> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/nemo/evaluator/nightly/deployment/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemo/evaluator/_mcp/server. # Deployment Guide ## Deployment Matrix | Environment | Sharding | Live Serve | Regression Gate | Effort | | ----------------------------------------------- | ----------------------------------- | ---------------------- | ------------------------ | ------- | | [Local Deployment](/deployment/local) | Manual via env vars | `nel serve` | `nel compare` | Minimal | | [Docker Compose](/deployment/docker) | Per-container env vars | docker-compose service | Script | Low | | [SLURM](/deployment/slurm) | `nel eval run` with executor config | `nel serve` | sbatch chain | Medium | | [Kubernetes Deployment](/deployment/kubernetes) | Indexed Job | Deployment + Service | CI pipeline | Medium | | [Ray Deployment](/deployment/ray) | `@ray.remote` tasks | N/A | Script | Medium | | [GitLab CI](/deployment/ci-regression) | CI variables per job | N/A | `regression:check` stage | Low | ## Which deployment to choose? ```mermaid flowchart TD START["How big is your eval?"] --> SIZE{"> 1000 problems
or > 4 repeats?"} SIZE -->|No| LOCAL["Local CLI
nel eval run"] SIZE -->|Yes| INFRA{"What infra?"} INFRA --> HPC["HPC cluster"] INFRA --> CLOUD["Cloud / K8s"] INFRA --> RAYC["Ray cluster"] HPC --> SLURM["SLURM
nel eval run config.yaml"] CLOUD --> K8S["K8s Indexed Job"] RAYC --> RAY["Ray launcher"] LOCAL --> CI{"Need regression gate?"} CI -->|Yes| GITLAB["GitLab CI"] CI -->|No| DONE["Done"] ```