1 Pull request merged by 1 person

• Update docs

  #841 merged Aug 16, 2024

1 Pull request opened by 1 person

• Fix typo in chat.md

  #866 opened Aug 20, 2024

13 Issues closed by 3 people

• Qwen2/Qwen2-7B-Instruct 使用V100 FP16推理/训练出现大量的数值溢出(nan)问题

  #868 closed Aug 21, 2024

• 【DashScope API流式返回】token分词问题，湖北，或其他中文的前方加入空格会导致乱码。

  #864 closed Aug 20, 2024

• 量化 lora 微调后的qwen2-72b 为4bit的模型

  #764 closed Aug 18, 2024

• 推理出现无端端的英文，怎么处理呢

  #766 closed Aug 18, 2024

• After executing 'qwen2', I interacted with it by posing several queries. Upon exiting the session and attempting to relaunch 'qwen2', I encountered an error

  #767 closed Aug 18, 2024

• qwen2 1.5B 评测结果无法对齐

  #700 closed Aug 16, 2024

• Is there any officially marlin quantized model ?

  #760 closed Aug 16, 2024

• 4块4090部署推理性能问题

  #763 closed Aug 16, 2024

• Qwen2 How to use fastapi to encapsulate the stream output interface

  #762 closed Aug 16, 2024

• 想确定一个SFT的一个细节

  #853 closed Aug 16, 2024

• Qwen2 7B具备Function Calling能力吗？

  #847 closed Aug 16, 2024

• AttributeError: 'Qwen2ForCausalLM' object has no attribute 'merge_and_upload'

  #706 closed Aug 16, 2024

• qwen2-72b 回复乱码，不符合预期

  #856 closed Aug 16, 2024

14 Issues opened by 13 people

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

  #872 opened Aug 22, 2024

• [TGI] V100部署qwen2-7b推理服务问题

  #871 opened Aug 22, 2024

• How do you fine tune acceleration for qwen2?

  #870 opened Aug 22, 2024

• qwen2不支持兼容openai的function_call吗？

  #869 opened Aug 21, 2024

• Is SWA used in Qwen2 long context pretraining?

  #867 opened Aug 21, 2024

• 请问一下qwen2-gptq的量化做了哪些具体的量化，只优化权重weight？

  #865 opened Aug 20, 2024

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

  #863 opened Aug 19, 2024

• 修改Qwen2-1.5B-Instruct里config.json的num_key_value_heads

  #862 opened Aug 19, 2024

• qwen2 7b 微调行业的function calling工具

  #861 opened Aug 19, 2024

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

  #860 opened Aug 18, 2024

• 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

  #859 opened Aug 18, 2024

• 在使用swift框架对原始base模型qwen2_0.5b进行推理时重复生成

  #858 opened Aug 18, 2024

• 请问Qwen2会出4B模型吗

  #857 opened Aug 16, 2024

• 如何设置参数，让每次回答的答案结果一致

  #855 opened Aug 16, 2024

14 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• What is the difference between id_to_token() and decode() in tokenizer?

  #782 commented on Aug 16, 2024 • 0 new comments

• 新的moe模型使用vllm启动报错AttributeError: 'MergedColumnParallelLinear' object has no attribute 'weight'

  #230 commented on Aug 16, 2024 • 0 new comments

• 关于量化模型过程中，对qwen2-72b部分权重参数进行padded的一些问题。

  #787 commented on Aug 17, 2024 • 0 new comments

• gptq版本与vllm版本冲突问题

  #784 commented on Aug 17, 2024 • 0 new comments

• Qwen2-7B-Instruct's performance gaps in evaluations of the GSM8K and MATH datasets with official scores

  #709 commented on Aug 18, 2024 • 0 new comments

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

  #791 commented on Aug 19, 2024 • 0 new comments

• Qwen2-72B-ins-gptq-int4 非常喜欢用英文回答问题，即使指定了中文。

  #800 commented on Aug 19, 2024 • 0 new comments

• [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

  #852 commented on Aug 20, 2024 • 0 new comments

• 在fine tune Qwen2-7B-Instruct 保存时候错误

  #835 commented on Aug 21, 2024 • 0 new comments

• 比较好奇Qwen的中文token的形式

  #537 commented on Aug 21, 2024 • 0 new comments

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

  #820 commented on Aug 21, 2024 • 0 new comments

• Any plan for Qwen2 14B and Qwen2-32B?

  #482 commented on Aug 22, 2024 • 0 new comments

• FunctionCall 能力开源实现咨询

  #738 commented on Aug 22, 2024 • 0 new comments

• Update llama_factory.rst which might be more user-friendly for beginners

  #850 commented on Aug 19, 2024 • 0 new comments