DGX Spark Playbooks
Collection of step-by-step playbooks for setting up AI/ML workloads on NVIDIA DGX Spark devices with Blackwell architecture.
About
These playbooks provide detailed instructions for:
- Installing and configuring popular AI frameworks
- Running inference with optimized models
- Setting up development environments
- Connecting and managing your DGX Spark device
Each playbook includes prerequisites, step-by-step instructions, troubleshooting guidance, and example code.
Available Playbooks
NVIDIA
- Register AI Compute with Brev
- Set Up CLI Coding Agents with Local Inference
- Generate Images and Videos with ComfyUI
- Connect Multiple DGX Sparks for Distributed Workloads
- Connect Three DGX Sparks in a Ring Topology
- Use NVIDIA Sync to Connect Remotely to Your AI Compute
- Connect Two DGX Sparks for Distributed Workloads
- Connect Two Nodes for Distributed Workloads
- Accelerate Data Science with CUDA-X Libraries
- Run cuPyNumeric Across Two DGX Sparks
- Run cuTile Kernels on DGX Spark
- Monitor Your AI Compute with DGX Dashboard
- Set Up Agent Skills for DGX Station
- Fine-Tune Specialized LLMs with Unsloth
- Fine-Tune FLUX.1 for Custom Image Generation
- Fine-Tune Isaac GR00T for Robot Skills
- Run Healthcare Agents with Local Inference
- Run Hermes Agent with a Local LLM
- Set Up Isaac Sim and Isaac Lab
- Accelerate JAX Performance
- Speed Up Fine-Tuning with Triton Kernels
- Stream Real-Time Video to a Vision Language Model
- Set up an AI Home Assistant
- Serve Models with llama.cpp
- Fine-Tune LLMs with LLaMA Factory
- Run Local LLMs
- Serve LLMs with LM Studio
- Set Up Claude Code with Local Inference
- Set Up Multi-Instance GPU (MIG)
- Build and Deploy a Multi-Agent Chatbot
- Build a Multi-GPU AI PC
- Run Multi-Modal Inference with TensorRT
- Connect Multiple DGX Sparks Through a Switch
- Run Chat Model Training with nanochat
- Set Up NCCL for Multi-Node GPU Communication
- Fine-Tune with NVIDIA NeMo
- Run NemoClaw with a Local LLM
- Set Up Example NemoClaw Agents
- Serve Nemotron Nano and Super Models
- Deploy NVIDIA NIM for LLM Inference
- Run NVFP4 Pretraining with Megatron Bridge
- Quantize Models to NVFP4 with NVIDIA Model Optimizer
- Chat with LLMs Using Open WebUI and Ollama
- Run OpenClaw with a Local LLM
- Secure AI Agents with OpenShell
- Install and Use NVIDIA PAIR with llama.cpp
- Accelerate Portfolio Optimization with cuOpt
- Fine-Tune with PyTorch
- Build a RAG Application with AI Workbench
- Serve LLMs with SGLang
- LLM Inference with SGLang
- Accelerate Single-Cell RNA Data Analysis
- Build an AI Photo Booth with Reachy and DGX Spark
- Speed Up Inference with Speculative Decoding
- Set Up Tailscale for Remote Access
- Example Test Playbook
- Accelerate Topic Modeling with BERTopic
- Serve LLMs with NVIDIA TensorRT-LLM
- Build Knowledge Graphs with txt2kg
- Fine-Tune Faster with Unsloth
- Set Up Vibe Coding in VS Code
- Generate Controlled Video with ComfyUI
- Generate Images and Video with ComfyUI
- Serve LLMs with vLLM on One Device
- Serve LLMs with vLLM Across DGX Sparks
- Fine-Tune Vision Language Models
- Set Up VS Code for Local and Remote Development
- Deploy a Video Search and Summarization Agent
Resources
- Documentation: https://www.nvidia.com/en-us/products/workstations/dgx-spark/
- Developer Forum: https://forums.developer.nvidia.com/c/accelerated-computing/dgx-spark-gb10
- Terms of Service: https://assets.ngc.nvidia.com/products/api-catalog/legal/NVIDIA API Trial Terms of Service.pdf
License
See:
- LICENSE for licensing information.
- LICENSE-3rd-party for third-party licensing information.