「train」の検索結果

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.

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。

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。

Train neural operators (FNO, DeepONet) to learn solution maps for parametric PDE families. Once trained, solve new PDE instances in milliseconds. Use when you need to solve many instances of the same PDE with different parameters/ICs/BCs.

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。

Visualize training metrics, debug models with histograms, compare experiments, visualize model graphs, and profile performance with TensorBoard - Google's ML visualization toolkit

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。

Train Physics-Informed Neural Networks (PINNs) using DeepXDE. Solve forward and inverse PDE problems by embedding physics equations into the neural network loss function. Supports 1D/2D/3D, time-dependent, and parametric PDEs.

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。

This skill should be used at the start of any computationally intensive scientific task to detect and report available system resources (CPU cores, GPUs, memory, disk space). It creates a JSON file with resource information and strategic recommendations that inform computational approach decisions such as whether to use parallel processing (joblib, multiprocessing), out-of-core computing (Dask, Zarr), GPU acceleration (PyTorch, JAX), or memory-efficient strategies. Use this skill before running analyses, training models, processing large datasets, or any task where resource constraints matter.

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。

Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with tool calls, or when you need 5× faster inference than vLLM with prefix sharing. Powers 300,000+ GPUs at xAI, AMD, NVIDIA, and LinkedIn.

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。

Fast tokenizers optimized for research and production. Rust-based implementation tokenizes 1GB in <20 seconds. Supports BPE, WordPiece, and Unigram algorithms. Train custom vocabularies, track alignments, handle padding/truncation. Integrates seamlessly with transformers. Use when you need high-performance tokenization or custom tokenizer training.

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。

Control LLM output with regex and grammars, guarantee valid JSON/XML/code generation, enforce structured formats, and build multi-step workflows with Guidance - Microsoft Research's constrained generation framework

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。

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

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。

End-to-end shotgun metagenomics workflow from FASTQ to taxonomic and functional profiles, orchestrating controls/host depletion, Kraken2+Bracken classification, MetaPhlAn marker profiling, and HUMAnN functional profiling. Covers the controls-first ordering, why Kraken2 read counts are not abundances and MetaPhlAn cell fractions do not equal Bracken read fractions, and the consistent-pipeline framing. Use when profiling shotgun metagenomic samples end to end, or chaining classification, abundance, and function. For resistome see metagenomics/amr-detection; for strains see metagenomics/strain-tracking; for assembly see genome-assembly/metagenome-assembly.

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。

Kerberos 委派攻击（非约束/约束/RBCD）。当 BloodHound 发现委派配置、或已获取有 SPN 的服务账号/机器账号控制权时使用。通过 S4U 协议滥用可实现跨服务模拟任意用户，常用于域内权限提升和横向移动。

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。

CadQuery (Python 製 3D CAD スクリプティングライブラリ) リファレンス。 Workplane API (lineTo / extrude / revolve / sweep / loft / fillet / chamfer / shell)、 Sketch API、Assemblies (Constraint solver, 9 種制約)、Free Function API、 Selectors、Class hierarchy (Shape / Vertex / Edge / Wire / Face / Shell / Solid)、 Import / Export (STEP / DXF / STL / glTF / SVG)、Visualization (VTK, Jupyter)、CQGI。

Android Jetpack Compose (Kotlin) の UI 基盤リファレンス。androidx, Material3 土台。 Composable, Modifier, CompositionLocal, remember, mutableStateOf, State, LaunchedEffect。 Column, Row, Box, LazyColumn, Pager, ConstraintLayout, WindowSizeClass, ペインスキャフォールド。 BasicText, BasicTextField, TextStyle, IME, clickable, ジェスチャ, フォーカス, セマンティクス。

UIKit iOS/iPadOS UI フレームワークリファレンス。 UIViewController, UINavigationController, UITabBarController, UISplitViewController, UIView, UILabel, UIButton, UITextField, UIScrollView, UIStackView, UITableView, UICollectionView, diffable data source, NSDiffableDataSourceSnapshot, UICollectionViewCompositionalLayout, Auto Layout, NSLayoutConstraint, NSLayoutAnchor, UILayoutGuide, intrinsicContentSize。

Android アプリ開発 (Kotlin) のバックグラウンド・非同期処理リファレンス。 Jetpack androidx WorkManager: Worker, CoroutineWorker, WorkRequest, Constraints, WorkInfo, WorkContinuation, expedited work, テスト。 Service: フォアグラウンドサービス, バウンドサービス, AIDL, Messenger。 AlarmManager, PendingIntent, BroadcastReceiver, JobScheduler, Doze。 Kotlin コルーチン, Flow, StateFlow, SharedFlow, Channel。

The live fieldwork layer for psycho-logic — go look at the actual target market NOW instead of reasoning from your training-data stereotype of it. Use when a psycho-logic move needs grounding in a specific place: the user asks 'what do competitors in [region/country] actually do', 'what's the real price band / norm here', 'how do local leaders position themselves', 'research this market before we localise', 'what signals trust to buyers in X', or is about to launch/adapt in a market you haven't studied. Runs structured web research in the local language — local competitors and their signals, price bands, category conventions, review/trust sources, local terminology, taboos and regulations — and synthesises a compact perception profile the four moves consume. Inspired by Rory Sutherland's *Alchemy*. Routed to from psycho-logic; pair with cultural-context (the lens that tells you what to look for). 日本語・中文の相談（「現地の実態を調べて」「调研当地市场」）にも使う。

Queries gnomAD v4 (807k samples), v3, v2.1.1, and constraint metrics with grpmax FAF95, bottleneck-group exclusion, LOEUF interpretation, SV/CNV/mtDNA catalogs, and Whiffin max-credible-AF framework. Use when filtering rare variants, applying ACMG BS1/BA1, ranking genes by LoF intolerance, or selecting between v2 (GRCh37 + chrX/Y constraint) and v4 (GRCh38 + 807k samples).

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。

Design scalable, normalized database schemas with proper relationships, indexes, constraints, and migration strategies for relational and NoSQL databases. Use when designing database schemas, planning table relationships and foreign keys, creating indexes for query optimization, defining constraints and validations, designing data models for scalability, planning database migrations, choosing between SQL and NoSQL, implementing sharding strategies, optimizing query performance, or establishing data integrity rules.

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。

Strip a problem to its irreducible fundamental truths and rebuild the solution from the ground up — free from analogy, convention, and inherited assumptions. Load when the user feels constrained by how something has always been done, when existing solutions feel expensive or inefficient for no good reason, when the user asks to think from first principles, challenge the fundamentals, or rebuild this from scratch. Also triggers on "why does it have to work this way", "what are the actual constraints here", "ignore what everyone else does", or when deep-thinking diagnoses a convention-break frame. Based on Aristotle's first principles method, popularised by Musk and Feynman. Produces genuinely novel solutions by eliminating convention.

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 the smallest, highest-signal context package for an AI coding task — goal, constraints, repo facts, boundaries, and a verification plan. Load when prompts are underspecified, the agent is missing key files or decisions, the user says "use the right context", "here's the repo", or when work is drifting due to missing constraints. Also triggers on "context engineering", "gather context", "what do you need from me", "before you start". Not for cross-session continuity (use memory-startup/memory-recall).

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。

Queries gnomAD v4 (807k samples), v3, v2.1.1, and constraint metrics with grpmax FAF95, bottleneck-group exclusion, LOEUF interpretation, SV/CNV/mtDNA catalogs, and Whiffin max-credible-AF framework. Use when filtering rare variants, applying ACMG BS1/BA1, ranking genes by LoF intolerance, or selecting between v2 (GRCh37 + chrX/Y constraint) and v4 (GRCh38 + 807k samples).

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 this skill to design an annual driver refresher training program. Covers what to cover, frequency, documentation, and ROI.

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 this skill to develop the capability to investigate crashes internally. Covers what to look for, when to escalate to outside experts, and documentation standards.

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。