Facico/Chinese-Vicuna

Issues

• 关于中断训练继续训练。

  #258 opened 9 months ago by xxyNeepu
  16

• 可以更新一下requirements吗？

  #257 opened a year ago by estuday
  1

• 运行generate脚本之后，在页面提问，很久没有产生回答，后台无报错

  #247 opened a year ago by mmmminyuhan
  2

• 如果更改數據集格式，要如何更改代碼

  #255 opened a year ago by alexaax
  0

• 多卡训练 bash scripts/finetune.sh报错

  #245 opened a year ago by hdjghjb
  1

• 可以使用原始文本微调吗

  #212 opened a year ago by gravitywp
  2

• 官方colab安裝套件失效

  #254 opened a year ago by alexaax
  0

• 可以提供一下huggingface上的Chinese-Vicuna/llama7b_4bit_128g模型的config.json和tokenizer么？

  #253 opened a year ago by jasoncow007
  0

• 使用finetune.sh来指令微调llama-33b，出现ZeroDivisionError: integer division or modulo by zero错误

  #252 opened a year ago by BIUBIUBIU-JIAZHOU
  2

• deepspeed跑模型相关问题

  #250 opened a year ago by sunpenglv
  0

• 从belle+guanaco数据集中抽取前5000条样本训练lora，效果不好

  #249 opened a year ago by huanghaifeng1234
  0

• OSError: Not enough disk space. Needed: Unknown size (download: Unknown size, generated: Unknown size, post-processed: Unknown size)

  #248 opened a year ago by thugbobby
  0

• 训练模型没有保存token信息

  #230 opened a year ago by apachemycat
  5

• 这几个不同路径下的模型是否有区别？

  #246 opened a year ago by hdjghjb
  0

• 运行chat_7B.sh聊两句话out of memory

  #244 opened a year ago by hdjghjb
  0

• 中文乱码

  #241 opened a year ago by NewEricWang
  5

• 请问llama7b_4bit_128g的input shape是多少呢

  #243 opened a year ago by KyrieZhang11
  1

• 请问多个lora模型怎么合并？

  #242 opened a year ago by Orangeices
  0

• 多卡finetune_chat时报mat1 and mat2 shapes cannot be multiplied (1024x2 and 1x11008)

  #240 opened a year ago by 18065013
  2

• RuntimeError: mat1 and mat2 shapes cannot be multiplied (164x4096 and 1x8388608)

  #228 opened a year ago by adaaaaaa
  3

• 是因为梯度为0吗？

  #239 opened a year ago by X1a0X
  0

• transformers和pydantic问题

  #238 opened a year ago by ww0o0
  1

• 有办法改成分类任务么，用LlamaForSequenceClassification模型类加载

  #237 opened a year ago by LeonhardtWang
  0

• ⁇ Below is an instruction that describes a task. Write a response

  #236 opened a year ago by vcbeaut
  0

• Traceback (most recent call last):RuntimeError: "addmm_impl_cpu_" not implemented for 'Half'

  #231 opened a year ago by alps008
  1

• 用checkpoint-11600跑部分问题(目测10-20%的问题)有奇怪的无限循环

  #235 opened a year ago by Tongcheng
  1

• 运行bash scripts/generate.sh或者bash scripts/chat_7B.sh后一般多久就可以进行推理了

  #234 opened a year ago by Junglesl
  1

• 简单的问题，finetune_other_continue.sh中step = 样本量/（MICRO_BATCH_SIZE*GRADIENT_ACCUMULATION_STEPS）。多卡的时候，是不是应该得是 step = 样本量/batch/（MICRO_BATCH_SIZE*GRADIENT_ACCUMULATION_STEPS*gpu数量）数量呢？这边一直不是很理解

  #233 opened a year ago by niuhuluzhihao
  0

• scripts中不同训练脚本的差异与训练效果的问题

  #207 opened a year ago by FURYFOR
  2

• 拉去最新分支之后，通过pip install安装好了bitsandbytes==0.37.2，但是通过finetune_other_continue执行的时候，报此模块没有__version__

  #219 opened a year ago by niuhuluzhihao
  5

• 7B 模型单卡3090后处理非常耗时

  #232 opened a year ago by f18298335152h
  0

• 现在哪个模型支持4060笔记本显卡下的推理或者训练吗？

  #209 opened a year ago by adaaaaaa
  1

• 在实际应用中我如何将num_beams=4，但最终输出的时候可保证输出过程和结果是一致的

  #213 opened a year ago by zjwlgr
  1

• 推理报错：RuntimeError: expected scalar type Half but found Float

  #210 opened a year ago by zhouchangju
  2

• 我使用7B参数的上游模型 + 100万个问答数据集做微调，时间需要48天，如何能加快

  #214 opened a year ago by zjwlgr
  1

• 官方 finetune colab 无法运行

  #215 opened a year ago by williamjqk
  1

• target_modules 各参数是什么意思，如何选择参数进行针对性的微调？

  #216 opened a year ago by pan365wang
  2

• finetune_deepspeed启动运行[ERROR] [launch.py:324:sigkill_handler]

  #218 opened a year ago by grantchenhuarong
  4

• scripts 中好像没有 直接从Chinese-Vicuna/Chinese-Vicuna-lora-7b-chatv1继续训练微调的版本

  #220 opened a year ago by svjack
  3

• 代码中关于EOS paddding的区别问题

  #221 opened a year ago by apachemycat
  1

• llama-13b-hf做推理，CUDA out of memory. 问题

  #224 opened a year ago by Bingohong
  2

• 运行generate.py推理报错ValueError: We need an `offload_dir` to dispatch this model

  #225 opened a year ago by kakuibeyond
  3

• Not an issue but a question for going forwards

  #227 opened a year ago by thusinh1969
  1

• 怎么区分user的问题是指令问题还是通用问题？

  #229 opened a year ago by suckseed5
  1

• 支持不同词表大小的 llama模型训练 lora

  #226 opened a year ago by greatewei
  0

• 为什么我在 kaggle.com 上训练的 LoRA 模型效果比较不错，模型下载到本地进行推理效果却很差？

  #217 opened a year ago by jianghushinian
  2

• Generation问题

  #223 opened a year ago by Jiangchenglin521
  0

• Infra问题

  #222 opened a year ago by Jiangchenglin521
  0

• 使用CPU运行13B的模型，有2个bin文件怎么选择呢

  #211 opened a year ago by hengxingtx
  1

• 运行chat_7B.sh报错

  #208 opened a year ago by hongshuo-wang
  0