lich99/ChatGLM-finetune-LoRA

Issues

• 有什么卡训练的，V100完全搞不定

  #15 opened a year ago by huzaizi2023
  10

• No such file or directory: '/root/.cache/huggingface/modules/transformers_modules/chatglm-6b/tokenization_chatglm.py'

  #29 opened a year ago by Data2Me
  8

• 报错

  #49 opened a year ago by zjbian
  0

• 报错

  #48 opened a year ago by fxb392
  0

• 不考虑GLM的双向注意力部分，注意力矩阵不是一个下三角矩阵吗

  #47 opened a year ago by fxb392
  0

• 显存问题

  #46 opened a year ago by zhangsanjava
  0

• 位置编码

  #45 opened a year ago by fxb392
  3

• 作者你好，这个项目对python的版本有什么要求吗，3.7是否可以？

  #44 opened a year ago by tjulh
  0

• ValueError: 150004 is not in list是什么回事？

  #28 opened a year ago by z1968357787
  3

• peft 0.3.0如何设置adapater_name

  #38 opened a year ago by moseshu
  8

• while cnt < retry_cnt:

  #43 opened a year ago by fxb392
  0

• 问答数据集如何构建

  #42 opened a year ago by Godlikemandyy
  0

• [deepspeed] OVERFLOW!

  #30 opened a year ago by JingerAI
  1

• 请问有交流群吗?

  #41 opened a year ago by roki1031
  0

• 关于ZeRO的疑问？

  #33 opened a year ago by MAxx8371
  3

• 有大佬们试过用8卡训练的吗？

  #40 opened a year ago by tu2022
  0

• 请问LORA模型参数怎么加载到原模型里呢？

  #39 opened a year ago by ZeyuBa
  0

• finetune没效果

  #34 opened a year ago by ChenBinfighting1
  4

• context_length = obj['prompt'].index(130004)

  #37 opened a year ago by moseshu
  2

• LoRA训练时间大概是多久呢？

  #36 opened a year ago by realcarlos
  3

• 这个显卡要求一定是bfloat16吗

  #26 opened a year ago by z1968357787
  3

• example.ipynb中进行训练测试loss为nan

  #32 opened a year ago by SilentMoebuta
  1

• torch.distributed.elastic.multiprocessing.errors.ChildFailedError, when running the train_new.py

  #27 opened a year ago by Skywalker-Harrison
  2

• LoRA的A矩阵一直不更新

  #22 opened a year ago by qz701731tby
  1

• 关于分布式GPU训练

  #25 opened a year ago by z1968357787
  1

• example_simple报错

  #24 opened a year ago by qishisurenhhh
  1

• here are my questions,I have more than 4 gpus to run the train.py,but it still out of memory,I check the usage of memory and find that one of them overflows and produce the bug,how can I solve it?

  #23 opened a year ago by z1968357787
  0

• subprocess.CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 1.

  #17 opened a year ago by xiamaozi11
  2

• 单机多卡报错

  #20 opened a year ago by ForgetThatNight
  0

• 您好，问一下，这个训练完有可以展示与原来基础模型进行对比的测试效果吗

  #21 opened a year ago by kunshou123
  0

• 请问下Finetune之后能实现企业定制FAQ的效果吗？可能有一百个问答这样

  #8 opened a year ago by terryops
  8

• 训练loss变为NaN

  #19 opened a year ago by qz701731tby
  2

• 能否使用量化后的chatGLM-6b-int4小模型进行微调？

  #7 opened a year ago by valkryhx
  4

• 可以给一下具体的环境requirement吗?

  #11 opened a year ago by Data2Me
  1

• 训练后的结果对应不上

  #4 opened a year ago by zhangyanbo2007
  1

• About multi-GPU

  #9 opened a year ago by zhongtao93
  6

• 如何设置batch_size个数，变动后train会变维度上的错误

  #12 opened a year ago by GUORUIWANG
  1

• train.py的命令行启动是什么？

  #13 opened a year ago by Data2Me
  1

• 请问DeepSpeed与Multi-gpu是绑定的吗？使用DeepSpeed提升效果有多大？

  #10 opened a year ago by nameless0704
  2

• LORAConfig报错：ValueError: Target modules ['q', 'k', 'v'] not found in the base model. Please check the target modules and try again.

  #14 opened a year ago by nameless0704
  6

• 已获取

  #18 opened a year ago by lbxcfx
  0

• RuntimeError: CUDA error: invalid device ordinal

  #16 opened a year ago by yanqiangmiffy
  1

• 两个epochs之间，loss值并没有降下来

  #6 opened a year ago by aizpy
  1

• NameError: name 'train_dataloader' is not defined

  #5 opened a year ago by wccccp
  1

• 训练超显存

  #3 opened a year ago by GaoPengGit
  5

• 数据集和微调模型的一些问题

  #2 opened a year ago by SarmonFish
  1

• Is there any details about dataset?

  #1 opened a year ago by 980202006
  2