Nailabx Logo

Quick Start Guide

Deploy your first AI cluster on Nailabx infrastructure in minutes. Get up and running with GPU compute, storage, and networking.

Prerequisites

  • Nailabx Account: Sign up at nailabx.com
  • API Access: Generate your API key from the dashboard
  • CLI Tool: Install Nailabx CLI (optional but recommended)
  • SSH Key: Add your public SSH key for secure access
1

Install the CLI

macOS / Linux
curl -fsSL https://cli.nailabx.com/install.sh | bash
Windows (PowerShell)
iwr https://cli.nailabx.com/install.ps1 | iex
2

Authenticate

nailabx auth login

This will open your browser to authenticate. Alternatively, use an API token:

export NAILABX_API_TOKEN="your-api-token-here"
3

Deploy Your First Cluster

Create a GPU cluster with 2x NVIDIA H100 GPUs for AI training:

nailabx clusters create \
  --name my-first-cluster \
  --gpu-type h100 \
  --gpu-count 2 \
  --region us-east-1 \
  --storage 500GB

This command provisions a cluster with 2x H100 GPUs, 500GB NVMe storage, and connects it to our Tier III+ datacenter in US-East-1.

4

Connect via SSH

Once your cluster is provisioned (usually takes 2-3 minutes), connect via SSH:

nailabx clusters ssh my-first-cluster

Or use standard SSH with the provided IP address and your configured SSH key.

5

Run Your First AI Workload

Test your cluster with a simple PyTorch workload:

# Install PyTorch with CUDA support
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

# Verify GPU access
python -c "import torch; print(torch.cuda.is_available()); print(torch.cuda.get_device_name(0))"

# Run a simple training script
python train.py --epochs 10 --batch-size 32

Your cluster comes pre-configured with CUDA 12.1, cuDNN 8.9, and common ML frameworks.

Common Use Cases

LLM Fine-tuning

Fine-tune large language models using Hugging Face Transformers:

nailabx clusters create \
  --gpu-type h100 \
  --gpu-count 8 \
  --storage 2TB

Inference API

Deploy a production inference endpoint:

nailabx deploy inference \
  --model-path ./model.pt \
  --port 8080 \
  --replicas 3

Distributed Training

Scale training across multiple nodes:

nailabx clusters create \
  --gpu-type h100 \
  --gpu-count 16 \
  --nodes 4 \
  --interconnect nvlink

Jupyter Notebooks

Launch a GPU-enabled Jupyter environment:

nailabx notebooks create \
  --gpu-type a100 \
  --memory 64GB \
  --storage 500GB

Next Steps

API Reference

Explore our REST API and SDKs

Hardware Specs

Detailed GPU and server specifications

Billing & Support

Manage costs and get help