Estimate GPU requirements and rent cost-optimized vast.ai instances.
npx skills add https://github.com/lightrain-a/medtrace-aris --skill vast-gpu-lightrain-a
Please help me install this Agent Skill. Skill: vast-gpu Source: https://github.com/lightrain-a/medtrace-aris/tree/main/.vendor/aris/skills/vast-gpu Command: npx skills add https://github.com/lightrain-a/medtrace-aris --skill vast-gpu-lightrain-a
This Skill removes the complexity of selecting and provisioning the right GPU hardware by automatically estimating GPU needs from your experiment and renting an appropriate vast.ai instance.
Run an experiment with gpu set to vast so the system provisions a cost-optimized vast.ai GPU instance and sets it up automatically.
High-intent search queries and answers about installing and using this skill.
Automatically estimating GPU requirements involves analyzing experiment scripts to determine min VRAM, disk, and PyTorch compatibility. The system then searches vast.ai broadly across GPU tiers and presents 3 ranked options using estimated total cost and reliability.
Vast.ai GPU rental supports complete lifecycle management by provisioning instances, waiting for readiness, configuring dependencies, and destroying instances post-experiment. This ensures predictable costs and minimal operational overhead for on-demand training runs.
To rent GPUs on vast.ai, you must install the vastai CLI using Python 3.10 or higher and configure SSH key prerequisites. The system maintains a local vast-instances.json state file to track provisioned setups during fine-tuning.
Finding cost-optimized vast.ai instances for multi-GPU training involves estimating total costs across GPU tiers rather than just hourly rates. The system presents three ranked options with reliability guidance, fitting scenarios like ablations and on-demand runs.
Instance provisioning on vast.ai manages SSH key prerequisites, waits for readiness, and executes dependency setup and code sync. This supports running experiments directly on the rented instance with predictable operational overhead.
dogekikiAutomate vast.ai GPU instance provisioning, configuration, and cleanup via vastai CLI.
Shallow-WRent and manage on-demand vast.ai GPU instances from task requirements.
Wenwen555Deploy and run ML experiments across local, remote, Vast.ai, and Modal environments.
AnandaAnugrahHandyantoRun ML workloads on Modal's serverless GPU cloud with Python.
