Skip to content

Latest commit

 

History

History
11 lines (10 loc) · 1.4 KB

File metadata and controls

11 lines (10 loc) · 1.4 KB

Credits

  • Qubitium and ModelCloud team for maintaining and improving GPT-QModel
  • AlpinDale for creating the Swordfish GPTQ/AWQ Blackwell (>= sm100/sm110) w4a16/w8a16 GEMM kernel family.
  • Elias Frantar, Saleh Ashkboos, Torsten Hoefler and Dan Alistarh: for creating GPTQ and Marlin.
  • PanQiWei: for creation of AutoGPTQ which this project code is based upon.
  • FXMarty: for maintaining and support of AutoGPTQ.
  • Qwopqwop200: for quantization code used in this project adapted from GPTQ-for-LLaMa.
  • Turboderp: for releasing Exllama v1 and Exllama v2 kernels adapted for use in this project.
  • FpgaMiner: for GPTQ-Triton kernels used in GPTQ-for-LLaMa which is adapted into this project.
  • Casper Hansen: for maintaining and support of AutoAWQ for which the early version of the awq integration code was adapted from.