/example_usage.md",{"path":584,"priority":440},"references/quick_reference.md",{"path":586,"priority":501},"scripts/paper_manager.py",{"path":588,"priority":501},"templates/arxiv.md",{"path":590,"priority":501},"templates/ml-report.md",{"path":592,"priority":501},"templates/modern.md",{"path":594,"priority":501},"templates/standard.md",{"basePath":345,"description":596,"displayName":347,"installMethods":597,"rationale":598,"selectedPaths":599,"source":317,"sourceLanguage":18,"type":259},"Look up and read Hugging Face paper pages in markdown, and use the papers API for structured metadata such as authors, linked models/datasets/spaces, Github repo and project page. Use when the user shares a Hugging Face paper page URL, an arXiv URL or ID, or asks to summarize, explain, or analyze an AI research paper.",{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-papers/SKILL.md",[600],{"path":327,"priority":311},{"basePath":393,"description":602,"displayName":395,"installMethods":603,"rationale":604,"selectedPaths":605,"source":317,"sourceLanguage":18,"type":259},"Use this skill when the user wants to build tool/scripts or achieve a task where using data from the Hugging Face API would help. This is especially useful when chaining or combining API calls or the task will be repeated/automated. This Skill creates a reusable script to fetch, enrich or process data.",{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-tool-builder/SKILL.md",[606,607,609,611,613,615,617,619],{"path":327,"priority":311},{"path":608,"priority":440},"references/baseline_hf_api.py",{"path":610,"priority":440},"references/baseline_hf_api.sh",{"path":612,"priority":440},"references/baseline_hf_api.tsx",{"path":614,"priority":440},"references/find_models_by_paper.sh",{"path":616,"priority":440},"references/hf_enrich_models.sh",{"path":618,"priority":440},"references/hf_model_card_frontmatter.sh",{"path":620,"priority":440},"references/hf_model_papers_auth.sh",{"basePath":377,"description":622,"displayName":379,"installMethods":623,"rationale":624,"selectedPaths":625,"source":317,"sourceLanguage":18,"type":259},"Track and visualize ML training experiments with Trackio. Use when logging metrics during training (Python API), firing alerts for training diagnostics, or retrieving/analyzing logged metrics (CLI). Supports real-time dashboard visualization, alerts with webhooks, HF Space syncing, and JSON output for automation.",{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-trackio/SKILL.md",[626,627,629,631],{"path":327,"priority":311},{"path":628,"priority":440},"references/alerts.md",{"path":630,"priority":440},"references/logging_metrics.md",{"path":632,"priority":440},"references/retrieving_metrics.md",{"basePath":417,"description":634,"displayName":419,"installMethods":635,"rationale":636,"selectedPaths":637,"source":317,"sourceLanguage":18,"type":259},"Trains and fine-tunes vision models for object detection (D-FINE, RT-DETR v2, DETR, YOLOS), image classification (timm models — MobileNetV3, MobileViT, ResNet, ViT/DINOv3 — plus any Transformers classifier), and SAM/SAM2 segmentation using Hugging Face Transformers on Hugging Face Jobs cloud GPUs. Covers COCO-format dataset preparation, Albumentations augmentation, mAP/mAR evaluation, accuracy metrics, SAM segmentation with bbox/point prompts, DiceCE loss, hardware selection, cost estimation, Trackio monitoring, and Hub persistence. Use when users mention training object detection, image classification, SAM, SAM2, segmentation, image matting, DETR, D-FINE, RT-DETR, ViT, timm, MobileNet, ResNet, bounding box models, or fine-tuning vision models on Hugging Face Jobs.",{"claudeCode":12},"SKILL.md frontmatter at skills/huggingface-vision-trainer/SKILL.md",[638,639,641,642,644,646,647,649,650,651,653,655],{"path":327,"priority":311},{"path":640,"priority":440},"references/finetune_sam2_trainer.md",{"path":534,"priority":440},{"path":643,"priority":440},"references/image_classification_training_notebook.md",{"path":645,"priority":440},"references/object_detection_training_notebook.md",{"path":538,"priority":440},{"path":648,"priority":440},"references/timm_trainer.md",{"path":552,"priority":501},{"path":554,"priority":501},{"path":652,"priority":501},"scripts/image_classification_training.py",{"path":654,"priority":501},"scripts/object_detection_training.py",{"path":656,"priority":501},"scripts/sam_segmentation_training.py",{"basePath":255,"description":10,"displayName":258,"installMethods":658,"rationale":659,"selectedPaths":660,"source":317,"sourceLanguage":18,"type":259},{"claudeCode":12},"SKILL.md frontmatter at skills/train-sentence-transformers/SKILL.md",[661,662,664,666,668,670,672,673,675,677,679,681,683,685,687,688,690,692,694,696,698,700,702,704,706,708,710,712],{"path":327,"priority":311},{"path":663,"priority":440},"references/base_model_selection.md",{"path":665,"priority":440},"references/dataset_formats.md",{"path":667,"priority":440},"references/evaluators_cross_encoder.md",{"path":669,"priority":440},"references/evaluators_sentence_transformer.md",{"path":671,"priority":440},"references/evaluators_sparse_encoder.md",{"path":532,"priority":440},{"path":674,"priority":440},"references/hf_jobs_execution.md",{"path":676,"priority":440},"references/losses_cross_encoder.md",{"path":678,"priority":440},"references/losses_sentence_transformer.md",{"path":680,"priority":440},"references/losses_sparse_encoder.md",{"path":682,"priority":440},"references/model_architectures.md",{"path":684,"priority":440},"references/prompts_and_instructions.md",{"path":686,"priority":440},"references/training_args.md",{"path":546,"priority":440},{"path":689,"priority":501},"scripts/mine_hard_negatives.py",{"path":691,"priority":501},"scripts/train_cross_encoder_distillation_example.py",{"path":693,"priority":501},"scripts/train_cross_encoder_example.py",{"path":695,"priority":501},"scripts/train_cross_encoder_listwise_example.py",{"path":697,"priority":501},"scripts/train_sentence_transformer_distillation_example.py",{"path":699,"priority":501},"scripts/train_sentence_transformer_example.py",{"path":701,"priority":501},"scripts/train_sentence_transformer_make_multilingual_example.py",{"path":703,"priority":501},"scripts/train_sentence_transformer_matryoshka_example.py",{"path":705,"priority":501},"scripts/train_sentence_transformer_multi_dataset_example.py",{"path":707,"priority":501},"scripts/train_sentence_transformer_static_embedding_example.py",{"path":709,"priority":501},"scripts/train_sentence_transformer_with_lora_example.py",{"path":711,"priority":501},"scripts/train_sparse_encoder_distillation_example.py",{"path":713,"priority":501},"scripts/train_sparse_encoder_example.py",{"basePath":409,"description":715,"displayName":411,"installMethods":716,"rationale":717,"selectedPaths":718,"source":317,"sourceLanguage":18,"type":259},"Use Transformers.js to run state-of-the-art machine learning models directly in JavaScript/TypeScript. Supports NLP (text classification, translation, summarization), computer vision (image classification, object detection), audio (speech recognition, audio classification), and multimodal tasks. Works in browsers and server-side runtimes (Node.js, Bun, Deno) with WebGPU/WASM using pre-trained models from Hugging Face Hub.",{"claudeCode":12},"SKILL.md frontmatter at skills/transformers-js/SKILL.md",[719,720,722,724,726,728,730,732],{"path":327,"priority":311},{"path":721,"priority":440},"references/CACHE.md",{"path":723,"priority":440},"references/CONFIGURATION.md",{"path":725,"priority":440},"references/EXAMPLES.md",{"path":727,"priority":440},"references/MODEL_ARCHITECTURES.md",{"path":729,"priority":440},"references/MODEL_REGISTRY.md",{"path":731,"priority":440},"references/PIPELINE_OPTIONS.md",{"path":733,"priority":440},"references/TEXT_GENERATION.md",{"sources":735},[736],"manual",{"closedIssues90d":243,"description":738,"forks":244,"homepage":739,"license":249,"openIssues90d":245,"pushedAt":246,"readmeSize":241,"stars":247,"topics":740},"Give your agents the power of the Hugging Face ecosystem","https://huggingface.co",[],{"classifiedAt":742,"discoverAt":743,"extractAt":744,"githubAt":744,"updatedAt":742},1778690772996,1778689536128,1778690770714,[225,227,226,224,223,222],{"evaluatedAt":253,"extractAt":293,"updatedAt":253},[],[749,777,806,829,849,871],{"_creationTime":750,"_id":751,"community":752,"display":753,"identity":759,"providers":764,"relations":771,"tags":773,"workflow":774},1778691799740.4905,"k17c27dcgjsqmxeggb19stv4xn86mf1z",{"reviewCount":8},{"description":754,"installMethods":755,"name":757,"sourceUrl":758},"Deep learning framework (PyTorch Lightning). Organize PyTorch code into LightningModules, configure Trainers for multi-GPU/TPU, implement data pipelines, callbacks, logging (W&B, TensorBoard), distributed training (DDP, FSDP, DeepSpeed), for scalable neural network training.",{"claudeCode":756},"K-Dense-AI/claude-scientific-skills","PyTorch Lightning","https://github.com/K-Dense-AI/claude-scientific-skills",{"basePath":760,"githubOwner":761,"githubRepo":762,"locale":18,"slug":763,"type":259},"scientific-skills/pytorch-lightning","K-Dense-AI","claude-scientific-skills","pytorch-lightning",{"evaluate":765,"extract":770},{"promptVersionExtension":215,"promptVersionScoring":216,"score":766,"tags":767,"targetMarket":228,"tier":229},100,[768,225,224,282,769],"pytorch","framework",{"commitSha":284,"license":249},{"repoId":772},"kd79rphh5gexy91xmpxc05h5mh86mm9r",[225,769,224,282,768],{"evaluatedAt":775,"extractAt":776,"updatedAt":775},1778693958717,1778691799740,{"_creationTime":778,"_id":779,"community":780,"display":781,"identity":787,"providers":792,"relations":800,"tags":802,"workflow":803},1778685991755.7163,"k17a5r03jke0mcdvkknyjfmsgh86mnqt",{"reviewCount":8},{"description":782,"installMethods":783,"name":785,"sourceUrl":786},"Provides guidance for interpreting and manipulating neural network internals using nnsight with optional NDIF remote execution. Use when needing to run interpretability experiments on massive models (70B+) without local GPU resources, or when working with any PyTorch architecture.",{"claudeCode":784},"davila7/claude-code-templates","nnsight-remote-interpretability","https://github.com/davila7/claude-code-templates",{"basePath":788,"githubOwner":789,"githubRepo":790,"locale":18,"slug":791,"type":259},"cli-tool/components/skills/ai-research/mechanistic-interpretability-nnsight","davila7","claude-code-templates","mechanistic-interpretability-nnsight",{"evaluate":793,"extract":799},{"promptVersionExtension":215,"promptVersionScoring":216,"score":794,"tags":795,"targetMarket":228,"tier":229},99,[796,768,797,224,225,798],"nnsight","interpretability","remote-execution",{"commitSha":284},{"repoId":801},"kd71fzn4s7r0269fkw47wt670n86ndz0",[225,797,224,796,768,798],{"evaluatedAt":804,"extractAt":805,"updatedAt":804},1778687638846,1778685991755,{"_creationTime":807,"_id":808,"community":809,"display":810,"identity":814,"providers":816,"relations":825,"tags":826,"workflow":827},1778691799740.4753,"k17dds8s9gvgqz6gqrb1r420fn86n2rv",{"reviewCount":8},{"description":811,"installMethods":812,"name":813,"sourceUrl":758},"This skill should be used when working with genomic interval data (BED files) for machine learning tasks. Use for training region embeddings (Region2Vec, BEDspace), single-cell ATAC-seq analysis (scEmbed), building consensus peaks (universes), or any ML-based analysis of genomic regions. Applies to BED file collections, scATAC-seq data, chromatin accessibility datasets, and region-based genomic feature learning.",{"claudeCode":756},"geniml",{"basePath":815,"githubOwner":761,"githubRepo":762,"locale":18,"slug":813,"type":259},"scientific-skills/geniml",{"evaluate":817,"extract":823},{"promptVersionExtension":215,"promptVersionScoring":216,"score":794,"tags":818,"targetMarket":228,"tier":229},[819,820,224,227,821,822],"genomics","bioinformatics","bed-files","scatac-seq",{"commitSha":284,"license":824},"BSD-2-Clause",{"repoId":772},[821,820,227,819,224,822],{"evaluatedAt":828,"extractAt":776,"updatedAt":828},1778692637736,{"_creationTime":830,"_id":831,"community":832,"display":833,"identity":835,"providers":836,"relations":845,"tags":846,"workflow":847},1778690773482.4878,"k17aqa68b1vx1r0j9feqm2kggh86nv3v",{"reviewCount":8},{"description":524,"installMethods":834,"name":322,"sourceUrl":14},{"claudeCode":12},{"basePath":320,"githubOwner":256,"githubRepo":257,"locale":18,"slug":322,"type":259},{"evaluate":837,"extract":844},{"promptVersionExtension":215,"promptVersionScoring":216,"score":794,"tags":838,"targetMarket":228,"tier":229},[839,226,840,841,842,843,282,224],"llm","trl","unsloth","huggingface-jobs","gguf",{"commitSha":284},{"parentExtensionId":262,"repoId":289},[226,843,842,839,224,282,840,841],{"evaluatedAt":848,"extractAt":293,"updatedAt":848},1778691314030,{"_creationTime":850,"_id":851,"community":852,"display":853,"identity":856,"providers":857,"relations":867,"tags":868,"workflow":869},1778690773482.4897,"k1758gawxcg89ba8b2srtfawhd86nmcs",{"reviewCount":8},{"description":715,"installMethods":854,"name":855,"sourceUrl":14},{"claudeCode":12},"Transformers.js",{"basePath":409,"githubOwner":256,"githubRepo":257,"locale":18,"slug":411,"type":259},{"evaluate":858,"extract":866},{"promptVersionExtension":215,"promptVersionScoring":216,"score":794,"tags":859,"targetMarket":228,"tier":229},[224,860,861,223,862,863,864,865],"javascript","typescript","computer-vision","audio","webgpu","wasm",{"commitSha":284,"license":249},{"parentExtensionId":262,"repoId":289},[863,862,860,224,223,861,865,864],{"evaluatedAt":870,"extractAt":293,"updatedAt":870},1778691465826,{"_creationTime":872,"_id":873,"community":874,"display":875,"identity":879,"providers":882,"relations":887,"tags":888,"workflow":889},1778691799740.4983,"k1797kqt7c6p7gytn30rckwzvh86nz20",{"reviewCount":8},{"description":876,"installMethods":877,"name":878,"sourceUrl":758},"This skill should be used when working with pre-trained transformer models for natural language processing, computer vision, audio, or multimodal tasks. Use for text generation, classification, question answering, translation, summarization, image classification, object detection, speech recognition, and fine-tuning models on custom datasets.",{"claudeCode":756},"Transformers",{"basePath":880,"githubOwner":761,"githubRepo":762,"locale":18,"slug":881,"type":259},"scientific-skills/transformers","transformers",{"evaluate":883,"extract":886},{"promptVersionExtension":215,"promptVersionScoring":216,"score":219,"tags":884,"targetMarket":228,"tier":229},[223,862,863,885,224,225,881,256],"multimodal",{"commitSha":284,"license":249},{"repoId":772},[863,862,225,256,224,885,223,881],{"evaluatedAt":890,"extractAt":776,"updatedAt":890},1778694649795]