KMnO4-zx/huanhuan-chat

Issues

• 数据集inputs和labels重合的问题

  #31 opened a month ago by YuGuilliman
  2

• 鲁迅

  #30 opened 3 months ago by world2025
  0

• 只能是Chatgpt的api吗？国内想获取他们的api都要手机验证，然后就不通过？说国家不支持啥的，能不能文心一言，或者其他国内大模型上获取的api呢？

  #29 opened 4 months ago by iamdrivenbymywonderfullife
  1

• Why is the input text written in "Instruction" instead of "input"?

  #4 opened a year ago by Alsacen
  2

• 微调后没起作用

  #26 opened 7 months ago by MrLuer
  3

• 为什么在训练的时候loss直接变0，梯度直接nan了？

  #23 opened 9 months ago by Sheldon090
  1

• 有GPU正常训练应该要多快

  #28 opened 6 months ago by TankKevin
  1

• 微调效果如何评估

  #27 opened 6 months ago by xielinzhen
  2

• pip安装依赖时报安装包冲突。

  #20 opened a year ago by zhanghui-china
  2

• error: the following arguments are required: --output_dir

  #25 opened 7 months ago by Liqian0506
  0

• 没有GPT的情况下怎样复现嬛嬛

  #19 opened a year ago by lslslslslslslslslslsls
  4

• pip install -r requirement.txt 安装需要进行如下的修改

  #24 opened 8 months ago by zhouqian177
  0

• 搭建一个小程序/网站

  #22 opened 10 months ago by lmgyuan
  1

• bash generation.sh报错：Error communicating with OpenAI:exceeded with url: vl/chat/completions (Caused by SSLError(CertificateError("hostnameapi.openai.com'

  #21 opened a year ago by zhanghui-china
  2

• 关于huanhuan-chat微调报错的问题

  #14 opened a year ago by JiQiangHuang
  3

• 几个问题

  #18 opened a year ago by Chenwe111
  4

• 请教sft训练数据集设计的依据是什么？

  #17 opened a year ago by Bryce1010
  1

• 多卡微调时报错

  #16 opened a year ago by surviveMiao
  9

• Error Report: pip install no such file or directory

  #15 opened a year ago by hsiong
  3

• 24G GPU can use for classffication in this project?

  #13 opened a year ago by CK-Cheung-CK
  1

• 请问sft数据集是如何整理的，可以提供方法吗？

  #10 opened a year ago by txy6666yr
  3

• System prompt 如何设计？

  #8 opened a year ago by lucasjinreal
  3

• 甄嬛传的剧本txt是如何下载的啊

  #9 opened a year ago by lucasjinreal
  1

• 无法使用peftmodel

  #3 opened a year ago by MZGuo111
  1

• 多轮对话数据

  #2 opened a year ago by mz2sj
  1