{"id":51239934,"url":"https://github.com/chrockey/kubeflow","last_synced_at":"2026-06-28T23:31:57.231Z","repository":{"id":350356969,"uuid":"1206462301","full_name":"chrockey/kubeflow","owner":"chrockey","description":null,"archived":false,"fork":false,"pushed_at":"2026-04-10T00:51:18.000Z","size":16,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-04-10T02:42:27.994Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Dockerfile","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/chrockey.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-04-10T00:06:49.000Z","updated_at":"2026-04-10T00:51:25.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/chrockey/kubeflow","commit_stats":null,"previous_names":["chrockey/kubeflow"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/chrockey/kubeflow","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/chrockey%2Fkubeflow","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/chrockey%2Fkubeflow/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/chrockey%2Fkubeflow/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/chrockey%2Fkubeflow/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/chrockey","download_url":"https://codeload.github.com/chrockey/kubeflow/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/chrockey%2Fkubeflow/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34907985,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-28T02:00:05.809Z","response_time":54,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-06-28T23:31:57.093Z","updated_at":"2026-06-28T23:31:57.220Z","avatar_url":"https://github.com/chrockey.png","language":"Dockerfile","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Kakao Cloud + Kubeflow training template\n\nA minimal template for running distributed PyTorch training on\n[Kakao Cloud](https://www.kakaocloud.com/) with\n[Kubeflow Trainer](https://www.kubeflow.org/docs/components/trainer/):\nbuild a CUDA image, push it to the Kakao Container Registry (KCR), and\nsubmit a `TrainJob`.\n\n## Prerequisites\n\n- A Kakao Cloud Kubernetes namespace with the Kubeflow Trainer CRDs installed\n- `kubectl` configured for that namespace\n- `docker` with push access to KCR\n\n## Workflow\n\n### 0. Clone this repo in your Kakao Cloud notebook\n\nOpen a terminal in your Kakao Cloud notebook instance (which already has\n`kubectl` and `docker` configured for your namespace) and clone the repo:\n\n```bash\ngit clone https://github.com/chrockey/kubeflow.git\ncd kubeflow\n```\n\nAll subsequent steps run from this directory.\n\n### 1. Create `.env` at the repo root\n\n```bash\nREGISTRY_URL=your-registry-url         # e.g. postech-a.kr-central-2.kcr.dev\nREGISTRY_NAMESPACE=your-kcr-namespace  # e.g. chunghyun\nREGISTRY_USERNAME=your-kcr-username\nREGISTRY_PASSWORD=your-kcr-password\nWANDB_API_KEY=your-wandb-key   # https://wandb.ai/authorize\n```\n\n`.env` is gitignored and holds per-user credentials and registry info.\n`docker_build.sh` reads everything from it, and step 3 below loads it into a\nKubernetes Secret for the TrainJob.\n\n\u003e [!NOTE]\n\u003e `REGISTRY_NAMESPACE` is your **Kakao Container Registry account**\n\u003e (e.g. `chunghyun`), *not* the Kubernetes namespace of your cluster\n\u003e (e.g. `kbm-g-np-postech-a`). The two are unrelated. The full image tag\n\u003e built by `docker_build.sh` will be\n\u003e `${REGISTRY_URL}/${REGISTRY_NAMESPACE}/kubeflow-train:latest`.\n\nThe image name itself (`kubeflow-train`) is defined in `docker/docker_build.sh`\nand must match the `image:` field in `kubeflow/training-runtime.yaml`. Change\nboth if you want a different name.\n\n### 2. Build and push the image\n\n```bash\n./docker/docker_build.sh latest --push\n```\n\nImage tag is built as `${REGISTRY_URL}/${REGISTRY_NAMESPACE}/kubeflow-train:latest`.\n\n### 3. Load `.env` into a Kubernetes Secret (one-time)\n\nSecrets are namespace-scoped, so in a shared namespace pick a unique name\n(e.g. `train-env-\u003cyour-name\u003e`) to avoid clobbering other users:\n\n```bash\nkubectl create secret generic train-env-\u003cyour-name\u003e \\\n  --from-env-file=.env \\\n  -n kbm-g-np-postech-a\n```\n\nThen update `kubeflow/example-training.yaml` so its `secretKeyRef.name`\nmatches the Secret you just created. The TrainJob will then read\n`WANDB_API_KEY` (and any other variable in `.env`) from your Secret with no\nper-job edits.\n\n### 4. Apply the TrainingRuntime\n\nReplace the `REGISTRY_URL` and `REGISTRY_NAMESPACE` placeholders in the\n`image:` field of `kubeflow/training-runtime.yaml` with your real values from\n`.env`, then:\n\n```bash\nkubectl apply -f kubeflow/training-runtime.yaml\n```\n\n### 5. Submit the TrainJob\n\n```bash\nkubectl apply -f kubeflow/example-training.yaml\nkubectl logs -f -n kbm-g-np-postech-a -l trainer.kubeflow.org/trainjob-name=example-training\n```\n\n`example-training.yaml` is a self-contained 2-GPU MNIST DDP job — it generates\nits own training script inline, so it works as a smoke test with no external\ncode or dataset.\n\n## Files\n\n| File | Purpose |\n|---|---|\n| `docker/Dockerfile` | Minimal CUDA 12.8 + PyTorch + torchvision + wandb image |\n| `docker/docker_build.sh` | Build / test / push to KCR (reads `.env`) |\n| `kubeflow/training-runtime.yaml` | Reusable `TrainingRuntime` with image and parallelism policy |\n| `kubeflow/example-training.yaml` | MNIST DDP smoke-test `TrainJob` |\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fchrockey%2Fkubeflow","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fchrockey%2Fkubeflow","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fchrockey%2Fkubeflow/lists"}