DankoZhang

minicpm-v采用W4A16量化，推理速度没什么变化

14

### Checklist - [x] 1. I have searched related issues but cannot get the expected help. - [x] 2. The bug has not been fixed in the latest version. ###...

### 是否已有关于该错误的issue或讨论？ | Is there an existing issue / discussion for this? - [X] 我已经搜索过已有的issues和讨论 | I have searched the existing issues / discussions ### 该问题是否在FAQ中有解答？ | Is there an...

### 是否已有关于该错误的issue或讨论？ | Is there an existing issue / discussion for this? - [X] 我已经搜索过已有的issues和讨论 | I have searched the existing issues / discussions ### 该问题是否在FAQ中有解答？ | Is there an...

[BUG] 为什么如何调整参数，显存占用都是接近80G？

3

### 是否已有关于该错误的issue或讨论？ | Is there an existing issue / discussion for this? - [X] 我已经搜索过已有的issues和讨论 | I have searched the existing issues / discussions ### 该问题是否在FAQ中有解答？ | Is there an...

[Bug]: for循环请求vllm进行推理，数据结果不一致

1

### Your current environment ### 🐛 Describe the bug A list contains multiple request texts, and when using a for loop to request inference from vllm, the following issues arise:...

[Docs] 在mlp1层之前，新增一个全连接层，应该怎么训练？

6

### 📚 The doc issue 在mlp1层之前，新增一个全连接层，应该怎么训练，视觉层和语言层都参与训练吗？辛苦帮忙解答下@czczup ### Suggest a potential alternative/fix _No response_

qwen2.5-vl语言模型初始化是使用的base版本还是instruct版本？

1

qwen2.5-vl语言模型初始化是使用的base版本还是instruct版本？

V2PE代码集成在哪里？

3

V2PE代码集成在哪里呀，没看到在哪里应用的

[Bug]: MilvusLite doesn't support nullable

3

### Is there an existing issue for this? - [x] I have searched the existing issues ### Describe the bug pymilvus==2.5.14 执行： `schema.add_field(field_name="configuration", datatype=DataType.JSON, description="配置", nullable=True) ` 报错： pymilvus.exceptions.MilvusException:

[Bug]: create json path index

2

### Is there an existing issue for this? - [x] I have searched the existing issues ### Describe the bug pymilvus==2.6.0 code is: error is: ### Expected Behavior _No response_...