DoubleClark

Pinned Repositories

• roctx4py

  Language:Cython0110

• SageAttention

  Quantized Attention that achieves speedups of 2.1-3.1x and 2.7-5.1x compared to FlashAttention2 and xformers, respectively, without lossing end-to-end metrics across various models.

  Language:Cuda788217243

DoubleClark's Repositories