Pods

Deploy GPU-accelerated containers for training and inference.

Overview

Pods are ephemeral GPU instances that you can spin up on demand. They are powered by RunPod but managed directly through NeuralHub, offering a simplified experience optimized for AI workflows.

Pod Templates

To make deployment easy, NeuralHub uses "Pod Templates". These are pre-configured environments with the necessary drivers, libraries, and tools installed.

Available templates include:

• Axolotl Fine-tuning: Optimized for fine-tuning LLMs.
• Unsloth: Fast fine-tuning with Unsloth.
• vLLM Inference: High-performance inference server.
• Open WebUI: A user-friendly chat interface.

Billing

Pods are billed hourly based on the GPU type and disk usage.

• Billing is deducted from your unified USD wallet.
• You must have a positive balance to start a pod.
• Pods are billed as long as they are active. Stop or terminate a pod to stop billing.

Auto-Termination

To prevent accidental overspending, you can set an "Auto-terminate" timer when creating a pod. The pod will automatically shut down after the specified duration.