Diffusion Models for Language
9 repos
Language models built on diffusion processes rather than traditional autoregressive or transformer decoding. These repositories explore using diffusion-based approaches for text generation, pre-training, and fill-mask tasks, often trained on large-scale datasets like OpenWebText. The cluster includes both core model implementations and variants exploring different architectural choices like block sizes and training configurations.
Python · 5
diffusion-language-models ·715
diffusion-models ·715
text ·715
language-model ·715
transformers ·24
custom_code ·24
fill-mask ·24
safetensors ·24
mdlm ·23
language-modeling ·1
