hahnyuan/RPTQ4LLM

Issues

• About R2 in figure 3 of the paper

  #13 opened 7 months ago by lucky9-cyou
  5

• If it would be better to cluster based on the values of Xmax-Xmin?

  #12 opened 9 months ago by tp-nan
  0

• 有配置perchannel的option选择吗？

  #6 opened 2 years ago by wangshankun
  6

• The accuracies of zero-shot tasks can not be reproduced

  #11 opened a year ago by NewDriverLee
  0

• inps in opt_reorder_quantize

  #10 opened a year ago by ZoePG
  0

• Share RoPE / LLaMA reordering code

  #9 opened a year ago by Artexety
  0

• 为什么 求MinMax 是在 [0, 1]轴上取

  #8 opened 2 years ago by MeJerry215
  0

• 使用对称量化配置进行量化，这一行导致返回NoneType

  #7 opened 2 years ago by MeJerry215
  0

• License issues

  #5 opened 2 years ago by AlpinDale
  1

• 如何计算weights/KV-cache/Dynamic Activation的占比

  #4 opened 2 years ago by xingyueye
  0

• RPTQ improvements

  #2 opened 2 years ago by qwopqwop200
  1

• Add support for RWKV model family

  #1 opened 2 years ago by bennmann
  0