Apppublic

NiteshBadgujar/llada-image-base-demo

sourceHugging Faceapache-2.0updated 14d agoView on Hugging Face

0likes

App README

LLaDA-Image Base

Interactive demo for inclusionAI/LLaDA-Image, a 6B unified image generator and editor. Try text-to-image, VQ-conditioned generation, or natural-language editing with a reference image.

The app follows the official Base recipe: BF16 inference, 50 sampling steps, and guidance scale 5. It includes example prompts in English and Chinese.

—Model card
—Source code
—Paper

Hardware

This Space requests ZeroGPU xlarge because the full BF16 Base checkpoint is approximately 49 GB. The first request after startup can take several minutes while the model is streamed into the GPU worker. The model is loaded from the public Hugging Face repository at runtime; no model weights are stored in this Space.

License

The model is released under Apache-2.0. See the model repository for the complete license and usage details.