87 件 ・ 関連度順
概要と使いどころ
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 <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.
日本語の概要は準備中です。原文の説明を表示しています。
Battle-tested PyTorch training recipes for all domains — LLMs, vision, diffusion, medical imaging, protein/drug discovery, spatial omics, genomics. Covers training loops, optimizer selection (AdamW, Muon), LR scheduling, mixed precision, debugging, and systematic experimentation. Use when training or fine-tuning neural networks, debugging loss spikes or OOM, choosing architectures, or optimizing GPU throughput.
日本語の概要は準備中です。原文の説明を表示しています。
Turns one-shot external recon into a continuous monitoring program. Covers the scheduled re-scan-and-diff loop (baseline snapshot -> interval sleep -> re-scan -> asset/finding delta -> threshold-gated webhook alert), the scan-to-scan diff engine (new/removed/changed assets by a tracked-attribute table, new/resolved findings by a stable cross-scan fingerprint), adversary CTI / chatter monitoring across six public feeds (ransomwatch, ransomware.live, HackerNews Algolia search, Reddit security-subreddit RSS, GitHub Gist code-search, public Telegram channel scraping) with a source-kind-aware severity engine (leak-site/forum/telegram/paste tiers, CRITICAL through INFO), literal/glob/regex watchlist pattern matching, full-corpus capture with retroactive rescan on new watchlist entries, infrastructure-tracking-over-time discipline (certificate-transparency, passive-DNS, port/service, and typosquat re-enumeration cadence, and what a genuine 'perimeter drift' event looks like in the diff output), a five-state finding-lifecycle state machine (open/triaged/risk_accepted/resolved/false_positive) with per-severity SLA and fingerprint-based cross-scan dedup and auto-resolve/reopen rules, the alert-fatigue trap where a lifecycle-unaware rule re-fires on an already-accepted finding, a durable retry/backoff alert-outbox pattern ('queued is not delivered'), and copy-paste bash-cron plus PowerShell-Scheduled-Task recipes for a re-scan+diff loop with the Slack-compatible webhook payload shape. Passive OSINT and analysis only -- no new active-intrusion technique. Use when setting up ongoing monitoring for a retainer or MSSP engagement, tuning alert thresholds to avoid fatigue, triaging a finding's lifecycle status, investigating adversary chatter about a brand, building a 'what changed on the perimeter since last week' report, or deciding whether a persisting finding should re-alert.
日本語の概要は準備中です。原文の説明を表示しています。
Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA/QLoRA optimization
日本語の概要は準備中です。原文の説明を表示しています。
Expert knowledge for Azure AI Personalizer development including troubleshooting, decision making, security, configuration, and integrations & coding patterns. Use when choosing single vs multi-slot setups, tuning policies/exploration, using apprentice mode, explainability, or local inference, and other Azure AI Personalizer related development tasks. Not for Azure AI Search (use azure-cognitive-search), Azure Machine Learning (use azure-machine-learning), Azure AI Metrics Advisor (use azure-metrics-advisor), Azure AI Anomaly Detector (use azure-anomaly-detector).
日本語の概要は準備中です。原文の説明を表示しています。
Mapping design, query optimization, aggregation patterns, index lifecycle management, and search relevance tuning.
日本語の概要は準備中です。原文の説明を表示しています。
Audit whether a paper's baseline comparisons are COMPLETE, FAIR, and SIGNIFICANT: a required recent SOTA baseline is missing while 'best/SOTA' is claimed (HP-MISSING-BASELINE); a baseline is undertuned / given less compute-tuning-data, run at a mismatched config, or the equal-budget ablation-as-baseline is absent (HP-WEAK-BASELINE); 'outperforms' is asserted over overlapping error bars or with no variance/seeds (HP-SIG-OVERLAP); and a cross-row 'improves over baseline by X%' is arithmetically wrong (HP-DELTA-ERROR, cross-row form only). A versioned per-domain baseline profile + a live leaderboard/recency search are assembled by the EXECUTOR as structured facts; a fresh cross-model reviewer (gpt-5.6-sol xhigh, read-only, fresh thread per dimension) PROPOSES findings, each span-anchored to a ledger claim_id; tools/adjudicate_findings.py DECIDES the verdict. Works at L0 (stated comparisons) and deepens at L2 (configs/result files). A completeness question it cannot settle internally becomes needs_external_check, never a guessed missing baseline. Emits baseline-comparison-audit.findings.json; computes NO verdict. Detect-only. Triggers: "baseline audit", "missing baselines", "is the comparison fair", "weak baseline", "baseline 误报", "SOTA earned?".
日本語の概要は準備中です。原文の説明を表示しています。
Elasticsearch guidance — index design, mappings, query DSL, aggregations, cluster operations, and performance tuning.
日本語の概要は準備中です。原文の説明を表示しています。
OpenAI's model connecting vision and language. Enables zero-shot image classification, image-text matching, and cross-modal retrieval. Trained on 400M image-text pairs. Use for image search, content moderation, or vision-language tasks without fine-tuning. Best for general-purpose image understanding.
日本語の概要は準備中です。原文の説明を表示しています。
OpenAI's model connecting vision and language. Enables zero-shot image classification, image-text matching, and cross-modal retrieval. Trained on 400M image-text pairs. Use for image search, content moderation, or vision-language tasks without fine-tuning. Best for general-purpose image understanding.
日本語の概要は準備中です。原文の説明を表示しています。
Axolotl: YAML LLM fine-tuning (LoRA, DPO, GRPO).
日本語の概要は準備中です。原文の説明を表示しています。
言語モデルに指示への応答や好ましい回答を学習させるため、TRLによる追加学習を案内するスキル。データ準備、学習方式の選択、報酬モデルの学習や評価を扱います。
Unsloth: 2-5x faster LoRA/QLoRA fine-tuning, less VRAM.
日本語の概要は準備中です。原文の説明を表示しています。
Expert guidance for deep security incident and entity investigations in Google SecOps. Use when investigating cases, analyzing entities (hosts, IPs, domains, hashes, users), extracting and searching UDM events, performing asset and user timeline analysis, and detecting lateral movement across enterprise networks. Don't use for detection rule authoring or YARA-L tuning (use secops-detection-engineering), proactive hypothesis-driven hunting (use secops-hunt), initial alert triage (use secops-triage), or basic case status updates (use secops-cases).
日本語の概要は準備中です。原文の説明を表示しています。
Run GPU workloads on Modal — training, fine-tuning, inference, batch processing. Zero-config serverless: no SSH, no Docker, auto scale-to-zero. Use when user says "modal run", "modal training", "modal inference", "deploy to modal", "need a GPU", "run on modal", "serverless GPU", or needs remote GPU compute.
日本語の概要は準備中です。原文の説明を表示しています。
Autonomous design space exploration loop for computer architecture and EDA. Runs a program, analyzes results, tunes parameters, and iterates until objective is met or timeout. Use when user says \"DSE\", \"design space exploration\", \"sweep parameters\", \"optimize\", \"find best config\", or wants iterative parameter tuning.
日本語の概要は準備中です。原文の説明を表示しています。
Autonomous design space exploration loop for computer architecture and EDA. Runs a program, analyzes results, tunes parameters, and iterates until objective is met or timeout. Use when user says "DSE", "design space exploration", "sweep parameters", "optimize", "find best config", or wants iterative parameter tuning.
日本語の概要は準備中です。原文の説明を表示しています。
Operates Amazon MSK Provisioned clusters (Standard and Express brokers). Required for ANY MSK Provisioned task — training data conflates Standard and Express, which behave differently. Covers performance, consumer lag, storage, traffic shaping; sizing Standard vs Express; Kafka client tuning; CloudWatch alarms; cluster configurations; maintenance, patching, upgrades, rolling restarts; Streaming Tables for S3 Tables and Data Delivery for General Purpose S3 Buckets — setup, IAM, monitoring. Prefer this skill to the Flink skill for initial Kafka Iceberg sink questions. Triggers: MSK Provisioned (Express/Standard), Kafka, `kafka.*` or `express.*` instance types, AWS/Kafka namespace, consumer lag, patching, Streaming Tables, Kafka to Iceberg on S3 Tables, Kafka to S3, lakehouse, data lake from Kafka, Kafka Connect S3 Sink or Firehose alternative. DO NOT USE for MSK Connect or Replicator — search documentation instead. Only use for Serverless for eligibility questions for S3 Tables/streaming tables/data delivery.
日本語の概要は準備中です。原文の説明を表示しています。
Audit Algolia usage drivers and propose measurable cost controls without hard-coded plan prices. Use when search spend, request volume, record count, replicas, or optional features need investigation. Trigger with "Algolia cost audit", "reduce Algolia usage", or "Algolia bill".
日本語の概要は準備中です。原文の説明を表示しています。
Battle-tested PyTorch training recipes for all domains — LLMs, vision, diffusion, medical imaging, protein/drug discovery, spatial omics, genomics. Covers training loops, optimizer selection (AdamW, Muon), LR scheduling, mixed precision, debugging, and systematic experimentation. Use when training or fine-tuning neural networks, debugging loss spikes or OOM, choosing architectures, or optimizing GPU throughput.
日本語の概要は準備中です。原文の説明を表示しています。
Build local BLAST databases and run searches using NCBI BLAST+ command-line tools. Use when running >50 queries, building custom databases with -parse_seqids and -taxid, downloading prebuilt NCBI databases via update_blastdb.pl, choosing -task variants (megablast/dc-megablast/blastn/blastn-short), tuning soft/hard masking, scaling threads, or extracting hits with blastdbcmd. Encodes BLAST v5 vs v4 database format, taxonomy filtering, makeblastdb pitfalls.
日本語の概要は準備中です。原文の説明を表示しています。
Deploy, manage, and optimize vector databases for AI applications. Covers Qdrant, Weaviate, pgvector, and Pinecone — collection management, indexing strategies, backup, and performance tuning for production RAG and semantic search workloads.
日本語の概要は準備中です。原文の説明を表示しています。
Use when generating or editing images with Google's Gemini image models (Nano Banana) on Scenario via MCP: text-to-image, natural-language instruction editing, identity locking or style transfer from reference images, multi-image fusion, pulling stills from a video clip, Google Search grounding, thinking level tuning, 512 to 4K output, or choosing between Flash, Pro, and Lite. Keywords: Gemini 3.1 Flash, Gemini 3.0 Pro, Gemini 3.1 Lite, Nano Banana 2, Nano Banana Pro, txt2img, img2img.
日本語の概要は準備中です。原文の説明を表示しています。
Expert knowledge for Azure Horizondb development including troubleshooting, best practices, decision making, architecture & design patterns, limits & quotas, security, configuration, integrations & coding patterns, and deployment. Use when tuning pgvector, azure_ai SQL functions, LangChain vector stores, Apache AGE graphs, or hybrid search, and other Azure Horizondb related development tasks. Not for Azure Cosmos DB (use azure-cosmos-db), Azure SQL Database (use azure-sql-database), Azure Table Storage (use azure-table-storage).
日本語の概要は準備中です。原文の説明を表示しています。