Zefan-Cai/PyramidKV

Issues

• Bug Report

  #21 opened 3 months ago by Patrick-Ni
  3

• Question about evaluating Llama-3-8B-Instruct on LongBench

  #12 opened 4 months ago by jameslahm
  8

• 对比了下StreamingLLM源码实现，貌似PyramidKV中的实现位置编码这块有问题

  #19 opened 3 months ago by free-stardust
  4

• 请问 kv_cluster.update_kv 只在 prefill 时被调用吗？

  #20 opened 3 months ago by DOG-wooooof
  3

• A serious issue in your code

  #15 opened 4 months ago by JulietLJY
  6

• 仅截取了输入数据的前4k和后4k token, 但longbench最长32k token？

  #18 opened 3 months ago by JulietLJY
  1

• Strange Results of Mistral

  #13 opened 3 months ago by yuhuixu1993
  1

• Merge into vLLM, is it possible?

  #14 opened 4 months ago by PatchouliTIS
  4

• 判断条件的依据

  #17 opened 4 months ago by exectrl
  3

• Settings and implementations of baselines

  #16 opened 4 months ago by bingps
  3

• Congratulation on the great work & hope for future versions

  #5 opened 5 months ago by doptime
  2

• Mistral 7B full kv cache out of memory

  #11 opened 4 months ago by monster119120
  6

• Llama3-8B model url

  #8 opened 4 months ago by monster119120
  8

• question about case selection in observations

  #10 opened 4 months ago by Cooperx521
  1

• confused about some code

  #6 opened 4 months ago by bohr
  3

• correct transformers version

  #7 opened 4 months ago by monster119120
  1

• mistral7B运行错误

  #9 opened 4 months ago by monster119120
  2

• What prompt was used in Needle in a Haystack test?

  #4 opened 5 months ago by 66RING
  4