DataCrunch - NVidia GPUs for FastAI offers low-cost dedicated GPU servers. We have been a proud supporter of FastAI for three years now!

Spot instances start at $0.25/h for a V100, a great card to get started with. This week we added the RTX6000 Ada 48GB, starting at $0.55/h, excellent for working on larger models.

Setup is very straightforward; you can simply select the FastAI image as disk, it has a preinstalled FastAI environment in JupyterLab.

These are the basic steps:

  1. Sign up
  2. Top up your account
  3. Order a new server, choose spot instance for lower pricing
  4. Create an SSH key
  5. Deploy your instance
  6. When you’re done, click hibernate to resume your work later.

Don’t hesitate to reach out via our chat if you have a question!

This is great info. Is there a free-tier GPU that can be used to try out the platform? Does it have git versioning support?