What problem does it solve?
Provisioning, customizing, and troubleshooting Domino compute environments is complex and prone to configuration drift, package conflicts, and failed builds; this skill provides clear, actionable guidance to produce reproducible Docker-based environments for workspaces and jobs.
Core Features & Use Cases
- Dockerfile customization: Guidance for adding system packages, pinned Python and R libraries, environment variables, and efficient layer management to reduce build times and improve reproducibility.
- Domino Standard Environments & GPU support: How to start from DSEs, configure CUDA/TensorFlow/PyTorch GPU stacks, and verify GPU availability at runtime.
- Package installation strategies & troubleshooting: When to install in the Dockerfile versus runtime, use of requirement files or pre/post-run scripts, and common failure diagnostics including permission issues, network timeouts, and missing packages.
- Use case: Prepare a reproducible ML training environment with pinned dependencies and GPU drivers, test locally via docker build and docker run, then publish a revision in Domino for consistent job execution.
Quick Start
Create or update a Domino environment by writing Dockerfile instructions to pin packages and set environment variables, then trigger a build in the Domino UI and verify the environment revision boots and runs your test command.