# kubeflow/trainer

Distributed AI Model Training and LLM Fine-Tuning on Kubernetes

Repository: https://github.com/kubeflow/trainer
Canonical: https://ross.abutalabs.com/products/trainer
Homepage: https://trainer.kubeflow.org/en/latest/
Language: Go
License: Apache-2.0
License Family: permissive
Topics: ai, distributed, fine-tuning, gpu, huggingface, jax, kubeflow, kubernetes, llm, machine-learning, mlops, python, pytorch, tensorflow, xgboost
Last push: 2026-08-26T19:42:30+00:00

## Health v2 (maintenance only)
Score: 95/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 86, longevity 100
- inputs: {"age_days": 3353, "days_push": 7, "days_rel": 15, "gap_med": 62, "n_releases_24m": 12}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2198, forks 1034 (observed 2026-08-28T04:06:25.352457+00:00)

## What it is
Kubeflow Trainer is a Kubernetes-native platform for distributed AI model training and LLM fine-tuning across frameworks like PyTorch, JAX, TensorFlow, HuggingFace, and XGBoost. It orchestrates multi-node, multi-GPU jobs (including MPI workloads) via TrainJob custom resources with a Python SDK.

## Use cases
- fine-tune LLMs on Kubernetes with multiple GPUs
- run distributed PyTorch training jobs across nodes
- train models with JAX or TensorFlow on a cluster
- orchestrate MPI workloads on Kubernetes
- scale HuggingFace model training with DeepSpeed
- manage ML training jobs as Kubernetes resources

## When to choose
- you already run Kubernetes and need scalable distributed training
- you want LLM fine-tuning with multi-node, multi-GPU orchestration
- you need framework-agnostic training runtimes (PyTorch, JAX, XGBoost, MPI)

## When to avoid
- you train small models on a single machine without a cluster
- you have no Kubernetes infrastructure or ops expertise
- you only need inference/serving rather than training

## Facets
- artifact type: framework
- maturity: active
- function: machine-learning, llm-training, gpu-computing, workflow-automation, deployment
- domain: machine-learning, deep-learning, large-language-models, microservices
- platform: python, go, cloud
- tags: distributed-training, llm-fine-tuning, pytorch, jax, tensorflow, huggingface, mpi, mlops, trainjob, kueue, kubernetes, docker

## Member repositories
- kubeflow/trainer (main) score 95

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:25.352457+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 2026-08-30T02:47:09.126067+00:00, confidence not recorded.
  - readme: https://github.com/kubeflow/trainer (fetched 2026-08-28T04:06:25.352457+00:00, sha f58edb23ae12)
- Data as of 2026-08-30T08:39:29.467469+00:00.