HandH1998/QQQ

Issues

• Transformer 4.46.1 compat

  #24 opened 17 days ago by Qubitium
  0

• Condition to achieve linear speedup?

  #15 opened 2 months ago by jiwonsong-dev
  18

• rotation+gptq data

  #20 opened a month ago by Andy0422
  7

• 关于qwen2-1.5b模型的问题

  #23 opened a month ago by darrenearl
  5

• Question about building W4A8 on AMD platform

  #22 opened a month ago by XIAOHUIL1
  2

• rotate + lm_head quantization

  #21 opened a month ago by RanchiZhao
  1

• Question on rotation

  #13 opened 3 months ago by cli99
  7

• Qwen2-1.5B 量化后精度完全不可用

  #17 opened a month ago by Juelianqvq
  18

• bugs: qqq_gemm.cu(183): error: identifier "__hfma2" is undefined

  #18 opened a month ago by Andy0422
  1

• 关于Marlin fetch_to_registers的问题

  #19 opened a month ago by darrenearl
  0

• How to use custom calib data?

  #1 opened 5 months ago by Juelianqvq
  7

• Possibility of using different group size setting

  #9 opened 3 months ago by NicoNico6
  6

• Qwen2-72B-Instruct packing failed

  #16 opened 2 months ago by Juelianqvq
  2

• Qwen2 supported？

  #14 opened 3 months ago by Juelianqvq
  5

• Plz share some calibration dataset or examples

  #11 opened 2 months ago by skykiseki
  2

• Does QQQ linear support H100?

  #12 opened 2 months ago by donglinz
  1

• smooth.py报错

  #8 opened 2 months ago by darrenearl
  1

• 关于group_size的问题

  #10 opened 3 months ago by darrenearl
  1

• 使用QQQ W4A8量化后的模型好像有问题。。。

  #7 opened 3 months ago by Zhao-Dongyu
  2

• [QST] Speedup of GEMM

  #3 opened 4 months ago by Hongbosherlock
  24

• Can MLA be smoothed?

  #6 opened 4 months ago by RanchiZhao
  5

• What is the prior for loss/error?

  #4 opened 4 months ago by RanchiZhao
  1

• [New Model Supported] MiniCPM-2.4B

  #5 opened 4 months ago by RanchiZhao
  3

• [QST] Scale factors and benchmarks

  #2 opened 5 months ago by jeromeku
  30