SageMaker AI
SageMaker fine-tuning end to end, plus HyperPod cluster debuggers for training failures.
Install
OfficialClaude Code: /plugin marketplace add anthropics/claude-plugins-official → /plugin install sagemaker-ai@claude-plugins-official
SageMaker'da model ince ayarı yapılırken ya da HyperPod eğitim kümesindeki arıza incelenirken
- Author
- Amazon Web Services
- License
- Apache-2.0
Two distinct jobs share one plugin. The first is a guided customisation path: define the use case, validate and convert the dataset, choose between SFT, DPO and RLVR, run the training job, evaluate with benchmarks or a judge model, then deploy to SageMaker or Bedrock. The second is failure diagnosis on HyperPod clusters, and it is the more specific half — NCCL hangs and AllReduce timeouts, EFA errors, per-node GPU and disk pressure, straggler detection, and Slurm nodes stuck in drain with miscounted GRES. Remote commands reach the nodes over SSM rather than SSH.
Similar skills
GitHub Actions Templates
SkillWorking workflow files instead of a blank .github directory.
New Relic
SkillFour New Relic skills — APM, cloud cost, Kubernetes and NRQL — over the hosted MCP server.
Kubernetes Operations
SkillShips runnable Kubernetes templates and a chart validator, not only prose.
Comments(0)
Sign in to comment