Quickstart with AMD
Start your container with an external volume so your work remains available after the container exits.
View Docker guideQuick start
From terminal to isolated compute.
Self-service AMD GPU compute with no sales handoff in the way.
First launch steps
01
Create an account
Log in to the terminal UI and create a team.
02
Add credits
Add credits with Stripe or crypto stablecoins (USDT and USDC).
03
Launch an instance
Choose a GPU configuration and start your isolated compute.
After launch
Next steps
Your VM already comes with a recent ROCm setup, and Docker or Podman is ready to go. AMD recommends using their dev containers, which is a lot easier than installing everything by hand, and their docs are solid. If you have any feedback, we’d be happy to pass it along to them.
Automate with dstack
Automate deployments and make better use of your VM capacity with our dstack API integration.
View dstack integrationChatXYZ + Open WebUI
Build a private ChatGPT-style interface with Open WebUI, vLLM, and an SSH tunnel to your GPU VM.
Read blog postOpenCode + vLLM
Connect OpenCode to a self-hosted vLLM server on Hot Aisle with SSH tunneling and AMD MI300X GPUs.
Read blog postPyTorch official guide
Use AMD’s official installation guide to get PyTorch running with ROCm.
View PyTorch guideTinyGrad setup
Follow the TinyGrad project setup instructions for an alternative lightweight stack.
View TinyGrad repositoryProgrammatic access
Build from your own tooling.
The same platform is available through the API, CLI, and cloud-init templates.
API docs
Reference the Hot Aisle API directly from your application or automation.
Read API docsCLI
Use the command line to inspect and manage compute from your terminal.
View CLICloud-init templates
Configure an instance at first boot with repeatable cloud-init templates.
View templatesQuestions