Code

New optimized kernels#365

Code

Merged

casper-hansen merged 8 commits intomaincasper-hansen/AutoAWQ:mainfrom new_kernelscasper-hansen/AutoAWQ:new_kernelsCopy head branch name to clipboardFeb 24, 2024

+260-20

Commits

Commits on Feb 24, 2024

• Fix quantization

  casper-hansen

  committed
  56120a4

  Copy full SHA for 56120a4

• Add gemv fast (new)

  casper-hansen

  committed
  0d97169

  Copy full SHA for 0d97169

• Add GEMVFast in quantizer and loader

  casper-hansen

  committed
  2ef380c

  Copy full SHA for 2ef380c

• scaled_zeros -> qzeros (standardize name)

  casper-hansen

  committed
  5fc7c0e

  Copy full SHA for 5fc7c0e

• Enable fused qkv

  casper-hansen

  committed
  a8c7a6a

  Copy full SHA for a8c7a6a

• Use correct extension

  casper-hansen

  committed
  396786c

  Copy full SHA for 396786c

• Minor fix

  casper-hansen

  committed
  31d9266

  Copy full SHA for 31d9266

• Remove triton requirement

  casper-hansen

  committed
  f0ed32c

  Copy full SHA for f0ed32c