QwenLM/Qwen2.5 issues and pull requests

#1055 - [Bug]: Use of the term "open source" to describe qwen when the training data is not open

Issue - State: open - Opened by phly95 11 days ago - 1 comment

#1052 - [Bug]: Model name error in vllm deployment

Issue - State: closed - Opened by JulioZhao97 12 days ago - 4 comments

#1049 - [Bug]: AttributeError: Model Qwen2ForCausalLM does not support BitsAndBytes quantization yet.

Issue - State: open - Opened by yananchen1989 15 days ago - 1 comment

#1040 - [REQUEST]:

Issue - State: closed - Opened by DAAworld 19 days ago - 1 comment

#1031 - [Bug]: vllm启动大模型，超过一定的上下文长度导致大模型回答答非所问

Issue - State: closed - Opened by Ave-Maria 24 days ago - 3 comments

#1023 - [Bug]: 在 4 卡 16GB V100 机器上采用 lmdeploy 部署 qwen2.5-32b-instruct-gptq-int4 模型，最高输出速度只有 80token/s ，请问这个速度正常吗？

Issue - State: open - Opened by SolomonLeon 30 days ago - 3 comments

#1015 - [Bug]: vllm 启动，openai的swarm 函数调用不正常

Issue - State: open - Opened by 18600709862 about 1 month ago - 2 comments

#1006 - [Bug]: Nvidia L20推理Qwen2.5 72B GPTQ-Int8模型不符合预期

Issue - State: open - Opened by renne444 about 1 month ago - 1 comment

#1005 - [Badcase]: 使用Qwen2-7b翻译中文，有额外的输出

Issue - State: open - Opened by cjjjy about 1 month ago - 1 comment

#998 - [Bug]: vllm部署后，用官方的例子调用报错openai.BadRequestError: Error code: 400 - {'object': 'error', 'message': "name 'Extension' is not defined", 'type': 'BadRequestError', 'param': None, 'code': 400}

Issue - State: open - Opened by 1gst about 2 months ago - 2 comments

#997 - [Bug]: 文档问答会忽略部分数据，比如证书号是12345 回答的是2345

Issue - State: open - Opened by daimashenjing about 2 months ago - 3 comments

#996 - docs: Add OpenLLM

Pull Request - State: open - Opened by Sherlock113 about 2 months ago

#994 - 关于function参数的格式问题

Issue - State: closed - Opened by XuyangHao123 about 2 months ago

#992 - [Badcase]: qwen2.5-72b 在昇腾910推理结果不符合预期

Issue - State: open - Opened by tianshiyisi about 2 months ago - 6 comments
Labels: help wanted

#991 - [Badcase]: 函数调用出现不正常token（iNdEx）

Issue - State: open - Opened by abiaoa1314 about 2 months ago

#990 - mlx-lm documentation is corrected

Pull Request - State: open - Opened by gringocl about 2 months ago

#986 - [Badcase]: Qwen2.5-72B-Instruct-GPTQ-Int4 input_size_per_partition

Issue - State: open - Opened by hyliush about 2 months ago - 5 comments

#985 - [Badcase]: Qwen2.5 14B Instruct can't stop generation

Issue - State: open - Opened by Jeremy-Hibiki about 2 months ago - 1 comment
Labels: enhancement

#957 - [Badcase]: qwen2.5 instruct 14B SFT后解码重复

Issue - State: open - Opened by 520jefferson about 2 months ago - 12 comments

#956 - [Question]: Can I use the QWEN2.5 model to generate sentence vectors?

Issue - State: closed - Opened by trundleyrg about 2 months ago - 3 comments

#945 - [Badcase]: Model inference Qwen2.5-32B-Instruct-GPTQ-Int4 appears as garbled text !!!!!!!!!!!!!!!!!!

Issue - State: open - Opened by zhanaali about 2 months ago - 12 comments

#935 - [Badcase]: 相同的数据，微调时在qwen2.5 72B预训练模型上的loss是qwen2 72B的3倍，请问2.5除了数据变多了，其他有什么不一样吗

Issue - State: closed - Opened by boundles about 2 months ago - 14 comments

#927 - [Bug]: The eos_token of the Qwen 2.5 base model is inconsistent between config.json and tokenizer_config.json.

Issue - State: closed - Opened by Songjw133 about 2 months ago - 9 comments

#921 - [Badcase]: Loss does not drop when using Liger Kernel at Qwen2.5

Issue - State: open - Opened by Se-Hun about 2 months ago - 2 comments

#918 - [Question]: qwen2.5 agent使用问题

Issue - State: closed - Opened by lonngxiang about 2 months ago - 8 comments

#890 - TensorRT Qwen2-72B-Instruct-GPTQ-Int4 can be converted and built into the engine normally, but the inference results are garbled. Have you ever encountered this?

Issue - State: open - Opened by tianzuishiwo 3 months ago - 3 comments

#888 - qwen模型可以用哪些gpu芯片？

Issue - State: closed - Opened by GhostISTA 3 months ago - 2 comments
Labels: inactive

#885 - About the vocabulary inconsistence

Issue - State: closed - Opened by patrick-tssn 3 months ago - 3 comments
Labels: inactive

#883 - how to finetuning qwen2 instruct model with long context

Issue - State: closed - Opened by ben-8878 3 months ago - 1 comment
Labels: inactive

#879 - Qwen2怎么扩展tokenizers词表？

Issue - State: open - Opened by LarryLong45 3 months ago - 4 comments

#876 - 您好，请问Qwen2-72B-Instruct公开榜单中以下评测集分别是采用多少shot评测的结果？

Issue - State: closed - Opened by 13416157913 3 months ago - 2 comments
Labels: inactive

#874 - ft qwen2的时候，flash attn 和core attn的输出相差较大，且attn_mask为false的token, flash attn输出的是全0向量，但core attn输出的是一个正常向量

Issue - State: open - Opened by seanM29 3 months ago - 9 comments
Labels: inactive

#872 - Qwen2通过transformer.pipeline调用时，怎么输出每个token的概率

Issue - State: closed - Opened by xin0623 3 months ago - 1 comment
Labels: inactive

#871 - [TGI] V100部署qwen2-7b推理服务问题

Issue - State: closed - Opened by charosen 3 months ago - 2 comments
Labels: inactive

#870 - How to fine tune qwen2?

Issue - State: closed - Opened by yangxue-1 3 months ago - 1 comment
Labels: inactive

#867 - [QUESTION] Is SWA used in Qwen2 long context pretraining?

Issue - State: closed - Opened by KKCDD 3 months ago - 2 comments

#863 - Parallel inference is wrong but single card inference is correct. (can't be solved by updating nvidia driver like #331)

Issue - State: closed - Opened by realCattleya 3 months ago - 5 comments
Labels: inactive

#860 - qwen1.5微调出错 ValueError: expected sequence of length 289 at dim 1 (got 291)

Issue - State: closed - Opened by lijiayi980130 3 months ago - 1 comment
Labels: inactive

#859 - 你好，我在自己的数据集上对qwen2-7b进行sft后，想在相同数据结构上进行评测，遇到了问题

Issue - State: closed - Opened by xiaomao19970819 3 months ago - 1 comment
Labels: inactive

#855 - 如何设置参数，让每次回答的答案结果一致

Issue - State: closed - Opened by xiangxinhello 3 months ago - 4 comments
Labels: inactive

#852 - [昇腾910B+AscendvLLM] 为什么Qwen2-7B-Instruct推理时总是在最前面加上一个链接

Issue - State: closed - Opened by zhufeizzz 3 months ago - 10 comments

#835 - 在fine tune Qwen2-7B-Instruct 保存时候错误

Issue - State: closed - Opened by JHaoGao 3 months ago - 3 comments
Labels: inactive

#826 - 请问qwen2的模型支持FP16部署吗，在用V100的推理过程中出现激活后数值的溢出的问题，请问qwen2支持v100推理吗？

Issue - State: closed - Opened by zhougekaibenchi 4 months ago - 4 comments
Labels: inactive

#820 - 使用官方提供微调工具finetune.py和finetune.sh,loss始终为0

Issue - State: closed - Opened by paulxs 4 months ago - 4 comments
Labels: inactive

#806 - 关于微调后qwen2-72B-instruct-int4-gptq后的压测和长文本测试

Issue - State: closed - Opened by lxb0425 4 months ago - 7 comments
Labels: inactive

#791 - Qwen1.5-MoE-A2.7B-Chat-GPTQ-Int4 模型加载时间过长（近 2 小时）

Issue - State: closed - Opened by TimVan1596 4 months ago - 3 comments
Labels: inactive

#717 - 运行支持的128K上下文后报错

Issue - State: closed - Opened by lonngxiang 5 months ago - 34 comments
Labels: inactive

#662 - Function calling support of openai style api for Qwen1.5 and Qwen2 model

Pull Request - State: open - Opened by Stephen-SMJ 5 months ago - 7 comments

#576 - 使用vllm部署qwen2-72b-instruct重复生成的问题

Issue - State: closed - Opened by Kk1984up 5 months ago - 10 comments
Labels: inactive

#387 - vllm本地部署Qwen1.5-72B-chat，在一次性生成长文本的任务中，出现超出字数限制的提示

Issue - State: closed - Opened by JSLW 6 months ago - 4 comments
Labels: enhancement

#368 - error on provided sample code: RuntimeError: CUDA error: device-side assert triggered

Issue - State: closed - Opened by Raychanan 7 months ago - 6 comments

#145 - [BUG] 运行openai_api.py报错，cli_demo.py和web_demo.py均正常启动

Issue - State: closed - Opened by 2277419213 9 months ago - 11 comments
Labels: inactive

#145 - [BUG] 运行openai_api.py报错，cli_demo.py和web_demo.py均正常启动

Issue - State: closed - Opened by 2277419213 9 months ago - 11 comments
Labels: inactive

#99 - [BUG] self.labels in SupervisedDataset

Issue - State: closed - Opened by gqc666 9 months ago - 1 comment

#98 - Fix model path in vllm.rst

Pull Request - State: closed - Opened by Michaelvll 9 months ago

#97 - [QA] Number of training tokens

Issue - State: closed - Opened by RicardoDominguez 9 months ago - 6 comments
Labels: inactive

#96 - Update index.rst

Pull Request - State: closed - Opened by ganeshkrishnan1 9 months ago

#96 - Update index.rst

Pull Request - State: closed - Opened by ganeshkrishnan1 9 months ago

#95 - Fix minor streamer typo

Pull Request - State: closed - Opened by osanseviero 9 months ago

#95 - Fix minor streamer typo

Pull Request - State: closed - Opened by osanseviero 9 months ago

#94 - Qwen1.5的模型参数名称和Qwen不同，考虑开放模型结构代码么？

Issue - State: closed - Opened by chenzhenbupt 9 months ago - 2 comments

#93 - qwen1.5沿用qwen1的lora代码和数据后，能力退化明显

Issue - State: closed - Opened by fanbooo 9 months ago - 8 comments

#93 - qwen1.5沿用qwen1的lora代码和数据后，能力退化明显

Issue - State: closed - Opened by fanbooo 9 months ago - 8 comments

#92 - qwen1.5-7b推理报错chat_stream

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

#92 - qwen1.5-7b推理报错chat_stream

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

#91 - Update README.md

Pull Request - State: closed - Opened by yijia2413 9 months ago - 1 comment

#90 - 使用ModelScope库pipeline方式加载模型报错

Issue - State: closed - Opened by chowsu 9 months ago - 2 comments

#90 - 使用ModelScope库pipeline方式加载模型报错

Issue - State: closed - Opened by chowsu 9 months ago - 2 comments

#89 - 有木有像Qwen一样，有自己的推理代码

Issue - State: closed - Opened by njhouse365 9 months ago - 2 comments

#89 - 有木有像Qwen一样，有自己的推理代码

Issue - State: closed - Opened by njhouse365 9 months ago - 2 comments

#88 - 问一下 Qwen1.5-72B-Chat 至少需要多少显存才可以跑呀，

Issue - State: closed - Opened by 1920853199 9 months ago - 6 comments

#88 - 问一下 Qwen1.5-72B-Chat 至少需要多少显存才可以跑呀，

Issue - State: closed - Opened by 1920853199 9 months ago - 6 comments

#87 - vllm中的例子跑不通，出现错误{"object":"error","message":"The model `Qwen/Qwen1.5-7B-Chat` does not exist.","type":"NotFoundError","param":null,"code":404}

Issue - State: closed - Opened by LHB-kk 9 months ago - 2 comments

#87 - vllm中的例子跑不通，出现错误{"object":"error","message":"The model `Qwen/Qwen1.5-7B-Chat` does not exist.","type":"NotFoundError","param":null,"code":404}

Issue - State: closed - Opened by LHB-kk 9 months ago - 2 comments

#86 - Qwen對繁體中文的識別及生成能力

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

#86 - Qwen對繁體中文的識別及生成能力

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

#85 - How well does this model support Traditional Chinese

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

#85 - How well does this model support Traditional Chinese

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

#84 - fine-tuning with SFTTrainer

Issue - State: closed - Opened by shao-shuai 9 months ago - 2 comments

#84 - fine-tuning with SFTTrainer

Issue - State: closed - Opened by shao-shuai 9 months ago - 2 comments

#83 - [BUG] HuggingFace Inference Endpoints 報錯

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

#83 - [BUG] HuggingFace Inference Endpoints 報錯

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

#82 - 请教一下，现在这个model.generate的方式怎么设置stop_words为“Observation:”呢？

Issue - State: closed - Opened by fataldemon 9 months ago - 2 comments

#82 - 请教一下，现在这个model.generate的方式怎么设置stop_words为“Observation:”呢？

Issue - State: closed - Opened by fataldemon 9 months ago - 2 comments

#81 - 关于仓库自带的Finetune代码target生成的疑惑

Issue - State: closed - Opened by TankNee 9 months ago - 3 comments

#81 - 关于仓库自带的Finetune代码target生成的疑惑

Issue - State: closed - Opened by TankNee 9 months ago - 3 comments

#80 - Can't find 'adapter_config.json' 用项目微调代码完成训练后，文件中包含有adapter_config.json但是一直提示找不到adapter_config.json。训练过程loss正常下降

Issue - State: closed - Opened by zwt0204 9 months ago - 5 comments
Labels: inactive

#79 - GPTQ 量化方法

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

#79 - GPTQ 量化方法

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

#78 - Qwen2: PAD token = EOS token?

Issue - State: closed - Opened by Qubitium 9 months ago - 2 comments

#77 - Qwen1.5-0.5b-chat 使用example中fintune.py 报错

Issue - State: closed - Opened by 128Ghe980 9 months ago - 5 comments

#77 - Qwen1.5-0.5b-chat 使用example中fintune.py 报错

Issue - State: closed - Opened by 128Ghe980 9 months ago - 5 comments

#76 - vllm的结果跟hf结果差距较大

Issue - State: closed - Opened by Nipi64310 9 months ago - 16 comments

#76 - vllm的结果跟hf结果差距较大

Issue - State: closed - Opened by Nipi64310 9 months ago - 17 comments

#75 - When I call the api '/v1/chat/completions' of FastChat API Server to access qwen1.5-72b-chat, it response incomplete results, but qwen-72b-chat response complete results

Issue - State: closed - Opened by coreyho 9 months ago - 2 comments
Labels: inactive

#75 - When I call the api '/v1/chat/completions' of FastChat API Server to access qwen1.5-72b-chat, it response incomplete results, but qwen-72b-chat response complete results

Issue - State: closed - Opened by coreyho 9 months ago - 2 comments
Labels: inactive

#74 - 运行API服务对话时提示：TypeError: 'NoneType' object is not iterable

Issue - State: closed - Opened by syusama 9 months ago - 5 comments

GitHub / QwenLM/Qwen2.5 issues and pull requests