skills.rest
duypham2801
duypham2801COMMUNITY
Explore all skills by this author →
GitHub

run-experiment

Deploy and run machine-learning experiments on local GPUs, SSH, Vast.ai, or Modal.

Updated May 25, 2026
Search Tags:#modal#vast.ai#ml experiments#training automation#gpu deployment#ssh screen#wandb logging
One-click install
npx skills add https://github.com/duypham2801/ThS_LLM --skill run-experiment-duypham2801
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: run-experiment
Source: https://github.com/duypham2801/ThS_LLM/tree/main/.claude/skills/run-experiment
Command: npx skills add https://github.com/duypham2801/ThS_LLM --skill run-experiment-duypham2801

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It eliminates the friction of launching ML training jobs by automatically handling environment detection, GPU pre-checks, code syncing, and remote execution for local servers, Vast.ai, or Modal.

Core Features & Use Cases

  • Automatic environment routing: Detects whether you’re running on local GPU, an SSH remote, Vast.ai, or Modal based on project CLAUDE.md.
  • Pre-flight GPU validation: Checks nvidia-smi (or MPS availability) and avoids assigning GPUs that are likely already in heavy use.
  • Robust deployment workflows: Launches jobs via screen on remote/Vast.ai, or delegates serverless GPU execution entirely to Modal.
  • Optional W&B enablement: If configured in CLAUDE.md, it adds/ensures W&B logging and validates login on the target machine.
  • Optional Vast.ai lifecycle management: Can auto-destroy rented instances after completion to reduce cost exposure.

Quick Start

Ask to run your experiment on a GPU by saying: "Deploy and run experiment with the command-line arguments I provide, using the project’s CLAUDE.md settings for local/SSH/Vast.ai/Modal."

Frequently Asked Questions about run-experiment

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate ML experiment deployment across local and remote GPUs?▼

ML experiment deployment is automated by detecting the environment from CLAUDE.md, checking GPU availability with nvidia-smi, syncing code, and launching training via screen or Modal serverless workflows.

Can I run training jobs on Vast.ai and automatically destroy instances after completion?▼

Vast.ai training jobs can execute via SSH screen sessions, and optional lifecycle management allows automatic destruction of rented instances after completion to reduce cost exposure.

How do I check GPU availability before launching remote training jobs?▼

GPU availability is checked during pre-flight validation by probing nvidia-smi or MPS, avoiding assignment of GPUs that are already in heavy use before launching remote training jobs.

Does this workflow support W&B logging for serverless GPU execution on Modal?▼

W&B logging is supported for Modal serverless GPU execution if configured in CLAUDE.md, which ensures logging is added and login is validated on the target machine.

What is the best way to manage SSH screen sessions for distributed ML training?▼

SSH screen sessions for distributed ML training are managed by applying safe code synchronization rules, launching jobs reliably through screen, and verifying execution steps across remote servers.

Do I need to specify environment settings manually for local and serverless GPU deployment?▼

Manual environment specification is unnecessary because automatic environment routing detects local GPUs, SSH remotes, Vast.ai, or Modal based on project CLAUDE.md settings.

Related Skills

Explore More Software Engineering Skills→
raja21068raja21068
2

run-experiment

Deploy and run machine learning experiments across local GPUs, SSH servers, Vast.ai, and Modal.

Community
Advanced
tbroadleytbroadley

read-inspect-eval

Extract samples, messages, events, and metrics from Inspect AI .eval logs.

Community
Intermediate
satsuki-64satsuki-64
2

run-experiment

Deploy and execute ML experiments on local or remote GPU servers via SSH.

Community
Advanced
jkfeejkfee

vast-gpu

Automate vast.ai GPU provisioning, setup, execution, and teardown for ML tasks.

Community
Advanced