Help us improve
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
End-to-end Ascend NPU development: migrate CUDA/TensorFlow models to NPU, develop custom operators (AscendC, Triton, CATLASS), deploy vLLM/Verl inference services, profile and optimize performance, manage clusters and containers, and generate documentation and tests across the full AI training and inference pipeline.
npx claudepluginhub ascend-ai-coding/awesome-ascend-skills --plugin npu-smiAutomates GitCode open-source repo merge workflow: commit → push → issue → PR → pipeline → review → /lgtm & /approve → merge. Invoke when user wants to submit code to a GitCode open-source repository.
GitHub Issue root cause analysis skill. Analyze specified GitHub issues to identify possible root causes by examining issue content, code repository, and related resources. Use when user asks to analyze issue root cause, investigate issue, RCA analysis, or troubleshoot GitHub issues. Provides investigation directions with probability estimates and explicitly states when root cause cannot be determined.
Analyze closed GitHub issues to create troubleshooting case studies with root cause analysis and lessons learned. Use when: (1) summarizing resolved issue troubleshooting process, (2) creating issue case studies, (3) documenting problem-solving methodology, (4) extracting lessons from closed issues, (5) building knowledge base from issue history. Triggers: "summarize this issue", "create case study from issue", "issue resolution summary", "troubleshooting case", "postmortem analysis".
AI for Science 昇腾 NPU 总入口 Skill,用于在用户只给出 AI for Science 需求、模型名、TensorFlow/Keras 项目或性能采集诉求时,判断应该进入 Profiling 采集、模型迁移或 TF 框架三条路线,并分流到对应子 skill。
AI for Science 场景下的昇腾 NPU Profiling 采集与性能分析 Skill,用于在华为 Ascend NPU 上使用 torch_npu.profiler 采集 L0、L1、L2 级性能数据,分析训练或推理中的算子耗时、调用栈、内存与瓶颈,并指导后续调优。
Share bugs, ideas, or general feedback.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Agent skills for the FlagOS multi-chip AI inference platform. Includes kernel generation and optimization, model migration, environment verification, vendor onboarding, and more.
Claude Code skill pack for CoreWeave (24 skills)
AI/ML development: LLM architecture, prompt engineering, ML ops, and NLP with production deployment focus
Agent Skills for AI/ML tasks including dataset creation, model training, evaluation, and research paper publishing on Hugging Face Hub
SkyPilot agent skill for launching cloud VMs, Kubernetes pods, and Slurm jobs across 25+ clouds
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
这是一个给昇腾 NPU 开发者使用的 skills 仓库。内容按 Skill 组织,可被 Claude Code、OpenCode、Cursor、Trae、Codex 等 AI 编程工具读取。
Awesome Ascend Skills 收集昇腾 NPU 开发中常用的排障、部署、迁移和分析经验。仓库里主要有四类内容:
npu-smi、hccl-testmindspeed-llm-skillsascend-base、ascend-inference当前目录模型:
skills/skills/<domain>/... 是本地 skill 的唯一正式路径external/ 是外部同步 skills 的独立目录,不参与本地路径规则第一次使用时,不必从完整列表里一个个挑。先看 快速开始,确定自己要装哪个方向,再去 安装指南 执行命令。
Start
├─ 你是第一次使用,或者还不确定该装什么?
│ ├─ Yes → 先安装 `ascend-base`
│ └─ No → 进入下一步
│
├─ 你的主要任务是什么?
│ ├─ 推理 / 模型转换 / 服务部署
│ │ └─ 安装 `ascend-base` + `ascend-inference`
│ ├─ 训练 / 通信 / MindSpeed-LLM
│ │ └─ 安装 `ascend-base` + `ascend-training`
│ ├─ Profiling 采集 / 性能瓶颈分析
│ │ └─ 安装 `ascend-base` + `ascend-profiling`
│ ├─ 算子开发 / Triton 迁移 / op-plugin 接入 / 算子调优
│ │ └─ 安装 `ascend-base` + `ascend-ops`
│ ├─ AI for Science 专项工作
│ │ └─ 安装 `ascend-base` + `ascend-ai-for-science`
│ └─ 我只想要一个非常具体的能力
│ └─ 使用 `-s <skill-name>` 安装单 skill
│
└─ 仍然拿不准?
└─ 先装 `ascend-base`,再按下面安装指南追加对应 bundle
ascend-baseascend-inference、ascend-training、ascend-profiling、ascend-ops使用 npx 安装到支持 Skills 的 AI 编程工具中。新同学先装对应方向的目录即可,不要一上来把全部 skills 都装进去:
# 基础环境包(推荐所有新同学先装)
npx skills add https://github.com/ascend-ai-coding/awesome-ascend-skills/tree/main/skills/base -s '*'
# 推理方向
npx skills add https://github.com/ascend-ai-coding/awesome-ascend-skills/tree/main/skills/inference -s '*'
# 训练方向
npx skills add https://github.com/ascend-ai-coding/awesome-ascend-skills/tree/main/skills/training -s '*'
# Profiling / 性能分析方向
npx skills add https://github.com/ascend-ai-coding/awesome-ascend-skills/tree/main/skills/profiling -s '*'
# 算子开发 / 迁移方向
npx skills add https://github.com/ascend-ai-coding/awesome-ascend-skills/tree/main/skills/ops -s '*'
# AI for Science 方向
npx skills add https://github.com/ascend-ai-coding/awesome-ascend-skills/tree/main/skills/ai-for-science -s '*'
# 安装单个 Skill
npx skills add ascend-ai-coding/awesome-ascend-skills -s npu-smi
# 安装全部 Skills(不建议新同学直接使用)
npx skills add https://github.com/ascend-ai-coding/awesome-ascend-skills -s '*'
支持的 AI 编程工具:Claude Code、OpenCode、Cursor、Trae、Codex 等。
如果你要维护仓库、补充 skill,先看这两个入口:
| 分类入口 | 适合维护什么 |
|---|---|
skills/ | 所有本地 skills 的统一入口,按 base / inference / training / profiling / ops / agent-tools / ai-for-science 分类组织 |
external/ | 外部同步 skills |
维护时注意:
skills/ 下。skills/README.md 继续按功能域分流。skills/<domain>/...,不要再在 root 下加平行目录。| 安装包 | 适合谁 | 包含内容 |
|---|---|---|
ascend-base | 所有新同学 | 基础环境、服务器连接、容器环境、设备检查、硬件诊断、虚拟化与 PyTorch NPU 基础能力 |
ascend-inference | 推理、模型转换、服务部署 | ATC、vLLM-Ascend、vLLM 服务部署、在线压测、ais-bench、量化、Diffusers、Wan 适配 |
ascend-training | 训练、通信、MindSpeed-LLM、MindSpeed-MM、VERL | HCCL、torch 通信测试、MindSpeed-LLM/MM 全流程、VERL quickstart |
ascend-profiling | Profiling 采集、性能分析 | Profiling 分析、训练 Profiling 采集、MFU 分析 |
ascend-ops | 算子开发、迁移、调优 | AscendC、op-plugin、Triton-Ascend 迁移、算子基准测试 |
ascend-ai-for-science | AI for Science 专项用户 | AI for Science 总入口及其子技能 |
ascend-profiling 和 ascend-ops 的区别:
ascend-profiling:你已经有模型/训练任务,重点是采集 Profiling、定位性能瓶颈、分析 hostbound / computing / communication。ascend-ops:你要做的是算子开发、算子迁移、op-plugin 接入、Triton-Ascend 改写或算子级调优。如果方向已经很明确,可以直接装更细的领域技能包:
| 技能包 | 说明 |
|---|---|
mindspeed-llm-skills | MindSpeed-LLM 训练全流程 |
mindspeed-mm-skills | MindSpeed-MM 多模态训练全流程 |
diffusers-ascend-skills | Diffusers 环境、权重准备与推理 |
profiling-analysis | Profiling 分析技能集 |
ai-for-science | AI for Science 技能集 |
hiascend-forum | 昇腾社区论坛抓取与反馈问题分析 |
如果无法使用 npx,可以手动复制所需的 skill 目录。
方式一:项目级安装(推荐)
将所需 skill 复制到项目根目录的 .agents/skills/ 下:
# 克隆仓库
git clone https://github.com/ascend-ai-coding/awesome-ascend-skills.git
# 复制基础环境相关 skills 到项目目录
cp -r awesome-ascend-skills/skills/base/npu-smi your-project/.agents/skills/
cp -r awesome-ascend-skills/skills/base/ascend-docker your-project/.agents/skills/
cp -r awesome-ascend-skills/skills/base/torch_npu your-project/.agents/skills/