Image-to-Image
Diffusers
Safetensors
QwenImage21Pipeline
qwen21
sdnq
4-bit precision
8-bit precision
Instructions to use SamuelTallet/Pruna-Qwen-Image-2.1-8steps-SDNQ-4bit-dynamic-hadamard256 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use SamuelTallet/Pruna-Qwen-Image-2.1-8steps-SDNQ-4bit-dynamic-hadamard256 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline from diffusers.utils import load_image # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("SamuelTallet/Pruna-Qwen-Image-2.1-8steps-SDNQ-4bit-dynamic-hadamard256", dtype=torch.bfloat16, device_map="cuda") prompt = "Turn this cat into a dog" input_image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png") image = pipe(image=input_image, prompt=prompt).images[0] - Notebooks
- Google Colab
- Kaggle
This is Qwen-Image-2.1 merged with Pruna 8 steps LoRA v0.1, optimized using SDNQ with UINT4 dynamic quantization and Hadamard Rotation (Group size: 256).
Texture fix VAE made by Ollin Boer Bohan is included to avoid the checkerboard artefacts in outputs.
Sample
Prompt:
Close portrait of a guitarist sitting on a bench in a lush green park, softly playing an acoustic guitar. Surrounding him are tall trees, a few fallen leaves, and a peaceful pathway. Golden late-afternoon light, a calm and warm atmosphere, cinematic rendering, realistic details, and natural depth of field.
Seed: 42
Usage
Install Torch, Diffusers (Git), SDNQ 0.2.0+ and Triton.
8 steps.
- Downloads last month
- 100
Model tree for SamuelTallet/Pruna-Qwen-Image-2.1-8steps-SDNQ-4bit-dynamic-hadamard256
Base model
Qwen/Qwen-Image-2.1