52 skills · sorted by relevance
Organize PyTorch training code with LightningModule and Trainer. Follow examples for validation, logging, model saving, and distributed training across multiple GPUs, with callbacks that add monitoring or stop training when progress stalls.
Reserve remote Android devices on Google's Developer Device Platform and connect through ADB, Android's device control tool. Check session status and expiration, extend usage time, or cancel sessions and release resources.
Adapt existing PyTorch training scripts to run across multiple GPUs or machines with Accelerate. This skill covers device setup, distributed training configuration, and mixed precision, which uses lower-precision numbers to reduce memory use.
Organize PyTorch training code with LightningModule and Trainer. This skill guides training, validation, logging, model checkpoint saving, early stopping when progress stalls, and configuration for running the same model across multiple GPUs.
Official NVIDIA-authored guidance for PhysicsNeMo ShardTensor domain parallelism — integrate domain parallelism into training/inference scripts (new or existing) with DDP or FSDP2, write and register shard patches to enable new layers/ops, and bootstrap multi-GPU correctness tests. Use when working with ShardTensor, scatter_tensor, domain parallelism, sequence/spatial sharding, ring attention, DeviceMesh + DDP/FSDP2 hybrid parallelism, or physicsnemo.domain_parallel. Do NOT use for generic PyTorch DDP/FSDP setup without domain parallelism, picking a PhysicsNeMo model or example (use physicsnemo-discover), or non-distributed training questions.
The English summary is being prepared. Showing the original description.
Implement a generative diffusion model (DDPM or score-based) with noise scheduling, U-Net architecture, training loop, and sampling procedures including DDIM acceleration. Use when building a generative model for image, audio, or molecular synthesis; implementing DDPM from a research paper; adding a custom noise schedule or conditioning mechanism; replacing a GAN-based generator with a diffusion alternative; or prototyping before scaling with production frameworks like diffusers.
The English summary is being prepared. Showing the original description.
Organize PyTorch models and data pipelines with PyTorch Lightning, automate training, save checkpoints, and track experiments. Includes templates and guidance for configuring training across multiple GPUs or TPUs.
Adapt PyTorch training scripts for multiple GPUs or machines with Accelerate. This skill guides code changes, configuration, and launch commands, including mixed precision, which combines numerical formats, and DeepSpeed or FSDP integration for distributed training.
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.
The English summary is being prepared. Showing the original description.
Integrate Apple Pay payments and Wallet passes using PassKit. Use when adding Apple Pay buttons, creating payment requests, handling payment authorization, adding passes to Wallet, configuring merchant capabilities, managing shipping/contact fields, or working with PKPaymentRequest, PKPaymentAuthorizationController, PKPaymentButton, AddPassToWalletButton, PKPass, PKAddPassesViewController, PKPassLibrary, Wallet pass distribution, or Apple Pay checkout flows for physical goods, real-world services, donations, and eligible recurring payments.
The English summary is being prepared. Showing the original description.
Build, upgrade, and operate Aspire 13.5.x C# or TypeScript application hosts with the current CLI, AppHost, ServiceDefaults, integrations, dashboard, testing, MCP, and deployment patterns for distributed apps. USE FOR: Aspire.AppHost.Sdk, Aspire.Hosting.*, DistributedApplication.CreateBuilder, apphost.mts, createBuilder, WithReference, WaitFor, AddProject, AddRedis, AddPostgres, aspire run, aspire init, aspire. DO NOT USE FOR: unrelated stacks; generic tasks that do not need this specific guidance. INVOKES: inspect the repository context, edit targeted files, and run relevant build, test, lint, or validation commands when changes are made.
The English summary is being prepared. Showing the original description.
B2B 大客户销售 (B2B 大客户销售 / 企业级复杂销售 (Enterprise B2B Sales / Complex Selling) — 面向企业客户的高价值、长周期、多干系人复杂销售的职业认知操作系统，从业者(AE/客户经理/SE 售前/销售负责人 VP Sales/CRO)、想入行者、以及服务这行(销售咨询/enablement/RevOps)的视角。覆盖: (a) 第一性张力 — 三大方法论流派之争: **买方主导的诊断式提问 (consultative/discovery-led: SPIN 提问 → 痛点 → 方案, 「像医生先诊断再开方」) ⇄ 卖方主导的教学式挑战 (Challenger: Teach-Tailor-Take control, 用商业洞见 reframe 客户认知, 「像老师挑战客户的既有想法」) ⇄ 资格审查纪律 (MEDDIC/MEDDPICC: 先严格 qualify/disqualify, 「no deal 好过 bad deal」)**; 三派不是非此即彼, 资深人混用, 但底层世界观分歧真实(问 vs 教 vs 筛); 更深层张力 — 「过程/方法论纪律 (process rigor: 预测准确、deal inspection、qualify) ⇄ 关系/信任 (relationship: champion 培养、高管关系、人情)」, 「卖结果/业务价值 (business outcome/ROI、卖给经济决策人) ⇄ 卖产品/功能 (feature selling)」, 「多线程触达买方委员会 (multithreading buying committee) ⇄ 单一 champion 依赖」; (b) 方法论正典 (最标准化、最易蒸出高质量+CLI 化) — MEDDIC/MEDDPICC (Metrics 量化指标/Economic buyer 经济决策人/Decision Criteria 决策标准/Decision Process 决策流程/Identify Pain 痛点/Champion 内部拥护者 + Paper process 签约流程/Competition 竞争), SPIN (Situation/Problem/Implication/Need-payoff 提问序列), Challenger (商业洞见 commercial insight + reframe + 建设性张力 constructive tension + Mobilizer 调动者), Solution Selling (痛点链 pain chain/愿景创建/buying vision), Sandler (前置合约 up-front contract/痛苦漏斗 pain funnel/潜艇七步), Command of the Message (Force Management: 业务价值框架/before-after/decision criteria), Gap Selling (Keenan: 现状 current state → 期望 future state → 差距 the gap/问题中心), Miller Heiman Strategic Selling (蓝表 blue sheet/buying influences: 经济/用户/技术买家 + coach), Value Selling, BANT (老式资格框架, 多被 MEDDIC 取代); (c) 行业结构与角色 — SDR/BDR (开发/约见) → AE/客户经理 (成交) → SE/售前/Solutions Engineer (技术验证/POC/demo) → CSM (客户成功/续约扩张) → Sales Manager/VP Sales/CRO; RevOps (收入运营), Sales Enablement (赋能); land and expand (先落地再扩张)、量化指标 quota/OTE(目标总收入)/配额达成率/pipeline coverage(管道覆盖率 ~3x 业内经验法则)/win rate(赢率)/sales velocity(销售速度)/ramp time(爬坡期)/ACV·ARR·NRR(年合同额/年经常性收入/净收入留存); (d) 销售周期 (the deal/opportunity lifecycle) — prospecting 开发 → discovery 需求挖掘 → qualification 资格审查(MEDDIC) → demo/technical validation/POC 技术验证 → business case/value 业务价值与 ROI → proposal 提案 → negotiation 谈判(procurement 采购/法务/security review) → close 成交 → onboarding/expansion 交付与扩张; 干系人: champion 拥护者 vs coach 线人 vs economic buyer 经济决策人 vs blocker 阻碍者; mutual action plan/close plan(双向行动计划/成交计划); (e) 产出物与节奏 — discovery call(需求电话)、demo、business case、proposal/SOW、MAP(mutual action plan)、forecast(预测: commit/best case/pipeline 类别)、pipeline review/deal review/QBR(季度业务回顾)/account plan(客户计划); (f) 职业/招聘/薪酬 — quota/OTE/comp plan(底薪+提成)/accelerator(加速器)/SPIFF/clawback、ramp、President's Club(顶尖销售俱乐部)、SDR→AE→管理 晋升路径、rep turnover(销售流失率高); (g) 争议/批判 — Challenger 是否过誉(CEB 原始数据被质疑、insight selling 批评)、MEDDIC 沦为填表剧场(box-checking theater, 填字段不真 qualify)、活动指标 vs 质量(spray-and-pray 群发/冷电话存废之争)、AI SDR/自动化淹没收件箱(2024-2025 AI BDR 泡沫 + 邮件送达率崩溃 backlash)、SDR 军团模式是否将死(Winning by Design/signal-based GTM 取代)、配额虚高与销售 burnout、季末打折促单、销售方法论的「宗教化」与培训-认证产业链、最大竞争对手其实是「no decision/维持现状」而非对手; (h) 流派/思想谱系 — 提问诊断派(Rackham SPIN/Bosworth&Eades Solution Selling/Keenan Gap Selling) vs 挑战教学派(Dixon&Adamson Challenger/Force Management Command of the Message) vs 资格纪律派(MEDDIC: Napoli&Dunkel 起源/John McMahon/MEDDICC Andy Whyte) vs 关系战略派(Miller Heiman Strategic/Large Account) vs SaaS 现代收入架构派(Aaron Ross Predictable Revenue/Mark Roberge Sales Acceleration Formula/Jacco van der Kooij Winning by Design bowtie/David Skok 指标) vs 开发-心态派(Jeb Blount Fanatical Prospecting/Sales EQ、Anthony Iannarino、Josh Braun) vs 谈判(Chris Voss Never Split the Difference 战术同理心)。不含: B2C/零售/电商导购销售、纯电话客服/客户支持、纯渠道/分销管理(channel 虽相关但聚焦直销复杂销售)、纯 marketing/需求生成(demand gen 虽上游但本 skill 聚焦销售执行)、纯 CSM 客户成功(虽相关但聚焦 new logo + expansion 的销售动作)。) Master OS — automated mastery of B2B 大客户销售 / 企业级复杂销售 (Enterprise B2B Sales / Complex Selling) — 面向企业客户的高价值、长周期、多干系人复杂销售的职业认知操作系统，从业者(AE/客户经理/SE 售前/销售负责人 VP Sales/CRO)、想入行者、以及服务这行(销售咨询/enablement/RevOps)的视角。覆盖: (a) 第一性张力 — 三大方法论流派之争: **买方主导的诊断式提问 (consultative/discovery-led: SPIN 提问 → 痛点 → 方案, 「像医生先诊断再开方」) ⇄ 卖方主导的教学式挑战 (Challenger: Teach-Tailor-Take control, 用商业洞见 reframe 客户认知, 「像老师挑战客户的既有想法」) ⇄
The English summary is being prepared. Showing the original description.
MEDDPICC-based deal qualification worksheet — guide Deal through structured qualification of any opportunity and produce a filled card + recommended next action. Use when asked to "qualify this deal", "run MEDDPICC on this opportunity", "should we pursue this", or "is this deal real".
The English summary is being prepared. Showing the original description.
API de treinamento distribuído mais simples. 4 linhas para adicionar suporte distribuído a qualquer script PyTorch. API unificada para DeepSpeed/FSDP/Megatron/DDP. Posicionamento automático de device, precisão mista (FP16/BF16/FP8). Config interativo, comando de launch único. Padrão do ecossistema HuggingFace.
The English summary is being prepared. Showing the original description.
Framework PyTorch de alto nível com classe Trainer, treinamento distribuído automático (DDP/FSDP/DeepSpeed), sistema de callbacks e boilerplate mínimo. Escala de laptop para supercomputador com o mesmo código. Use quando quiser loops de treinamento limpos com boas práticas integradas.
The English summary is being prepared. Showing the original description.
High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.
The English summary is being prepared. Showing the original description.
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.
The English summary is being prepared. Showing the original description.
High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.
The English summary is being prepared. Showing the original description.
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.
The English summary is being prepared. Showing the original description.
Helps organize PyTorch model training with LightningModule and Trainer, including validation, logging, model saving, and early stopping. It also guides distributed training across multiple GPUs and settings for reduced-precision computation.
Build and evaluate single-agent reinforcement learning experiments, where an agent learns actions from rewards. Use Stable Baselines3 to choose algorithms, create environments, monitor training, and save models with the normalization data needed for consistent evaluation.
Production-ready reinforcement learning algorithms (PPO, SAC, DQN, TD3, DDPG, A2C) with scikit-learn-like API. Use for standard RL experiments, quick prototyping, and well-documented algorithm implementations. Best for single-agent RL with Gymnasium environments. For high-performance parallel training, multi-agent systems, or custom vectorized environments, use pufferlib instead.
The English summary is being prepared. Showing the original description.
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.
The English summary is being prepared. Showing the original description.
Convert existing PyTorch Lightning training code into an NVFLARE federated job using the Lightning Client API patch, local validation, and job export; use only when the request names federated/NVFLARE conversion or asks multiple sites to train collaboratively while keeping each site's data local, and either names PyTorch Lightning or preliminary source inspection identifies one Lightning owner; do not use for non-federated Lightning work such as DDP, profiling, inference serving, or training-loop changes, nor for plain PyTorch, TensorFlow/Keras, other frameworks, deployment, POC/production lifecycle, or experiment workflows.
The English summary is being prepared. Showing the original description.