Documentation

Create Pod

The Create Pod page allows you to configure and deploy a new compute pod with GPU resources. Follow this guide to understand all configuration options.

Pod Configuration

Basic Settings

Pod Name

  • Default: "Your Pod"
  • Enter a custom name to identify your pod
  • Names help you organize multiple pods

Template Selection

The Template Selection section allows you to choose the environment configuration (template) that will be used to create your pod. A template defines the base Docker image, software stack, and environment settings for your GPU container.

How Template Selection Works

  • Browse from a list of available templates, including official and custom templates
  • Each template specifies a Docker image, version, and may include labels
  • Selecting a template determines the software environment (Python version, CUDA version, pre-installed libraries)

Example: PyTorch Template

Suppose you want to run PyTorch with CUDA support. You might select:

Template: Pytorch (Cuda) - daturaal/pytorch Image: 2.70-py3.12-cuda12.8.0-devel-ubuntu24.04 Labels: ā€¢ Previously used: You have used this template before ā€¢ Most popular: Commonly used and cached on the node

By choosing this template, your pod will be provisioned with PyTorch installed, CUDA 12.8.0 support, and based on Ubuntu 24.04. This is suitable for deep learning tasks requiring GPU acceleration.

šŸ’” Tip

Templates marked as "Most popular" are cached on nodes, resulting in faster deployment times.

Select # of GPUs

This section allows you to specify how many GPUs your pod will use. If the node supports GPU splitting, you can rent a subset of the available GPUs rather than the entire machine.

How it works

  • Nodes with GPU splitting enabled allow you to rent individual GPUs
  • Total rental cost = GPU Price Ɨ Number of GPUs selected
  • This feature enables cost optimization by renting only what you need
  • GPU providers set a minimum GPU count requirement

Resource Allocation (Proportional)

When renting only part of a node's GPUs, resources scale proportionally:

# CPU Allocation cpu = total_cpu * rented_gpu_count / total_gpu_count # Memory Allocation memory = (total_memory_in_gb - 2gb) * rented_gpu_count / total_gpu_count # Storage Allocation available_space = free_space * rented_gpu_count / total_gpu_count volume_limit = int(available_space * 2 / 3) storage_limit = int(available_space / 3)

Minimum GPU Count

GPU providers can configure a minimum number of GPUs that must be rented per pod. For example, a node may have 8 available GPUs with a minimum rental requirement of 2 GPUs. In this case, you can rent between 2 and 8 GPUs, but cannot rent only 1 GPU.

SSH Access

This section allows you to configure how you will securely connect to your pod using SSH (Secure Shell).

SSH Key Configuration

SSH keys authenticate your access to the pod without needing a password:

  • Current Selection: Shows the SSH key that will be installed on your pod
  • Dropdown: Select from your list of previously added SSH keys
  • Add SSH Key: Upload a new SSH key if needed
šŸ” Security Note

Only you (or someone with your SSH private key) can access the pod after deployment, providing secure and convenient remote access.

Storage

External Volume Configuration

  • Volume Selection: Use the "Choose Volume" dropdown to select from available external storage volumes
  • Add Volume: Click to attach additional external storage volumes to your pod

For more details about how volumes work, see the Volumes documentation.

Restore Options

The Restore option allows you to restore a backup immediately when your pod is deployed. This is useful if you want your new pod to start with data from a previous backup.

  • Volume Path: /root (the directory where the backup will be restored)
  • Backup: Option to "Enter backup ID directly"
  • Select Backup: Button to choose from your existing backups

For more details on restores, see the Restore documentation.

Pod Summary

This section provides an overview of all the key settings and resources you have chosen for your pod before deployment. The summary panel helps you quickly review:

  • Hardware configuration (CPU, memory, disk, GPU)
  • Network details
  • Geographic location
  • Estimated pricing

Reviewing this summary ensures that your pod will be provisioned exactly as intended and allows you to confirm your selections before finalizing the deployment.

Advanced Options

Agent SSH Key

VoltageGPU automatically adds an agent SSH key to your pod to enable features like the web terminal and to monitor pod health.

  • Skip agent SSH key: Check this box if you do not want the agent SSH key added
āš ļø Warning

If you skip the agent SSH key:

  • The web terminal will be disabled
  • VoltageGPU cannot check SSH connectivity or monitor your pod
  • Connection issues cannot be verified or reported to the GPU provider

Recommendation: Keep the agent SSH key enabled unless you have a specific reason to disable it.

Deployment

Once all configurations are set, click the Deploy button to create and start your pod. The pod will be provisioned with the selected hardware, template, and access configurations.

Next Steps After Deployment

  • Connect to your pod via SSH using your configured key
  • Access your development environment
  • Monitor pod performance and costs through the dashboard
  • Manage volumes and backups as needed
āœ… Pro Tip

Make sure your SSH key is properly configured before deployment to ensure secure access to your pod.

Ready to Create Your Pod?

Browse available GPUs and deploy in under 60 seconds.

Browse Available GPUsView Templates