Integrations / Kohya
RunDocumented
Run LoRA training with Kohya
Start the Kohya trainer on a cloud GPU. Upload your dataset, train a LoRA, download the result.
The steps follow the tool's documented settings. No tested run is recorded for this page yet.
Why use Badgr for Kohya
- Badgr interface
- Run
- Where Kohya fits
- Training runtime
- Why Badgr
- LoRA fine-tuning
- Train a LoRA on a GPU with enough memory without buying one.
- Start the trainer from a built-in template instead of building the environment yourself.
- Spend is capped, and you stop it when training is done.
What you can do
- Train a character or style LoRA for an image model
- Train on a GPU larger than the one you own
- Repeat training with a different dataset
Quickstart: Start the Kohya trainer
badgr template info kohya-ss badgr serve template kohya-ss
How it works
- 1.You start the trainer
- 2.Badgr gives you an address for its web interface
- 3.You upload data and start training there
- 4.You download the LoRA and stop the service
What you get
- LoRA weight files
- Training logs
Requirements
- The Badgr CLI installed and logged in (npm install -g badgr-cli)
- A funded Badgr balance
- A prepared dataset (images with captions) in the layout Kohya expects
- Enough GPU memory for your base model and resolution
What it costs
- This template is interactive, so it bills until you stop it, not until a command ends. Stop it with badgr down <id> when training is done.
- Pass --max-cost to bound spend.
See pricing for how billing works.
Proof
No tested run is recorded for this integration yet. A receipt and example output will be added here after a real run, not before.
Troubleshooting
Out of memory during training
Lower the resolution or batch size, enable gradient checkpointing, or use a GPU with more memory.
Dataset not found
Check the folder layout and the repeat-count prefix on image folders, which Kohya requires.
Service left running
Run badgr status and stop it with badgr down <id>. Files on the service are removed when it stops, so download first.
Developer reference
Ready to start with Kohya?
Every run or service has a spending limit you set before it starts.