gpu-provisioner

What problem does it solve?

This Skill helps users find and manage cost-effective GPU instances for training large language models, ensuring strict confirmation gates for all spending.

Core Features & Use Cases

• GPU Instance Sourcing: Queries multiple cloud providers (RunPod, Vast.ai, Lambda, Kaggle) for suitable GPU hardware.
• Cost Optimization: Ranks instances based on total expected run cost, VRAM fit, throughput, and reliability, not just hourly rates.
• Spend Control: Implements mandatory user approval for provisioning and alerts for cost overruns.
• Use Case: A researcher needs to train a new LLM and wants to find the most economical way to secure the necessary GPU compute, ensuring they don't overspend.

Quick Start

Use the gpu-provisioner skill to find the cheapest viable GPU instance for training a 7B parameter model with a batch size of 16 and a sequence length of 2048.