Skip to content

Getting Started

Jump to bottom
The Duskfall Portal Crew edited this page Jun 12, 2026 · 4 revisions

Getting Started

v0.2.1-BETA Cross-platform LoRA training interface (Next.js + FastAPI)

This guide walks you through installing and running your first training session. For detailed platform-specific instructions, see Installation.

1. Prerequisites

Component Requirement Notes
GPU NVIDIA with CUDA 12.1+ 12 GB VRAM minimum, 24 GB recommended for SDXL/Flux
Python 3.10 or 3.11 Other versions may cause dependency conflicts
Node.js 20.19+ 22.x recommended
Disk Space 50 GB+ free For models, datasets, and cache
Git Latest For cloning and updates

2. Quick Install

Windows

git clone https://github.com/Ktiseos-Nyx/Ktiseos-Nyx-Trainer.git
cd Ktiseos-Nyx-Trainer
install.bat

Install to a path you own (e.g. C:\Users\YourName\Projects\). Avoid system directories, OneDrive, or network drives.

Linux

git clone https://github.com/Ktiseos-Nyx/Ktiseos-Nyx-Trainer.git
cd Ktiseos-Nyx-Trainer
./install.sh

What the installer does:

  1. Checks for Python 3.10/3.11 and Node.js 20+
  2. Creates a virtual environment (.venv/) unless --no-venv is passed
  3. Installs PyTorch with CUDA 12.1 index
  4. Installs Python dependencies
  5. Runs npm install && npm run build in frontend/

Installer flags:

Flag Effect
--venv Create venv without prompting
--no-venv Skip venv creation
--auto Non-interactive (implies --venv)
--no-comfyui Skip ComfyUI installation
--verbose Show full pip/npm output

3. Starting the Application

start_services_local.bat    # Windows
./start_services_local.sh   # Linux

Services:

URL Description
http://localhost:3000 Frontend UI
http://localhost:8000 Backend API
http://localhost:8000/docs API documentation

Custom ports:

start_services_local.bat --port 4000 --backend-port 9000
./start_services_local.sh  --port 4000 --backend-port 9000

Quick restart (no rebuild):

restart.bat     # Windows
./restart.sh    # Linux

4. Cloud Deployment

One-click deploy buttons are available in the README. Both auto-configure on launch.

VastAI: The template runs vastai_setup.sh on first boot. Services restart automatically via Supervisor.

RunPod: The template runs provision_runpod.sh on first boot.

See Installation#Cloud-Deployment for manual setup.

5. Your First Training Run

  1. Open the UI: Navigate to http://localhost:3000
  2. Configure Training:
    • Model Path: Use a HuggingFace path (e.g. runwayml/stable-diffusion-v1-5) or local path
    • Dataset Path: Path to your training data
    • Optimizer: Start with AdamW8bit
  3. Save Config: Click "Save" to persist settings
  4. Start Training: Click "Start Training" and monitor the output panel
  5. Verify:
    • Training logs appear in real-time
    • GPU activity via nvidia-smi

6. Updating

git pull
install.bat       # Windows
./install.sh      # Linux

The installer is idempotent — re-running on an existing install only updates what changed.

If reinstalling doesn't fix issues:

python clean_slate.py --dry-run    # Preview what will be removed
python clean_slate.py              # Remove build artifacts, venv, node_modules
python clean_slate.py --nuclear    # Also remove models, datasets, outputs

After clean_slate.py, re-run the installer.

7. Next Steps

Need Help?

  1. Check Troubleshooting
  2. Run diagnose.bat (Windows) or ./diagnose.sh (Linux)
  3. Open a GitHub Issue with the diagnostic output

Clone this wiki locally