登录

skypilot

skypilot-orgApache-2.0活跃维护

把分散在多家云上的 GPU 资源统一调度成一台「AI 超算」，自动选最便宜的可用算力跑训练、微调和推理任务，省去手动比价与跨云管理。

The AI Compute Platform for frontier teams. SkyPilot turns fragmented AI compute into one AI supercomputer, so frontier AI teams build custom intelligence faster.

v0.14.2发布于 2026-10-28 18:26 UTC10,816star +10 / 7 天1,337fork同步于 2026-11-10 17:29 UTC
多云调度GPU 算力成本优化分布式训练任务队列
对比 GitHub 官网cloud-computingcloud-managementcost-optimizationdeep-learningdistributed-training

怎么上手

由 README 提炼，执行前请确认
原文 
uv pip install "skypilot[kubernetes,aws,gcp,azure,oci,nebius,lambda,runpod,fluidstack,paperspace,cudo,ibm,scp,seeweb,shadeform,verda]"
硬件要求按 README 说明

README 未说明具体显存或 GPU 型号要求；示例使用 8x NVIDIA A100，并支持 GPU、TPU、CPU 工作负载，实际资源取决于所选云或集群。

下载：源码 zip·GitHub Releases（安装包）本站不镜像安装包

项目说明

由 README 自动翻译提炼
没人工点评？写下你的经验

统一管理多云与 K8s 上的 AI 算力，用一个 YAML 就能把任务跑到任意基础设施上

把分散在多家云上的 GPU 资源统一调度成一台「AI 超算」，自动选最便宜的可用算力跑训练、微调和推理任务，省去手动比价与跨云管理。

核心能力

  • 用 YAML 或 Python API 声明资源、数据和命令，同一任务可跨云迁移
  • 自动寻找最便宜可用的 GPU/TPU/CPU 并做故障转移
  • 支持 Kubernetes、Slurm 及 20+ 云厂商的统一控制面
  • 提供 gang scheduling、多集群调度与空闲资源自动回收
  • 支持 SSH 进 pod、同步代码、连接 IDE 的本地开发体验

容易踩的坑

安装时需按所用云厂商选择对应的 extras，否则无法对接该云；示例需要目标云上的 GPU 实例配额；任务在你自己云账号的 VPC 和集群内启动，需先完成云凭证配置。

内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README

最近发版

5 个版本v0.14.2
全部 release
  1. v0.14.2Release v0.14.2最新正式版2026-10-28 · 2026-10-28 18:26 UTC

    SkyPilot v0.14.2

    Get it now with:

    uv pip install "skypilot>=0.14.2"

    What's Changed

    Full Changelog: https://github.com/skypilot-org/skypilot/compare/v0.14.1...v0.14.2

  2. v0.14.1Release v0.14.12026-10-19 · 2026-10-19 21:48 UTC
    查看更新说明

    SkyPilot v0.14.1

    Get it now with:

    uv pip install "skypilot>=0.14.1"

    What's Changed

    Full Changelog: https://github.com/skypilot-org/skypilot/compare/v0.14.0rc1...v0.14.1

  3. v0.14.0Release v0.14.02026-10-05 · 2026-10-05 23:56 UTC
    查看更新说明

    SkyPilot v0.14.0: Dynamic Job Groups, Multi-User Slurm, AWS Neuron, Read-Only Workspaces, and More

    SkyPilot v0.14.0 introduces dynamic job groups, letting a running job group launch more managed jobs that are tracked and cancelled with it, and makes Slurm a multi-user backend: jobs are submitted as the authenticated user, with per-user QoS/account settings and stop/start for container clusters. This release also adds AWS Trainium/Inferentia (Neuron) on Kubernetes, read-only visibility for private workspaces, and a conda-free default image.

    Get it now with:

    uv pip install "skypilot[all]==0.14.0"

    Or, upgrade your team SkyPilot API server:

    NAMESPACE=skypilot
RELEASE_NAME=skypilot
VERSION=0.14.0

helm repo update skypilot
helm upgrade -n $NAMESPACE $RELEASE_NAME skypilot/skypilot \
  --set apiService.image=berkeleyskypilot/skypilot:$VERSION \
  --version $VERSION --devel --reuse-values

    Deprecations & Breaking Changes:

    • Conda is no longer installed by default (#10132). provision.install_conda now defaults to false, and conda has been removed from the default Kubernetes images, which instead ship system Python plus a default user venv (~/sky-user-env). Set provision.install_conda: true to opt back in; the opt-in install now uses a current Miniforge with conda-forge as the only channel.
    • Job Groups enforce in-group networking by default (#10297, #10330). Groups whose tasks cannot be co-located now fail at submission instead of being silently spread across clusters, and networking setup failures fail the job. Set inter_connection: false on the group to allow cross-cluster placement.
    • Tighter API authorization: request listing and tracking endpoints are scoped to the calling user (#10274, #10278), sky debug-dump is admin-only (#10277), download paths are authorized against the user's own roots (#10683), principals with an unrecognized role are denied (#10501), and non-admins can no longer grant the admin role via service accounts (#9932, #10482).
    • rbac.restrict_config_to_admins now defaults to true (#10042): users with the user role can no longer read the API server config. Set it to false to restore the previous behavior.
    • Kubernetes: use disk_size instead of the ephemeral_storage field (#9273).
    • Async request submission no longer prints the API stream link (#10081).

    Highlights

    [New] Dynamic Job Groups

    A task in a job group can now launch further managed jobs with sky jobs launch or sky.jobs.launch, and they attach to the group automatically as dynamic tasks (#10698, #10701, #10702, #10725, #10734, #10736, #10732). Dynamic tasks are listed under the group in sky jobs queue and the dashboard, numbered after the group's own tasks, addressed like any other task (`sky jobs logs

  4. v0.14.0rc1Release v0.14.0rc12026-09-22 · 2026-09-22 23:35 UTC
    查看更新说明

    SkyPilot v0.14.0rc1

    Get it now with:

    uv pip install "skypilot>=0.14.0rc1"

    What's Changed

  5. v0.13.2rc1Release v0.13.2rc1pre-release2026-09-18 · 2026-09-18 18:42 UTC
    查看更新说明

    SkyPilot v0.13.2rc1

    Get it now with:

    uv pip install "skypilot>=0.13.2rc1"

    What's Changed

用户点评

登录用户的真实使用记录，可按版本和硬件筛选
共 0 条

    还没有人点评。第一个说出真实体验的人最有价值。

    README 原文

    未翻译
    在 GitHub 打开

    SkyPilot is a system to run, manage, and scale AI workloads on any AI infrastructure.

    SkyPilot gives AI teams a simple interface to run jobs on any infra. Infra teams get a unified control plane to manage any AI compute — with advanced scheduling, scaling, and orchestration.

    :fire: News :fire:

    • [Oct 2026] SkyPilot's scheduler bumps NVL72 training throughput by 14%: blog
    • [Oct 2026] Post-training a 4B model with RL to fly the Tesla Roadster 21% better than GPT-6 Astra: blog
    • [Sep 2026] RL is Everything, Everywhere, All at Once: blog
    • [Sep 2026] SkyPilot Agent Sessions: Mission Control for Coding Agents, on Your Infrastructure: blog
    • [Jul 2026] Serving Kimi K3 on your own GPUs with SkyPilot: blog
    • [Jul 2026] Run AI workloads on any cloud, store on Hugging Face, no egress fees to worry about: Hugging Face blog
    • [Jun 2026] SkyPilot Endpoints: production-ready inference on every cluster you own: blog
    • [Jun 2026] Announcing SkyPilot Sandboxes: run untrusted, LLM-generated code on the Kubernetes clusters you already own. Learn more, join early access
    • [May 2026] How Multiverse doubled their GPU utilization with SkyPilot: case study
    • [Apr 2026] Introducing GPU Compass: One dashboard to browse, compare pricing, and launch across every GPU cloud. Try it at gpus.skypilot.co.
    • [Apr 2026] Research-Driven Agents: Agents read arxiv papers before coding, landed 5 llama.cpp kernel fusions and +15% faster flash attention in ~3 hours for ~$29: blog, HackerNews
    • [Mar 2026] Scaling Karpathy's Autoresearch: Autoresearch runs 1 experiment at a time. We gave it 16 GPUs and let it run in parallel: blog, HackerNews
    • [Mar 2026] How H Company Unlocked Online RL and Unified their AI Platform: case study

    Overview

    SkyPilot is easy to use for AI users:

    • Quickly spin up compute on your own infra
    • Environment and job as code — simple and portable
    • Easy job management: queue, run, and auto-recover many jobs

    SkyPilot makes Kubernetes easy for AI & Infra teams:

    • Slurm-like ease of use, cloud-native robustness
    • Local dev experience on K8s: SSH into pods, sync code, or connect IDE
    • Turbocharge your clusters: gang scheduling, multi-cluster, and scaling

    SkyPilot unifies multiple clusters, clouds, and hardware:

    • One interface to use reserved GPUs, Kubernetes clusters, Slurm clusters, or 20+ clouds
    • Flexible provisioning of GPUs, TPUs, CPUs, with smart failover
    • Team deployment and resource sharing

    SkyPilot maximizes GPU fleet utilization:

    • Autostop: automatic cleanup of idle resources
    • Binpacking: workload binpacking on shared clusters
    • Intelligent scheduler: automatically schedule on the most available infra

    SkyPilot supports your existing GPU, TPU, and CPU workloads, with no code changes.

    Install with uv (also supported: pip, nightly, from source)

    # Choose your clouds:
uv pip install "skypilot[kubernetes,aws,gcp,azure,oci,nebius,lambda,runpod,fluidstack,paperspace,cudo,ibm,scp,seeweb,shadeform,verda]"

    To use SkyPilot directly with your agent (Claude Code, Codex, etc.), install the SkyPilot Skill. Tell your agent:

    Fetch and follow https://github.com/skypilot-org/skypilot/blob/HEAD/agent/INSTALL.md to install the skypilot skill

    Current supported infra: Kubernetes, Slurm, AWS, GCP, Azure, OCI, CoreWeave, Nebius, Lambda Cloud, RunPod, Fluidstack, Cudo, Digital Ocean, Paperspace, Cloudflare, Samsung, IBM, Vast.ai, VMware vSphere, Seeweb, Prime Intellect, Shadeform, Verda Cloud, VastData, Crusoe.

    Getting started

    Install SkyPilot in 1 minute. Then, launch your first cluster in 2 minutes in Quickstart.

    SkyPilot is BYOC: Everything is launched within your cloud accounts, VPCs, and clusters.

    Benefits of SkyPilot on Kubernetes

    SkyPilot makes Kubernetes AI-native.

    It turbocharges your existing Kubernetes clusters by accelerating AI/ML velocity:

    • AI-friendly interface to launch jobs and deployments
    • Much simplified interactive dev for K8s (SSH / sync code / connect IDE to pods)

    ...and optimizing GPU scheduling, utilization, and scaling:

    • Advanced scheduling: Gang scheduling, multi-node jobs, and queueing
    • Multi-cluster support: Bring all your clusters under one control plane
    • Multi-cloud support: One consistent interface to manage many providers

    See SkyPilot vs Vanilla Kubernetes and this blog post for more details.

    SkyPilot in 1 minute

    A SkyPilot task specifies: resource requirements, data to be synced, setup commands, and the task commands.

    Once written in this unified interface (YAML or Python API), the task can be launched on any available infra (Kubernetes, Slurm, cloud, etc.). This avoids vendor lock-in, and allows easily moving jobs to a different provider.

    Paste the following into a file my_task.yaml:

    resources:
  accelerators: A100:8  # 8x NVIDIA A100 GPU

num_nodes: 1  # Number of VMs to launch

# Working directory (optional) containing the project codebase.
# Its contents are synced to ~/sky_workdir/ on the cluster.
workdir: ~/torch_examples

# Commands to be run before executing the job.
# Typical use: pip install -r requirements.txt, git clone, etc.
setup: |
  cd mnist
  pip install -r requirements.txt

# Commands to run as a job.
# Typical use: launch the main program.
run: |
  cd mnist
  python main.py --epochs 1

    Prepare the workdir by cloning:

    git clone https://github.com/pytorch/examples.git ~/torch_examples

    Launch with sky launch (note: access to GPU instances is needed for this example):

    sky launch my_task.yaml

    SkyPilot then performs the heavy-lifting for you, including:

    1. Find the cheapest & available infra across your clusters or clouds
    2. Provision the GPUs (pods or VMs), with auto-failover if the infra returned capacity errors
    3. Sync your local workdir to the provisioned cluster
    4. Auto-install dependencies by running the task's setup commands
    5. Run the task's run commands, and stream logs

    See Quickstart to get started with SkyPilot.

    Runnable examples

    See SkyPilot examples that cover: development, training, serving, LLM models, AI apps, and common frameworks.

    Latest featured examples:

    TaskExamples
    TrainingVerl, Finetune Llama 4, TorchTitan, PyTorch, DeepSpeed, NeMo, Ray, Unsloth, Jax/TPU, OpenRLHF
    ServingvLLM, SGLang, Ollama
    ModelsDeepSeek-R1, Llama 4, Llama 3, CodeLlama, Qwen, Kimi-K2, Kimi-K2-Thinking, Mixtral
    AI appsRAG, vector databases (ChromaDB, CLIP)
    Common frameworksAirflow, Jupyter, marimo

    Source files can be found in llm/ and examples/.

    Learn more

    To learn more, see SkyPilot Overview, SkyPilot docs, and SkyPilot blog.

    SkyPilot adopters: Testimonials and Case Studies

    Follow updates:

    Questions and feedback

    We are excited to hear your feedback:

    For general discussions, join us on the SkyPilot Slack.

    Contributing

    We welcome all contributions to the project! See CONTRIBUTING for how to get involved.

    相关项目

    同分类里 star 最高的项目
    对比全部 →
    ollama
    ollama·Go

    本地一键跑起 Kimi、DeepSeek、Qwen、Gemma 等大模型，省去环境配置和依赖折腾，适合快速做推理和原型验证。

    本地推理+3大模型推理服务入门
    186548+35819kv0.40.152026-11-10 01:49 UTC
    llama.cpp
    ggml-org·C++

    用 C/C++ 实现的大模型推理引擎，能在普通 CPU 和消费级显卡上跑 LLaMA 等模型，支持量化以降低显存占用。

    推理引擎+4大模型推理服务入门
    133932+33525kb122332026-11-10 17:13 UTC
    vllm
    vllm-project·Python

    面向大语言模型的高吞吐、低显存推理与服务引擎，通过 PagedAttention 等优化显著提升并发吞吐并降低显存占用，适合部署在线推理服务。

    推理引擎+4大模型推理服务进阶
    96680+18725kv0.33.0rc32026-11-10 17:01 UTC
    jan
    janhq·Rust

    在本地电脑上离线运行大模型的桌面客户端，是 ChatGPT 的开源替代品，无需联网即可对话，数据不出本机。

    本地推理+4大模型推理服务
    45655+1133.2kv0.8.82026-11-10 17:02 UTC