Instructions to use xerofounder/XeroDesign-1.0 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use xerofounder/XeroDesign-1.0 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Tongyi-MAI/Z-Image", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("xerofounder/XeroDesign-1.0") prompt = "argonz_design poster" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
XeroDesign-1.0 (Z-Image 6B LoRA Adapter)
XeroDesign-1.0 is a specialized Rank-32 LoRA adapter fine-tuned on Tongyi-MAI/Z-Image (6B S3-DiT + Qwen3-4B Text Encoder) over the curated 696-poster argonz-curated-design dataset (Swiss International Typographic Style, Bauhaus geometric grids, asymmetric editorial layouts, and high-contrast modernist poster compositions).
Verified Endpoint Showcase Output (832x1216)
- Endpoint Route:
POST /v1/images/generations(Bearer CC1D7FF46EFFD318) - Prompt:
"argonz_design poster, Minimalist Swiss International Typographic Style exhibition poster for 'XERODESIGN 1.0 โ BAUHAUS ARCHIVE ZURICH', bold geometric sans-serif typography, asymmetric 4-column modular grid layout, high-contrast vermilion red and matte black geometric forms on warm off-white archival paper substrate, generous negative space corridor, razor-sharp vector edges." - Resolution:
832 ร 1216(1,517,613bytes, SHA-256:c87a172dda67c56cef97b0a08dd02d824600dd41cc02ad0a957b86e0017bb6a8) - Pixel Telemetry:
mean = 176.41,std = 70.18(Zero NaN / Zero Overflow)
Model Specifications
| Parameter | Specification |
|---|---|
| Model Name | XeroDesign-1.0 (xerofounder/XeroDesign-1.0) |
| Base Architecture | Tongyi-MAI/Z-Image (04cc4abb7c5069926f75c9bfde9ef43d49423021) |
| Adapter Type | PEFT LoRA (r=32, lora_alpha=32, Target: to_q, to_k, to_v, to_out.0) |
| Weights File | pytorch_lora_weights.safetensors (133,728,936 bytes / 133.7 MB) |
| Weights SHA-256 | 946fce0bc018a101825f9ec8429fbd3841f8e94e9d0e95d229d4b52065e268a1 |
| Dataset | xerocore/argonz-curated-design (696 high-resolution curated posters, multi-aspect buckets 1216x832, 832x1216, 1024x1024) |
| Precision & Numerics | 4-bit NF4 Base Quantization (bnb_4bit_use_double_quant=True) + Full FP32 Activation Compute Path (0 NaN steps; finite loss 0.243 - 0.324) |
| Trigger Token | argonz_design poster |
Quickstart: Load & Generate in 5 Minutes (server_xerodesign.py)
This repository includes server_xerodesign.py, a self-contained OpenAI/Llama-compatible FastAPI gateway (/v1/images/generations and /health) protected by the 16-digit Bearer key CC1D7FF46EFFD318.
pip install "diffusers @ git+https://github.com/huggingface/diffusers@e04f0c5f24392fcd34b32d18d0ce03668f9bcd50" "accelerate==1.11.0" "transformers==5.22.1" "peft==0.21.2" "bitsandbytes==0.48.1" "safetensors==0.8.0" "sentencepiece==0.2.1" "Pillow==11.3.0" "torchao==0.15.0" fastapi uvicorn huggingface_hub
python3 server_xerodesign.py
Example curl Request
curl -X POST http://localhost:8000/v1/images/generations \
-H "Authorization: Bearer CC1D7FF46EFFD318" \
-H "Content-Type: application/json" \
-d '{
"model": "XeroDesign-1.0",
"prompt": "argonz_design poster, Minimalist Swiss International Typographic Style exhibition poster, bold geometric sans-serif typography, vermilion red and matte black on warm off-white archival paper",
"size": "832x1216",
"steps": 24,
"guidance_scale": 4.0,
"seed": 42
}'
- Downloads last month
- 8
Model tree for xerofounder/XeroDesign-1.0
Base model
Tongyi-MAI/Z-Image