Configuration

Complete guide to configuring your NeuroShard node.

Configuration Methods

NeuroShard can be configured via:

1. Command-line arguments (highest priority)
2. Environment variables
3. Configuration file

Command-Line Arguments

See CLI Reference for complete list.

neuroshard --token YOUR_TOKEN --port 8000 --memory 8192

Environment Variables

Variable	Description	Example
NEUROSHARD_TOKEN	Wallet token	abc123...
NEUROSHARD_PORT	HTTP port	8000
NEUROSHARD_TRACKER	Tracker URL	https://neuroshard.com/api/tracker
NEUROSHARD_MEMORY	Memory limit (MB)	8192
NEUROSHARD_LOG_LEVEL	Log verbosity	INFO

export NEUROSHARD_TOKEN=your_token
export NEUROSHARD_PORT=8000
neuroshard

Resource Configuration

Memory Limits

By default, NeuroShard uses 70% of available RAM.

# Limit to 4GB
neuroshard --token YOUR_TOKEN --memory 4096

# Limit to 50% of system RAM
neuroshard --token YOUR_TOKEN --memory $(free -m | awk '/^Mem:/{print int($2*0.5)}')

Memory allocation:

• Model weights: ~40%
• Gradients: ~20%
• Optimizer states: ~20%
• Activations: ~15%
• Overhead: ~5%

CPU Threads

# Use only 4 threads
neuroshard --token YOUR_TOKEN --cpu-threads 4

TIP

Leave at least 1-2 cores free for the HTTP server and network operations.

Storage Limits

# Limit Genesis data cache to 50MB
neuroshard --token YOUR_TOKEN --max-storage 50

Network Configuration

Port Configuration

NeuroShard uses two ports:

• HTTP Port: Dashboard and API (default: 8000)
• gRPC Port: Peer communication (HTTP port + 1000, default: 9000)

neuroshard --token YOUR_TOKEN --port 8000
# Opens: 8000 (HTTP), 9000 (gRPC)

NAT Traversal

If behind NAT, configure announce addresses:

neuroshard --token YOUR_TOKEN \
  --announce-ip 203.0.113.50 \
  --announce-port 8001

Firewall Rules

# UFW (Ubuntu)
sudo ufw allow 8000/tcp  # HTTP
sudo ufw allow 9000/tcp  # gRPC

# iptables
sudo iptables -A INPUT -p tcp --dport 8000 -j ACCEPT
sudo iptables -A INPUT -p tcp --dport 9000 -j ACCEPT

Training Configuration

DiLoCo Settings

# Sync more frequently (faster convergence, more bandwidth)
neuroshard --token YOUR_TOKEN --diloco-steps 100

# Sync less frequently (slower convergence, less bandwidth)
neuroshard --token YOUR_TOKEN --diloco-steps 1000

Inner Steps	Bandwidth	Convergence	Use Case
100	High	Fast	Good network
500	Medium	Normal	Default
1000	Low	Slow	Bad network

Disable Training

# Inference-only mode
neuroshard --token YOUR_TOKEN --no-training

GPU Configuration

CUDA (NVIDIA)

NeuroShard auto-detects CUDA. To use a specific GPU:

CUDA_VISIBLE_DEVICES=0 neuroshard --token YOUR_TOKEN

Apple Metal (MPS)

Auto-detected on Apple Silicon. No configuration needed.

Force CPU

CUDA_VISIBLE_DEVICES="" neuroshard --token YOUR_TOKEN

File Locations

Data Directory

Default: ~/.neuroshard/

Path	Contents
~/.neuroshard/checkpoints/	Model checkpoints
~/.neuroshard/data_cache/	Genesis data cache
~/.neuroshard/logs/	Node logs

Custom Data Directory

export NEUROSHARD_DATA_DIR=/data/neuroshard
neuroshard --token YOUR_TOKEN

Docker Configuration

Docker Compose

version: '3.8'
services:
  neuroshard:
    image: neuroshard/node:latest
    ports:
      - "8000:8000"
      - "9000:9000"
    environment:
      - NEUROSHARD_TOKEN=${NEUROSHARD_TOKEN}
      - NEUROSHARD_MEMORY=8192
      - NEUROSHARD_LOG_LEVEL=INFO
    volumes:
      - neuroshard_data:/data
    deploy:
      resources:
        limits:
          memory: 8G
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

volumes:
  neuroshard_data:

Docker Run

docker run -d \
  --name neuroshard \
  --gpus all \
  -p 8000:8000 \
  -p 9000:9000 \
  -e NEUROSHARD_TOKEN=YOUR_TOKEN \
  -e NEUROSHARD_MEMORY=8192 \
  -v neuroshard_data:/data \
  neuroshard/node:latest

Systemd Configuration

/etc/systemd/system/neuroshard.service:

[Unit]
Description=NeuroShard Node
After=network.target

[Service]
Type=simple
User=ubuntu
Environment="NEUROSHARD_TOKEN=your_token_here"
Environment="NEUROSHARD_PORT=8000"
Environment="NEUROSHARD_MEMORY=8192"
Environment="NEUROSHARD_LOG_LEVEL=INFO"
ExecStart=/usr/local/bin/neuroshard --headless
Restart=always
RestartSec=10

[Install]
WantedBy=multi-user.target

sudo systemctl daemon-reload
sudo systemctl enable neuroshard
sudo systemctl start neuroshard

Tuning Guide

Low Memory (<4GB)

neuroshard --token YOUR_TOKEN \
  --memory 2048 \
  --max-storage 25 \
  --diloco-steps 500

High Performance (>16GB)

neuroshard --token YOUR_TOKEN \
  --memory 32768 \
  --max-storage 500 \
  --diloco-steps 500

Low Bandwidth

neuroshard --token YOUR_TOKEN \
  --diloco-steps 1000 \
  --no-training  # Or inference-only

See Also

• CLI Reference
• Running a Node
• Troubleshooting