ssbuild/chatglm_finetuning

Issues

• python train.py执行训练报错，求解。

  #214 opened a year ago by pan365wang
  9

• 请问adalora能用deepspeed训练吗

  #266 opened a year ago by Yu-Yuqing
  0

• LoRA和ptv2微调均发生OOM

  #264 opened a year ago by shenzhyzzz
  4

• INFO:lightning_fabric.utilities.rank_zero:`Trainer.fit` stopped: No training batches.

  #261 opened a year ago by hasakikiki
  2

• 有谁用过Mac Studio微调的

  #260 opened a year ago by xsailor511
  0

• 如何使用evaluate.py对测试集进行验证

  #236 opened a year ago by lawrencelxy
  4

• AttributeError: module 'inspect' has no attribute 'ArgSpec'

  #255 opened a year ago by SeekPoint
  1

• 修改max_seq_length好像并没有生效？

  #256 opened a year ago by tjulh
  4

• 怎么控制每训练n轮就保存一次模型呢

  #259 opened a year ago by tjulh
  1

• AttributeError: module 'torch.optim' has no attribute 'adam'

  #258 opened a year ago by evanweiguohua
  5

• 推理时怎么指定用哪几张卡

  #257 opened a year ago by tjulh
  2

• 您好，非常感谢您的工作。我进行全参数微调后，调用infer_finetuning.py的时候，出现Missing key(s) in state_dict: "_TransformerLightningModule__backbone.transformer.lm_head.weight".请问您有遇到过这个问题吗

  #252 opened a year ago by Xuan-ZW
  2

• 显示可训练参数数量问题

  #254 opened a year ago by xxll88
  0

• 缺省Lora训练显存消耗 60G

  #253 opened a year ago by is
  0

• 关于数据的instruction，input，output

  #230 opened a year ago by sanwei111
  3

• 模型训练只使用到了单个GPU

  #247 opened a year ago by GZJAS
  1

• ptuning v2 如何启动quantization_bit 4

  #245 opened a year ago by xxll88
  1

• 用单轮数据集。 p-tuning微调chatGLM之后出现的问题。

  #243 opened a year ago by SMR-S
  1

• 使用ptv2的方式进行finetune，总是OOM（正常微调、lora方式都没问题）

  #218 opened a year ago by lxw0109
  7

• 请问一下，mac系统装不了deep_training？

  #217 opened a year ago by WHJTC
  1

• ptv2显存不够？

  #232 opened a year ago by sanwei111
  11

• 运行lora的训练代码， int8=True ，推理时报错 RuntimeError: expected scalar type Half but found Float，这是什么原因？

  #239 opened a year ago by MathamPollard
  14

• should be load_sft_weight?

  #241 opened a year ago by HenryYuxuanWang
  1

• 单轮的QA回答的还不错，但是多轮对话时，就会出现瞎回答的情况。我感觉应该是多轮对话的history记录干扰了回答的效果？目前还不确定是不是这样的情况，有兄弟遇到类似的情况吗？交流一波啊！

  #242 opened a year ago by SMR-S
  0

• Lora int8微调，推理时出错

  #202 opened a year ago by crellian
  4

• 整体微调以后，领域内的知识记住了，但是问常规问题，比如你好，你叫什么，他也回答领域内的知识

  #206 opened a year ago by heiheiwangergou
  5

• 关于需要多少显卡资源

  #234 opened a year ago by sanwei111
  1

• deepspeed 和普通训练（lora ptuning） batch_size 只能设置4以下 不然会OOM

  #233 opened a year ago by markWJJ
  21

• 执行infer_lora_finetuning.py报错：‘NoneType’ objectg has no attribute 'learning_rate'

  #238 opened a year ago by paizhongxing
  8

• input_ids格式是否需要<CLS>

  #237 opened a year ago by Jong-Won
  0

• Overriding torch_dtype=None with `torch_dtype=torch.float16` due to requirements of `bitsandbytes` to enable model loading in mixed int8. Either pass torch_dtype=torch.float16 or don't pass this argument at all to remove this warning.

  #235 opened a year ago by sanwei111
  0

• 单机两卡指令怎么样

  #231 opened a year ago by sanwei111
  2

• 关于数据格式

  #228 opened a year ago by sanwei111
  6

• RuntimeError: expected scalar type Half but found Float 你们在训练的时候遇到过这样的情况吗？

  #226 opened a year ago by SMR-S
  3

• 大佬 ，能讲一下如何合并lora权重到原来的模型中吗？

  #203 opened a year ago by cywjava
  5

• 设置 LoRa微调的 'target_modules' 后，运行报错 "AssertionError"

  #213 opened a year ago by ngbruce
  4

• 大佬好，请问关于scheduler

  #210 opened a year ago by IamRoBota
  4

• 加载lora模型出错～

  #223 opened a year ago by zlht812
  0

• 请问如何试用一般新闻语料对ChatGLM进行继续finetuing呢？

  #221 opened a year ago by yang9112
  1

• 第一次f16 lora双卡训练成功，第二次int8 lora单卡训练成功，第三次 换会f16 lora双卡训练失败，详情请进～

  #219 opened a year ago by zlht812
  3

• Lora推理2分30s正常吗？

  #216 opened a year ago by jikhunb
  2

• web/api_lora_demo.py 如何多张卡推理

  #220 opened a year ago by lxw0109
  0

• deepspeed如何设置可以避免OOM

  #209 opened a year ago by lianrzh
  2

• Deepspeed stage3保存模型权重维度为0

  #212 opened a year ago by Jong-Won
  2

• Lora训练后推理问题

  #215 opened a year ago by jikhunb
  2

• 大佬好，请问使用lora和ptv2进行微调分别需要修改哪些配置？

  #211 opened a year ago by mircop1t
  19

• 大佬好，请问下数据构造中的特殊token

  #208 opened a year ago by IamRoBota
  2

• 数据集

  #207 opened a year ago by renmengjie7
  0

• 训练数据集的q和a有长度限制吗，和max_seq_length是什么关系

  #205 opened a year ago by lancexiao
  0

• infer_lora_finetunning.py 报错

  #204 opened a year ago by philipyao
  4