OpenMOSS/CoLLiE

Issues

• 如果想应用更多的模型，例如InstructBLIP，请问应该怎么修改呢？

  #181 opened 9 months ago by Listever
  1

• Zero 2 gets stuck when initializing optimizer states

  #149 opened 9 months ago by tengxiaoliu
  1

• 希望能支持 safetensors 格式的权重

  #157 opened 9 months ago by WillQvQ
  1

• trainer.py文件中保存peft_config时会出错

  #159 opened 10 months ago by Mr-nnng
  2

• 关于增加千问模型的支持

  #153 opened a year ago by Jieni05
  11

• No log information

  #163 opened 10 months ago by BeastyZ
  2

• The interpetation about the transposition operation when spliting weight to tensor parallel group

  #154 opened a year ago by SparkJiao
  2

• lomo训练65b llama实测 Lomo is incompatible with pipeline parallelism

  #152 opened a year ago by zlh1992
  1

• 能否增加一个从头预训练的例子？

  #151 opened a year ago by liujuncn
  1

• LOMO优化器使用梯度裁剪导致训练时间翻倍？

  #150 opened a year ago by Jieni05
  2

• Evaluating is too slow

  #121 opened a year ago by JinchaoLove
  2

• chatGLM2 使用张量并行报错

  #135 opened a year ago by BlueSkyyyyyy
  6

• 关于模型中断，重启的问题，怎么让模型继续训练

  #142 opened a year ago by 459737087
  6

• 支持 RLHF

  #145 opened a year ago by KYLN24
  0

• 你好，怎么让保存的模型能够分片，而不是保存一个几十G的大模型

  #143 opened a year ago by 459737087
  2

• 能不能重新训练啊？

  #127 opened a year ago by 459737087
  3

• RendezvousConnectionError，跑着跑着就有这个报错

  #141 opened a year ago by 459737087
  5

• AttributeError: 'PeftModelForCausalLM' object has no attribute 'set_cache'

  #140 opened a year ago by JiafeiSun
  2

• 关于 adalomo 没有 loss_scaler 只有 loss_scale 的问题

  #139 opened a year ago by HappyLynn
  1

• 训练loss为NaN

  #107 opened a year ago by fuqianya
  6

• 使用的Megatron-LM的版本

  #126 opened a year ago by liaosnow
  11

• 是否可以新增chatglm3 支持？

  #134 opened a year ago by hijeffwu
  3

• A100单卡跑llama2 finetune lora报错oom

  #138 opened a year ago by JiafeiSun
  2

• No module named 'collie.callbacks.pefts'

  #137 opened a year ago by JiafeiSun
  2

• chatGLM2 好像目前不支持ptuning训练，有计划什么时候支持么

  #136 opened a year ago by BlueSkyyyyyy
  2

• __init__() missing 'init_method' and 'config'

  #133 opened a year ago by yueg-security
  1

• AdaLomo optimizer step method

  #132 opened a year ago by winglian
  3

• [BUG] Evaluation 时使用并行可能不会完整地遍历一遍数据

  #119 opened a year ago by KYLN24
  1

• 张量并行流水并行可以和lora一起使用么？报错ValueError: Target module ColumnParallelLinearWithoutBias() is not supported. Currently, only `torch.nn.Linear` and `Conv1D` are supported.

  #131 opened a year ago by BlueSkyyyyyy
  3

• 是否可以增加baichuan-2 的fine-tuning支持? 或者 是否可以给一个 如何 新增微调模型的guide？ 谢谢

  #130 opened a year ago by wuchangping
  3

• collie和lomo不兼容

  #100 opened a year ago by LZY-the-boys
  3

• 使用数据类_ShardContainer遇到错误

  #123 opened a year ago by xuguohai
  1

• ImportError: cannot import name 'PeftConfig' from 'peft.utils'

  #115 opened a year ago by lisherlock
  7

• 该项目能否用于对模型进行二次预训练

  #120 opened a year ago by Zheng-Jay
  3

• [Feature] examples 里是否可以新增一个 internLM的用例？

  #124 opened a year ago by wuchangping
  1

• 基于CoLLie训练7B Moss模型，无法使用Huggingface的AutoModelForCausalLM加载吗？

  #114 opened a year ago by xuguohai
  3

• How to convert parallel state_dict to normal state_dict?

  #122 opened a year ago by JinchaoLove
  3

• [BUG] 使用 CollieDatasetForClassification 在 helm 风格下进行分类评测时，max new token 截取存在问题

  #118 opened a year ago by KYLN24
  0

• ColumnParallelLinearWithoutBias is not supported by peft

  #116 opened a year ago by JinchaoLove
  5

• Could Lomo class support `param_groups`?

  #117 opened a year ago by JinchaoLove
  0

• [问题]有关训练可视化

  #113 opened a year ago by RickMeow
  2

• [QUESTION]Multi-node multi-gpu training

  #110 opened a year ago by RickMeow
  2

• [BUG] ImportError: cannot import name 'PeftConfig' from 'peft.utils'

  #109 opened a year ago by RickMeow
  2

• Error： llama2 70B LlamaForCausalLM.from_pretrained 开启Zero3，会消耗大量内存导致 OOM

  #98 opened 2 years ago by xiaopqr
  5

• 替换tokenizer后载入报错

  #105 opened a year ago by 2793145003
  12

• lr_scheduler设置的问题

  #106 opened a year ago by YuxiangZhang0114
  1

• Llama2 70B 训练报错

  #104 opened 2 years ago by xiaopqr
  3

• 训练出错但没有报错信息

  #102 opened 2 years ago by 2793145003
  2

• tensor parallel + zero3 error

  #99 opened 2 years ago by LZY-the-boys
  1

• V100上执行examples/alpaca/train.py碰到错误No module named 'petrel_client，请问有人知道怎么解决吗

  #94 opened 2 years ago by JiafeiSun
  2