ymcui/Chinese-LLaMA-Alpaca-3

Issues

• 如何加快推理速度

  #109 opened 2 months ago by Mike-ihr
  1

• 扩充词表后，如何做增量预训练

  #108 opened 2 months ago by mc112611
  1

• 在设置了相同seed的情况下，多次运行完全相同的脚本得到的输出不同

  #107 opened 2 months ago by dreamingshao
  1

• llama-3中文是否经过对齐训练

  #103 opened a month ago by zzu-hzc
  2

• 我给出字数较长的复杂翻译任务时会答非所问。+竞技场是否能添加个自己写问题指定模型对比(独立记分或者不记分)？还有是否存在无内容审查版本计划？

  #102 opened a month ago by CheerfulStone
  2

• 跑原始的推理脚本出错，

  #105 opened 2 months ago by dreamlychina
  1

• v3 比 v2 Prompt 理解更差了

  #106 opened 2 months ago by osabc
  1

• 请问有可以学习的预训练模型吗？

  #111 opened a month ago by ShikangPang
  0

• 读取模型ggml-model-q8_0.gguf出错

  #88 opened 3 months ago by phoenixlucky
  3

• 进一步预训练可以全靠lora吗

  #92 opened 3 months ago by ymourenya
  4

• 中文数据集上的模型性能

  #87 opened 3 months ago by jerrywyn
  3

• 关于在lora版本上继续训练的问题

  #101 opened 2 months ago by gotimeqwer
  2

• llama cpp没有GUI

  #104 opened 2 months ago by LukeLIN-web
  1

• Instruct-v3 模型融合具体是怎么融合的，以及融合的出发点的什么

  #98 opened 3 months ago by Play2Boy
  2

• 已经在一张卡上执行了微调训练，现在想再执行另一个预训练，但脚本执行出错，请问要怎样修改脚本

  #95 opened 3 months ago by czhcc
  2

• 想詢問一下這次的訓練程式沒有使用deepspeed的原因

  #84 opened 3 months ago by ian08005454
  4

• 基于我自己的长文本训练数据，如何训练一个中文长文本模型，最长可以支持多长的长度

  #93 opened 3 months ago by jy-101361-1810897
  4

• sft训练数据问题

  #94 opened 3 months ago by leo-young
  2

• 模型微调时构建数据代码Output缺少终止符（输出重复）

  #59 opened 6 months ago by fangzheng123
  2

• 多机多卡训练，两机执行到这步后没有后续步骤了 [INFO|trainer.py:641] 2024-07-18 14:06:18,182 >> Using auto half precision backend

  #91 opened 3 months ago by cc8476
  2

• 多卡训练会报错 terminate called after throwing an instance of 'c10::Error' what(): CUDA error: unspecified launch failure

  #89 opened 4 months ago by cc8476
  2

• 提问不同问题显存会增加,相同问题则不会

  #83 opened 4 months ago by Chenhuaqi6
  3

• chinese-llama-2-13b-hf可否直接用bf16继续预训练？

  #85 opened 4 months ago by NLP-Learning
  3

• 请问在精调这样的数据格式是正确的吗？

  #86 opened 4 months ago by NiniAndy
  2

• 如何使用非alpaca格式的数据微调如pclue？

  #80 opened 4 months ago by lotus0903
  3

• Merging Instruct-v1 and Instruct-v2

  #74 opened 4 months ago by HuuHuy227
  4

• llama3的分词器

  #77 opened 4 months ago by ymourenya
  4

• 关于开始训练时出现了建立dataset失败事宜

  #82 opened 4 months ago by hk63560892
  2

• multi-node inference for llama3 70b

  #79 opened 4 months ago by Abolfazl-kr
  3

• 多卡训练会报错 terminate called after throwing an instance of 'c10::Error' what(): CUDA error: unspecified launch failure

  #90 opened 4 months ago by cc8476
  1

• sft执行慢的情况，请大佬帮忙看看

  #75 opened 4 months ago by lingaoan2024
  3

• 复现该项目的精调报错，具体如图

  #78 opened 4 months ago by hbs429469861
  2

• 在微调时，如果输入的prompt太长超过max_seq_length时，prompt最后的结束符会被丢弃

  #76 opened 4 months ago by seal-wang
  2

• 新人请教

  #72 opened 5 months ago by lingaoan2024
  4

• 微调时候Loss为0，grad_norm为Nan

  #64 opened 5 months ago by aa200647963
  7

• 请问什么时候能给llama3增加一个网页版demo脚本来使用

  #67 opened 5 months ago by dasaffa
  2

• 模型融合

  #63 opened 5 months ago by xiaoxiaoto
  2

• 请问中文对战平台是如何实现的，会开源相关代码吗

  #56 opened 5 months ago by Infinity4B
  3

• Colab中微调报错： CUDA out of memory

  #44 opened 5 months ago by chenmonster
  11

• 评价结果疑问

  #49 opened 5 months ago by fakerms
  2

• 使用 inference_hf.py 推理异常

  #52 opened 5 months ago by Xiaoshu-Zhao
  2

• 如何设置多卡训练？

  #51 opened 5 months ago by TDlemon-1900
  2

• hfl/ruozhiba_gpt4 的数据集有问题啊

  #73 opened 5 months ago by wencan
  3

• Merge完的模型在inference出錯

  #69 opened 5 months ago by MonetCH
  4

• merge lora model 时出現 error

  #68 opened 5 months ago by MonetCH
  1

• 训练过程种，异常中断问题

  #47 opened 5 months ago by AnonymousDestroyer
  4

• ruozhiba数据相关，有很多并不是高质量的回答，有一些是gpt4没有发现的陷阱

  #48 opened 5 months ago by AIchenkai
  2

• 使用原生llama3模型和使用Chinese-LLaMA-Alpaca-3模型推理的差异和问题

  #58 opened 6 months ago by yaoyonstudio
  4

• MacOS（苹果M3芯片）下指令精调报错

  #57 opened 6 months ago by yaoyonstudio
  3

• checkpoint file error

  #54 opened 6 months ago by jeffersyuan1976
  1