license: mit library_name: transformers pipeline_tag: text-generation
We introduce LLaDA, a diffusion model with an unprecedented 8B scale, trained entirely from scratch, rivaling LLaMA3 8B in performance.
Project Page
Code