skills.rest
OpenLAIR
OpenLAIROFFICIAL
Explore all skills by this author →
GitHub

aris-run-experiment

Deploys and runs ML training experiments on local, remote, Vast.ai, or Modal GPU resources.

1.1k|116|Updated Feb 26, 2026
Search Tags:#training#GPU#ssh#modal#experiment deployment#Machine learning#Vast.ai
One-click install
npx skills add https://github.com/OpenLAIR/dr-claw --skill aris-run-experiment
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: aris-run-experiment
Source: https://github.com/OpenLAIR/dr-claw/tree/main/skills/aris-run-experiment
Command: npx skills add https://github.com/OpenLAIR/dr-claw --skill aris-run-experiment

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires vastai, modal, wandb.

What problem does it solve?

Launching ML experiments across heterogeneous GPU environments (local machines, SSH servers, Vast.ai rentals, Modal serverless) requires repetitive manual work: checking GPU availability, syncing code, managing screen sessions, and tracking costs. This Skill automates the full deployment workflow so training jobs start correctly on the right hardware.

Core Features & Use Cases

  • Compute Resource Guard: Verifies GPU availability via nvidia-smi, MPS checks, or Modal token verification before launching anything, and stops rather than fabricating results when no compute exists.
  • Multi-Backend Deployment: Detects the target environment from CLAUDE.md (local, remote SSH, Vast.ai, or Modal) and handles code sync via rsync or git, dependency installation, and screen-session or background-process launch.
  • Experiment Lifecycle Management: Optionally adds W&B logging to training scripts, sends Feishu notifications, and auto-destroys Vast.ai instances after collecting results to stop billing.
  • Use Case: A researcher says "run the fine-tuning experiment on the remote server" and the Skill checks free GPUs, rsyncs the Python code, launches training in a detached screen session with CUDA_VISIBLE_DEVICES set, and reports the GPU, log file, and estimated time.

Quick Start

Ask the assistant to run your training script as an experiment, for example: run my train.py experiment on the remote GPU server with 8 A100s.

Frequently Asked Questions about aris-run-experiment

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

FAQPage Schema
How do I run an ML training experiment on a remote GPU server?▼

Add your server details (SSH alias, conda environment, code directory) to the project CLAUDE.md with gpu: remote, then ask to run the experiment. The Skill checks free GPUs with nvidia-smi, syncs code via rsync, and launches training in a detached screen session.

How do I run experiments on Vast.ai rented GPUs?▼

Set gpu: vast in CLAUDE.md after installing the vastai CLI and setting your API key. The Skill provisions a cost-optimized instance, syncs code to /workspace/project/, runs the experiment, downloads results, and auto-destroys the instance to stop billing.

Can I run experiments on Modal serverless GPUs without SSH?▼

Yes. Set gpu: modal in CLAUDE.md after running pip install modal and modal setup. The Skill delegates to a Modal launcher that mounts local code, selects a GPU based on VRAM needs, and returns results via a Volume with no SSH or code sync required.

What happens if no GPU is available when I launch an experiment?▼

The Skill runs a mandatory compute resource guard first using nvidia-smi, MPS checks, or Modal token verification. If no compute is available it stops immediately, reports what is missing, and suggests alternatives rather than fabricating results.

Does the experiment runner support Weights & Biases logging?▼

Yes. When wandb: true is set in CLAUDE.md with a wandb_project name, the Skill adds wandb.init and wandb.log calls to training scripts for metrics like train/loss, eval/loss, and GPU memory, and verifies wandb login on the target machine.

How do I avoid being billed for idle Vast.ai instances?▼

Keep auto_destroy: true in CLAUDE.md, which is the default. After the experiment completes, the Skill downloads results and logs, destroys the instance with vastai destroy instance, and reports the duration and estimated cost.

Related Skills

Explore More Education & Research Skills→
Mang30Mang30

run-experiment

Launches ML training jobs on local, SSH, Vast.ai, or Modal GPUs with logging.

Community
Advanced
wanshuiyinwanshuiyin
15.6k

monitor-experiment

Monitor running ML experiments and collect results from remote training backends.

Community
Intermediate
lightrain-alightrain-a

run-experiment

Deploys ML experiments to local, SSH, Vast.ai, or Modal GPUs with verification.

Community
Advanced
JesusFigueroa25JesusFigueroa25

mlops-engineer

Automate ML infrastructure setup and MLOps workflows across AWS, Azure, and Google Cloud.

Community
Advanced