Dice Cuda Generation

Train diffusion language models to generate optimized CUDA kernels using bi-phase reinforcement learning. First phase masks and regenerates core kernel logic with provided scaffolding to prevent PyTorch shortcuts. Second phase enables end-to-end generation. Leverage diffusion's global context awareness for non-sequential code generation.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit c41de39da4

Frequently asked questions

npx get-skill add adu2021/dice-cuda-generation