THUDM/ChatGLM2-6B issues and pull requests

#681 - GLM-4 模型已经发布，欢迎使用

Pull Request - State: closed - Opened by zRzRzRzRzRzRzR 3 months ago

#680 - [BUG/Help] 用vllm 起INT4量化版本的模型报错类型不匹配 self_attention.dense.weight int4 shape [4096,2048] mismatch fp16 shape [4096, 4096]

Issue - State: open - Opened by yjjiang11 4 months ago

#679 - [BUG/Help] <title>运行api.py报错

Issue - State: open - Opened by cqray1990 4 months ago

#678 - [BUG/Help] <title>python api.py 后，点击链接报错405 Method Not Allowed

Issue - State: open - Opened by cqray1990 4 months ago

#677 - [BUG/Help] ModuleNotFoundError: No module named 'transformers.models.mpt'

Issue - State: closed - Opened by EinKung 4 months ago - 1 comment

#676 - [Help] 我在把项目fork下来之后，替换为我自己的接口，用axios不是fetch，我可以确定的是接口是正常的流式返回，但是在本地跑的时候，调用这个接口会显示一次性返回，不是流式输出，我看了头信息都是没有问题的，但是部署在服务器中是没有问题的，是因为什么呢，要怎样解决呢，希望得到作者或者同行的帮助，谢谢

Issue - State: open - Opened by wisdomlsh 4 months ago

#675 - [BUG/Help] <title>peft , deepspeed 下 p-tuning , AttributeError: 'NoneType' object has no attribute 'shape'

Issue - State: open - Opened by xxhh1212 5 months ago

#674 - [Help] <title>关于微调ptuning不能达到训练集的效果，且不破坏原有结构实验

Issue - State: open - Opened by Bingoyww 5 months ago

#673 - [BUG/Help] <title>python3.7 cli_demo.py 报错cannot set version_counter for inference tensor

Issue - State: open - Opened by mxldjt 6 months ago

#672 - test

Pull Request - State: open - Opened by SeanHH86 6 months ago

#671 - 运行python cli_demo.py报错

Issue - State: open - Opened by mxldjt 6 months ago

#670 - [BUG/Help]UnstructuredFileLoade 读取 txt 文件报错： zipfile.BadZipFile: File is not a zip file

Issue - State: closed - Opened by PedroNeal 6 months ago

#669 - [BUG/Help] <title>最后运行的时候出现这个问题，不知道怎么解决

Issue - State: open - Opened by Oraclty 6 months ago

#668 - Unable to load weights from pytorch checkpoint file for '/home/qianlab03/rjs/Langchain-Chatchat-0.2.7/chatglm2-6b/pytorch_model-00001-of-00007.bin

Issue - State: open - Opened by iaoxuesheng 7 months ago

#667 - [BUG/Help] 为什么chatglma2量化后weight的size会改变<title>

Issue - State: open - Opened by Paradise59 7 months ago

#666 - [BUG/Help] <title>按照官方给出的多轮问答数据集构建问答数据之后，运行脚本命令出现Traceback (most recent call last): File "/mnt/ChatGLM2-6B/ptuning/main.py", line 411, in <module> main() File "/mnt/ChatGLM2-6B/ptuning/main.py", line 229, in main train_dataset = train_dataset.map( File "/root/anaconda3/envs/GLM2/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 592, in wrapper out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs) File "/root/anaconda3/envs/GLM2/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 557, in wrapper out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs) File "/root/anaconda3/envs/GLM2/lib/python3.9/site-packages/datasets/arrow_dataset.py", line 3180, in map with Pool(len(kwargs_per_job)) as pool:

Issue - State: open - Opened by nevesaynever1 7 months ago

#665 - 请问现在大家目前有微调的效果比较好的方案吗，目前微调效果一直不理想。

Issue - State: open - Opened by nevesaynever1 7 months ago

#664 - [BUG/Help] <title>chatglm2首token时延增长随输入长度成倍快速增长

Issue - State: open - Opened by woaipichuli 8 months ago

#664 - [BUG/Help] <title>chatglm2首token时延增长随输入长度成倍快速增长

Issue - State: open - Opened by woaipichuli 8 months ago

#663 - [友情链接] <Colossal-AI加速ChatGLM2>

Issue - State: open - Opened by Yanjia0 8 months ago

#663 - [友情链接] <Colossal-AI加速ChatGLM2>

Issue - State: open - Opened by Yanjia0 8 months ago

#662 - [BUG/Help] cannot import name '_sentencepiece' from partially initialized module 'sentencepiece'

Issue - State: open - Opened by mericalandintent 8 months ago - 5 comments

#661 - Update README for NPU inference

Pull Request - State: open - Opened by wangshuai09 8 months ago

#661 - Update README for NPU inference

Pull Request - State: open - Opened by wangshuai09 8 months ago

#660 - [BUG/Help] <pytorch_model.bin don't exists>

Issue - State: open - Opened by sc-carson 8 months ago - 4 comments

#659 - 请问chatglm1、chatglm2、chatglm3训练了多少种语言呢？

Issue - State: open - Opened by fxb392 8 months ago

#659 - 请问chatglm1、chatglm2、chatglm3训练了多少种语言呢？

Issue - State: open - Opened by fxb392 8 months ago

#658 - [BUG/Help] <title>微调后为什么聊着聊着说我chatbot为空

Issue - State: open - Opened by LeoQianQY 8 months ago - 1 comment

#658 - [BUG/Help] <title>微调后为什么聊着聊着说我chatbot为空

Issue - State: open - Opened by LeoQianQY 8 months ago - 1 comment

#657 - [BUG/Help] <title>如何通过API来调用ChatGLM？

Issue - State: open - Opened by 20130216 8 months ago

#657 - [BUG/Help] <title>如何通过API来调用ChatGLM？

Issue - State: open - Opened by 20130216 8 months ago

#656 - [BUG/Help] <title> ptuning的时候可以自定义loss吗？

Issue - State: open - Opened by hhy150 8 months ago

#656 - [BUG/Help] <title> ptuning的时候可以自定义loss吗？

Issue - State: open - Opened by hhy150 8 months ago

#655 - [Help] chatGLM2-6B是否支持直接注入整篇文档进行微调，如果支持怎么处理

Issue - State: open - Opened by wangfenglei-hehe 8 months ago - 1 comment

#655 - [Help] chatGLM2-6B是否支持直接注入整篇文档进行微调，如果支持怎么处理

Issue - State: open - Opened by wangfenglei-hehe 8 months ago - 1 comment

#654 - [Help] 请问如何能做到微调过程中不保存早期的checkpoint

Issue - State: open - Opened by ybdesire 8 months ago - 3 comments

#654 - [Help] 请问如何能做到微调过程中不保存早期的checkpoint

Issue - State: open - Opened by ybdesire 8 months ago - 3 comments

#653 - [BUG/Help] <title>训练数据时显示ValueError: None is not in list，尝试过自己的json文件和AdvertiseGen提供的json文件

Issue - State: open - Opened by Siqi-c 9 months ago

#652 - [Help] 6b-int4 lora 微调使用 adam 优化器时梯度爆炸

Issue - State: closed - Opened by wizardforcel 9 months ago - 1 comment

#651 - 当我的transformers是4.36.2时，chatglm2不能正常加载

Issue - State: open - Opened by Congcong-Song 9 months ago - 2 comments

#651 - 当我的transformers是4.36.2时，chatglm2不能正常加载

Issue - State: open - Opened by Congcong-Song 9 months ago - 2 comments

#650 - [BUG/Help] Nonetype bug for build_stream_inputs

Issue - State: closed - Opened by Dinghow 9 months ago - 1 comment

#650 - [BUG/Help] Nonetype bug for build_stream_inputs

Issue - State: closed - Opened by Dinghow 9 months ago - 1 comment

#649 - how to transfer chatglm2-6b int4 model to npu device

Issue - State: open - Opened by woaipichuli 9 months ago

#649 - how to transfer chatglm2-6b int4 model to npu device

Issue - State: open - Opened by woaipichuli 9 months ago

#648 - [Help] 请问ChatGLM2-6B 目前是否通过中国相关部门的安全审查

Issue - State: open - Opened by fengbrute 9 months ago

#647 - [BUG/Help] <title>6b-chat基模型输出bug

Issue - State: open - Opened by HaomingX 9 months ago

#647 - [BUG/Help] <title>6b-chat基模型输出bug

Issue - State: open - Opened by HaomingX 9 months ago

#646 - [BUG/Help] <win10微调失败：ValueError: None is not in list>

Issue - State: open - Opened by wangyingdong 9 months ago - 2 comments

#645 - [Help] <title>用lora微调遇到这个问题RuntimeError: Expected to mark a variable ready only once.

Issue - State: open - Opened by wfllyzh 9 months ago - 1 comment

#645 - [Help] <title>用lora微调遇到这个问题RuntimeError: Expected to mark a variable ready only once.

Issue - State: open - Opened by wfllyzh 9 months ago - 1 comment

#642 - [BUG/Help] 求问在做基于ChatGLM2-6B的Ptuning v2的微调任务时损失/目标函数的公式是什么样子的

Issue - State: open - Opened by Magicsmx 9 months ago - 1 comment

#642 - [BUG/Help] 求问在做基于ChatGLM2-6B的Ptuning v2的微调任务时损失/目标函数的公式是什么样子的

Issue - State: open - Opened by Magicsmx 9 months ago - 1 comment

#640 - [BUG/Help] readme中清华云盘的模型文件版本不对，跑出来是乱码，需使用huggingface.co模型，麻烦更新

Issue - State: open - Opened by CN-zhouyk 10 months ago - 2 comments

#635 - [BUG/Help] <title>ptuning微调后想要在此基础上进一步微调

Issue - State: closed - Opened by fan-xh 10 months ago

#633 - [Help]有没有人尝试过把优化器改成SGD来减少显存占用的

Issue - State: open - Opened by 31-ryougishiki 10 months ago - 1 comment

#633 - [Help]有没有人尝试过把优化器改成SGD来减少显存占用的

Issue - State: open - Opened by 31-ryougishiki 10 months ago - 1 comment

#623 - [BUG/Help] ptuning微调之后向模型提问，返回空

Issue - State: open - Opened by 2021QKA 11 months ago - 5 comments

#621 - [BUG/Help] 单张A800 显存80g 跑chatglm2没问题，但是使用两张A40，一张A40显存48g，跑chatglm2报了torch.cuda.OutOfMemoryError: CUDA out of memory.

Issue - State: open - Opened by zhengdacheng 11 months ago - 1 comment

#608 - [BUG/Help] 请求大神对微调参数设置进行详解

Issue - State: open - Opened by YueleiFu 11 months ago - 3 comments

#600 - 如何改造原有模型并达到私有化使用。

Issue - State: open - Opened by TzyTman 11 months ago - 3 comments

#600 - 如何改造原有模型并达到私有化使用。

Issue - State: open - Opened by TzyTman 11 months ago - 3 comments

#596 - 运行web_demo.py，向GLM提问遇到JS错误

Issue - State: open - Opened by wrl1224 11 months ago - 4 comments

#594 - [Help] 长文本推理OOM

Issue - State: open - Opened by Wohoholo 11 months ago - 3 comments

#594 - [Help] 长文本推理OOM

Issue - State: open - Opened by Wohoholo 11 months ago - 3 comments

#593 - 【Help】使用lm-evaluation-harness评估，ChatGLM2-6B在CEval上准确率很低？

Issue - State: open - Opened by Kevin-KWH 11 months ago - 2 comments

#583 - [Help] lora微调合并后模型推理速度明显慢了好多

Issue - State: open - Opened by daydayup-zyn 11 months ago - 2 comments

#572 - lora微调-训练完直接预测得到的预测指标f1 ≠ 加载保存模型进行预测后得到的预测指标f1

Issue - State: open - Opened by Doufanfan 12 months ago - 3 comments

#572 - lora微调-训练完直接预测得到的预测指标f1 ≠ 加载保存模型进行预测后得到的预测指标f1

Issue - State: open - Opened by Doufanfan 12 months ago - 3 comments

#570 - [BUG/Help] windows11 chatglm2-6b-int4 量化版本 webui打开了，但是无法提交和回复

Issue - State: open - Opened by jhjade 12 months ago - 9 comments

#569 - 如何在web_demo中修改代码增加systemprompt呢？

Issue - State: open - Opened by BryanMurkyChan 12 months ago - 1 comment

#565 - [Feature] <title>6b-32k的多卡部署一直报错，但是2k的没问题，求助？

Issue - State: open - Opened by 300id 12 months ago - 2 comments

#565 - [Feature] <title>6b-32k的多卡部署一直报错，但是2k的没问题，求助？

Issue - State: open - Opened by 300id 12 months ago - 2 comments

#557 - [BUG/Help] <回答到一半就开始重复输出某个字符>

Issue - State: open - Opened by 1250681923 about 1 year ago - 5 comments

#557 - [BUG/Help] <回答到一半就开始重复输出某个字符>

Issue - State: open - Opened by 1250681923 about 1 year ago - 5 comments

#553 - 加载模型时报错，提示信息：Only Tensors of floating point and complex dtype can require gradients

Issue - State: open - Opened by Arkon2021 about 1 year ago - 3 comments

#535 - 为什么chatglm1和chatglm2的get_masks的实现不一样？即对应的attention_mask的实现方式不同

Issue - State: open - Opened by Doufanfan about 1 year ago - 12 comments

#535 - 为什么chatglm1和chatglm2的get_masks的实现不一样？即对应的attention_mask的实现方式不同

Issue - State: open - Opened by Doufanfan about 1 year ago - 12 comments

#521 - [BUG/Help] 使用LoRA微调之后，推理过程中卡住了

Issue - State: open - Opened by LuckyFanpu about 1 year ago - 2 comments

#521 - [BUG/Help] 使用LoRA微调之后，推理过程中卡住了

Issue - State: open - Opened by LuckyFanpu about 1 year ago - 2 comments

#520 - 微调时torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 报错不定时的好

Issue - State: closed - Opened by ZCQ0628 about 1 year ago - 1 comment

#506 - [BUG/Help] <chatglm-6b和chatglm2-6b 垂直领域Lora微调差异大>

Issue - State: open - Opened by LivinLuo1993 about 1 year ago - 2 comments

#506 - [BUG/Help] <chatglm-6b和chatglm2-6b 垂直领域Lora微调差异大>

Issue - State: open - Opened by LivinLuo1993 about 1 year ago - 2 comments

#495 - [BUG/Help] 微调之后模型部署，设置了多卡，但还是只用第0卡，显示内存不足

Issue - State: open - Opened by SebastianHan about 1 year ago - 2 comments

#494 - [BUG/Help] <title>dataclasses.FrozenInstanceError: cannot assign to field generation_max_length

Issue - State: open - Opened by leoluopy about 1 year ago - 9 comments

#464 - [BUG/Help] <title>

Issue - State: open - Opened by shnyyds about 1 year ago - 3 comments

#464 - [BUG/Help] <title>

Issue - State: open - Opened by shnyyds about 1 year ago - 3 comments

#458 - [BUG/Help] 在 Windows 上运行，加载了 Linux 共享库

Issue - State: open - Opened by wizardforcel about 1 year ago - 1 comment

#458 - [BUG/Help] 在 Windows 上运行，加载了 Linux 共享库

Issue - State: open - Opened by wizardforcel about 1 year ago - 1 comment

#443 - [BUG/Help] <各位大佬救一下>梯度回传出现了一个bug。

Issue - State: open - Opened by xtceh-yh about 1 year ago - 1 comment

#441 - 全参数精调 [launch.py:315:sigkill_handler] Killing subprocess

Issue - State: open - Opened by jakeywu about 1 year ago - 4 comments

#441 - 全参数精调 [launch.py:315:sigkill_handler] Killing subprocess

Issue - State: open - Opened by jakeywu about 1 year ago - 4 comments

#436 - [BUG/Help] <微调训练完毕发现output文件夹中只有几个json文件后续怎么使用>

Issue - State: open - Opened by wys2641970184 about 1 year ago - 5 comments

#431 - [Feature] <title>如何基于上次训练的结果使用其他的数据进行继续训练

Issue - State: open - Opened by tito-dt about 1 year ago - 5 comments

#431 - [Feature] <title>如何基于上次训练的结果使用其他的数据进行继续训练

Issue - State: open - Opened by tito-dt about 1 year ago - 5 comments

#426 - 请问是不是不支持AMD显卡？Torch not compiled with CUDA enabled

Issue - State: open - Opened by rtsbtx about 1 year ago - 3 comments

#426 - 请问是不是不支持AMD显卡？Torch not compiled with CUDA enabled

Issue - State: open - Opened by rtsbtx about 1 year ago - 3 comments

#414 - 旋转位置编码

Issue - State: open - Opened by xtceh-yh about 1 year ago - 2 comments

#407 - [BUG/Help] 有个疑问，模型刚开始运行的时候推理很慢，但是过一段时间（问一些问题或者两三个小时后）推理就很快了，这个是什么原因？

Issue - State: open - Opened by ToviHe about 1 year ago - 13 comments

#406 - Fix bug in openai_aip.py caused by Pydantic

Pull Request - State: closed - Opened by chzhyang about 1 year ago - 1 comment

GitHub / THUDM/ChatGLM2-6B issues and pull requests