# tencentmusic/cube-studio

cube studio开源云原生一站式机器学习/深度学习/大模型AI平台，mlops算法链路全流程，算力租赁平台，notebook在线开发，拖拉拽任务流pipeline编排，多机多卡分布式训练，超参搜索，推理服务VGPU虚拟化，边缘计算，标注平台自动化标注，deepseek等大模型sft微调/奖励模型/强化学习训练，vllm/ollama/mindie大模型多机推理，私有知识库，AI模型市场，支持国产cpu/gpu/npu 昇腾生态，支持RDMA，支持pytorch/tf/mxnet/deepspeed/paddle/colossalai/horovod/ray/volcano等分布式

Repository: https://github.com/tencentmusic/cube-studio
Canonical: https://ross.abutalabs.com/products/cube-studio
License: NOASSERTION
License Family: other
Topics: kubernetes, inference, mlops, workflow, ai, pytorch, spark, argo, kubeflow, automl, aihub, gpt, llmops, notebook, pipeline, vgpu, deepseek
Last push: 2026-07-11T06:55:31+00:00

## Health v2 (maintenance only)
Score: 86/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 92, release rhythm 71, longevity 100
- inputs: {"age_days": 1842, "days_push": 53, "days_rel": 116, "gap_med": 89, "n_releases_24m": 8}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5074, forks 881 (observed 2026-08-28T04:09:09.414073+00:00)

## What it is
Cube Studio is an open-source, cloud-native, one-stop machine learning platform covering the full MLOps lifecycle: online notebooks, drag-and-drop pipeline orchestration, distributed multi-node training, hyperparameter search, LLM fine-tuning (SFT, reward models, RLHF), and multi-node LLM inference with vLLM/Ollama. It also provides vGPU-virtualized inference serving, edge computing, automated data labeling, a private knowledge base (RAG), an AI model marketplace, and support for domestic CPUs/GPUs/NPUs (Ascend) with RDMA networking.

## Use cases
- run an end-to-end mlops platform on kubernetes
- orchestrate machine learning pipelines with drag-and-drop
- fine-tune large language models like deepseek with sft or reinforcement learning
- serve llm inference across multiple nodes with vllm or ollama
- rent out gpu compute with vgpu virtualization
- search hyperparameters for distributed training jobs
- build a private knowledge base chatbot over company documents
- automate data labeling for training datasets

## When to choose
- you need a self-hosted, all-in-one mlops platform on kubernetes covering training, tuning, labeling, and inference
- you must support heterogeneous accelerators including ascend npus, domestic gpus, and rdma clusters
- you want llm fine-tuning and multi-node llm serving integrated with workflow orchestration

## When to avoid
- you only need a lightweight experiment tracker or single-feature tool rather than a full platform
- you cannot operate a kubernetes-based cloud-native deployment
- you require a permissively licensed project - the license is non-standard (NOASSERTION)
- you want actively maintained upstream - this repository was archived in 2026 and moved to data-infra/cube-studio

## Facets
- artifact type: application
- maturity: abandoned
- function: machine-learning, deep-learning, llm-training, llm-inference, rag, workflow-automation, scheduling, container-orchestration, data-visualization, agent-framework
- domain: machine-learning, deep-learning, large-language-models, artificial-intelligence, data-science, cloud-computing, self-hosted
- platform: self-hosted, cloud, python
- tags: mlops, llmops, pipeline-orchestration, vgpu, distributed-training, hyperparameter-tuning, model-serving, notebook, data-labeling, model-marketplace, ascend-npu, rdma, archived-repo, containers, kubernetes, docker, gpu, web-server

## Member repositories
- tencentmusic/cube-studio (main) score 86

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:09.414073+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:02:46.786099+00:00, confidence not recorded.
  - readme: https://github.com/tencentmusic/cube-studio (fetched 2026-08-28T04:09:09.414073+00:00, sha 354d7a8f948b)
- Data as of 2026-08-30T08:39:29.467469+00:00.