run-experiment

To deploy ML experiments on a remote GPU server, the skill handles environment detection, code synchronization, and deployment using configurations defined in a CLAUDE.md file. It automates pre-flight checks to ensure GPU availability before executing your training run.

Yes, you can run ML training on Vast.ai instances automatically. The skill supports vast.ai workflows alongside local and remote SSH targets, managing code synchronization, deployment, and monitoring to streamline orchestration across these flexible compute environments.

Yes, this tool supports optional W&B integration for monitoring ML training runs. It provides built-in tracking capabilities across its supported local, remote, and vast.ai deployment targets, allowing you to oversee experiment metrics and progress during execution.

Before running ML experiments on remote GPUs, you need to configure a CLAUDE.md file with your remote environment details. This configuration provides the necessary deployment targets and connection parameters for the skill to execute pre-flight checks and synchronize code.

The skill handles GPU availability by performing automated pre-flight checks and environment detection before deploying experiments. This safeguard ensures that local or remote GPU resources are actually present and ready, preventing execution errors before code synchronization begins.

Yes, you can orchestrate parallel ML experiments across multiple deployment targets. The skill supports multi-target orchestration, allowing you to manage concurrent local, remote, and vast.ai workflows while maintaining deterministic execution and monitoring.