run-experiment

You can deploy ML experiments to remote GPU servers by providing an experiment description; the Skill handles SSH connection, environment detection, pre-flight GPU checks, code synchronization, and execution using your configured conda environment.

Yes, you can run ML experiments on local GPUs. The Skill supports both local execution and remote SSH deployment, automatically detecting the environment and conducting GPU pre-flight checks to ensure the experiment launches reliably in either context.

Yes, running ML experiments supports optional wandb logging. During remote deployment or local execution, you can enable wandb integration to track experiment metrics and outputs alongside the automated SSH orchestration and GPU pre-flight checks.

Before deploying ML experiments, the Skill performs environment detection and GPU pre-flight checks to verify the target environment. It also synchronizes your code via SSH to prevent deployment failures and ensure the remote GPU is ready for execution.

CLAUDE.md configuration is needed for ML experiment deployment because it provides the necessary server and environment parameters. The Skill relies on this configuration to correctly identify target GPUs, establish SSH connections, and execute code within the proper conda environment.