charent/ChatLM-mini-Chinese

Issues

• Some NCCL operations have failed or timed out.

  #47 opened 8 months ago by dbcSep03
  7

• 训练集与微调集数据重叠

  #60 opened 2 months ago by AndyZZt
  0

• 3.4预训练运行出现 unsupported operand type(s) 错误，求帮忙

  #55 opened 2 months ago by summerFF
  2

• 运行3.4python ptr_train.py时报错OSError: Can't load tokenizer for 'D:/pycharmenv/ChatLM-mini-Chinese/model_save/'.

  #56 opened 2 months ago by summerFF
  4

• tokenizer的字典中有不少token带有下划线，请问这种是什么意思

  #53 opened 5 months ago by Mactarvish
  3

• 4080显卡，基本跑不了多少数据，过万条训练数据就报错

  #54 opened 5 months ago by iissy
  6

• 数据清洗代码

  #58 opened 4 months ago by Mrkkew
  1

• tokenizer训练OOM 。内存60G

  #59 opened 4 months ago by musexiaoluo
  1

• 大佬，能不能分享一下清洗后的数据集呀，loss一直在4.0下不来

  #57 opened 5 months ago by KTVICTORY18
  0

• 可以用a卡训练吗

  #52 opened 5 months ago by alexhan1012
  1

• 预训练，用了160万数据，共2G句子对，使用A40的48G显存，无论使用1/2/3/4卡，都会报OOM

  #51 opened 5 months ago by JaymzWang
  1

• 这种只能通过问答对的方式，有没有办法MLM的方式学习知识体系。

  #50 opened 5 months ago by BShark-YB
  1

• 是否考虑将预训练的模型和仅stf后的模型也上传的平台呢

  #49 opened 5 months ago by seal-wang
  1

• sft_train

  #48 opened 5 months ago by dbcSep03
  1

• train_3.5M_CN数据处理问题

  #42 opened 5 months ago by wflying000
  1

• 是否有计划针对agent函数调用微调

  #19 opened 5 months ago by lucasjinreal
  4

• 用train.py出现shape的mismatch

  #36 opened 8 months ago by huluk98
  10

• 请问这些预训练数据加起来有多少token呀

  #44 opened 8 months ago by StarCycle
  2

• 如何加载sft后的模型？

  #41 opened 8 months ago by Liuxinhao12
  1

• 这个模型好像没有长文对话的能力，该如何训练它让它有这个能力？

  #43 opened 8 months ago by Liuxinhao12
  1

• 非常不错的开源项目

  #45 opened 8 months ago by DataXujing
  1

• 预训练数据集必须是{“prompt”: "response":}的格式么？

  #46 opened 8 months ago by dbcSep03
  2

• RuntimeError: No executable batch size found, reached zero

  #40 opened 9 months ago by suiyueyousan
  2

• 考虑出一个支持llama的版本吗

  #39 opened 9 months ago by leondada
  1

• 微调后预测三元组不正确原因

  #35 opened 9 months ago by qiutzh
  5

• sft微调时报错

  #37 opened 9 months ago by ama0zarashi
  4

• 如何提取中间层的输出？

  #38 opened 9 months ago by W-void
  2

• 运行·pre_train报错，TypeError: Accelerator.__init__() got an unexpected keyword argument 'use_seedable_sampler'

  #31 opened 9 months ago by JaymzWang
  1

• 使用Lora 和 sft_train.py 训练效果好像没有，有没有好的方法？

  #21 opened 9 months ago by yugu91
  7

• 请问，如果有新的内容需要添加，是否需要全部重新训练？

  #29 opened 9 months ago by kideve
  2

• Why do I get stuck loading the dataset after running it

  #24 opened 9 months ago by anyiz
  11

• 大佬请教一下，只做中文RAG的话，这个跟你另外一个phi，哪个效果比较好？

  #26 opened 9 months ago by xianzhisheng
  1

• Great Work! Does it support multimodal ability?

  #32 opened 9 months ago by LianghuiGuo
  1

• 预训练数据集

  #34 opened 9 months ago by rabintang
  2

• 项目怎么使用fastchat 进行调试

  #33 opened 10 months ago by zhilangtaosha
  1

• 请问数据预处理里面bell_open_source/train_0.8M_CN.json是在哪里下载的呀

  #30 opened a year ago by PshySimon
  7

• 多卡情况下，同一份数据集会加载多次吗

  #27 opened a year ago by shinerdeng
  6

• 请教“3.3 Tokenizer训练”如何运行？

  #25 opened a year ago by ybdesire
  2

• 有考虑将模型分发的https://modelscope.cn/么？

  #22 opened a year ago by qmjy
  2

• readme可以提供下封装了环境加模型的docker镜像吗?

  #20 opened a year ago by zack-sys
  1

• 如果在更好的设备上训练效果区别大吗

  #18 opened a year ago by aiwillcoming
  1

• 请教一个问题，生成的回复重复

  #17 opened a year ago by shinerdeng
  2

• 為甚麼我啟動API會出現這個

  #16 opened a year ago by Adolph3671
  1

• Hello, 第一次使用，请问运行时出现 unsupported operand type(s) for |: 'types.GenericAlias' and 'type' 是什么问题？

  #15 opened a year ago by yugu91
  2

• 是否可以在服务器上运行？

  #14 opened a year ago by yanyilin3344
  5

• 基于提供的模型进行sft报错

  #13 opened a year ago by cq1316
  13

• 清洗好的数据集会开源吗？

  #12 opened a year ago by echo-valor
  1

• 如何运行呢？

  #10 opened a year ago by meng25meng
  17

• 关于小模型ChatLM-mini-Chinese 信息抽取的 sft_train.json文件

  #9 opened a year ago by pengcheng-yan
  4

• 在 SFT 微调途中出现报错

  #23 opened a year ago by aoguai
  11