https://github.com/geocine/flux
Easy way to train Flux LoRA
https://github.com/geocine/flux
Last synced: 2 months ago
JSON representation
Easy way to train Flux LoRA
- Host: GitHub
- URL: https://github.com/geocine/flux
- Owner: geocine
- Created: 2024-08-18T14:16:40.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-06-14T18:46:34.000Z (4 months ago)
- Last Synced: 2025-06-14T20:00:43.218Z (4 months ago)
- Language: Python
- Homepage:
- Size: 61.5 KB
- Stars: 16
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Train Flux LoRA
> [!WARNING]
> If you're not a beginner, check out the original repo: https://github.com/ostris/ai-toolkit. This script is just for beginners.
https://github.com/user-attachments/assets/2c65f498-5185-4e0d-8344-cbd88068e6b2
This is meant to be run on the following spec server. Register at https://console.quickpod.io
| Specification | Recommended Value |
|-----------------|--------------------|
| GPU | Minimum 24GB VRAM. Use a 3090/4090 |
| Disk Space Size | 100GB |
## Setup
1. Initialize
```
curl -s https://geocine.github.io/flux/init.sh | sh
```
2. Paste your [huggingface](https://huggingface.co/settings/tokens) token on the `token` file.
3. Upload your photos on `data` folder
4. Modify `prompts` and `trigger_word` on `config.yaml`
5. Start training
```
./run.sh
```
6. After training your model, you can download it from the `output` folder. Checkout the [generate images](./docs/GENERATE.md) documentation.