Instructions to use quarterturn/qwen-image-20b-bbwhot with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use quarterturn/qwen-image-20b-bbwhot with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Qwen/Qwen-Image", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("quarterturn/qwen-image-20b-bbwhot") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
A Qwen-image 20b LoRA for making BBWs (big beautiful women).
Note:
- Qwen seems to be sensitive to pulling the image off the LoRA style if you prompt for a style not in the dataset, i.e. "chibi". The LoRA was trained on photographs so it will do best with realisim prompts.
- steps 40, cfg 3.5 seemed to make good images
- I tested with LoRA strength 1.0
Trained using diffusion-pipe on 147 images to epoch 95 using the 'automagic' optimizer, rank 32.
- Downloads last month
- 558
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for quarterturn/qwen-image-20b-bbwhot
Base model
Qwen/Qwen-Image