Aquanode DocsVM Templates
Launch a pre-configured VM in seconds, no setup needed.
Every VM machine starts from one of these ready-to-use OS templates. Pick one when you launch a pod, no manual OS install needed.
Not every template is available on every provider, check the Supported by list below before you pick a provider you're set on. If your provider isn't listed, the deploy path falls back to a template that is.
Operating System templates
| Template | Version | GPU / vCPU / RAM / Storage | Supported by |
|---|---|---|---|
| Ubuntu VM | 24 | 1 GPU · 1 vCPU · 4 GB RAM · 20 GB | Akash, Verda, Voltage Park, Hot Aisle |
| Alpine VM | 3.18.4 | 1 GPU · 1 vCPU · 2 GB RAM · 10 GB | Akash |
| Debian VM | 12 | 1 GPU · 1 vCPU · 4 GB RAM · 20 GB | Akash |
| CentOS VM | 7 | 1 GPU · 1 vCPU · 4 GB RAM · 20 GB | Akash |
| Fedora VM | 39 | 1 GPU · 1 vCPU · 4 GB RAM · 20 GB | Akash |
| Arch Linux VM | latest | 1 GPU · 1 vCPU · 4 GB RAM · 20 GB | Akash |
- Ubuntu VM: lightweight, full SSH access, ideal for configuring your own stack from scratch. The widest provider support of the six, so it's the safe default when you're not set on a specific OS.
- Alpine VM: minimal footprint, custom configurations.
- Debian VM: stable, for production environments and enterprise applications.
- CentOS VM: Red Hat-based deployments and workloads requiring CentOS compatibility.
- Fedora VM: cutting-edge development, testing, and package availability.
- Arch Linux VM: advanced users who need fine-grained customization.
Machine Learning templates
| Template | Version | GPU / vCPU / RAM / Storage | Supported by |
|---|---|---|---|
| PyTorch Conda CUDA | 2.1.0-cuda12.1-cudnn8-devel | 1 GPU · 4 vCPU · 24 GB RAM · 100 GB | Akash |
| TensorFlow | 2.14.0-gpu | 1 GPU · 4 vCPU · 24 GB RAM · 75 GB | Akash |
- PyTorch Conda CUDA: PyTorch with Conda and CUDA preconfigured, for GPU-accelerated deep learning development.
- TensorFlow: official TensorFlow image, ready for model development, training, and deployment.
Both ML templates currently only run on Akash.
How these relate to app templates
The templates above are plain OS/framework images, they boot a box and nothing more. There are separately two app templates (ComfyUI, Torch + Jupyter) that install a managed service on top of a box and surface an authenticated app URL once it's running: see ComfyUI and Torch + Jupyter.
Both sets of templates, OS-only and app-managed, appear as starting points on the New pod screen. See Launching a pod for how that flow works.
Next steps
- Go to New pod and pick a starting point
- Choose GPU and resources
- Add an SSH key for access
- Deploy
Best practices
Choose the right template
- Start with Ubuntu for general use, it has the widest provider support
- Use Alpine for minimal resource needs
- Pick a PyTorch or TensorFlow template for ML workloads that want CUDA preconfigured
Security
- Always add your SSH key at launch
- Update packages after first login
- Configure firewall rules as needed
Machines
A Machine is the rented GPU box behind a pod: full root access, direct GPU access, and the flexibility to install anything you need.
Browser Terminal
Open a root shell on a running machine from the console, without an SSH client, and understand which providers support it and why the others don't.