New
introspective-diffusion/dlmrt — inference engine for block diffusion language models
introspective-diffusion/dlmrt is an inference engine that processes block diffusion language models one request at a time, structured to run the denoising loop on the GPU.