YidaHu comments

Results 17 comments of


                                            YidaHu

请问怎么转化成pb文件，输入层和输出层是什么呢

同问

Invalid markdown format in kubernetse/docs/V1EnvVar.md

Let me change that.

[bug] Question Classifier working wrong in the workflow process

> @crazywoola The DSL file here: [Customer Review Analysis Workflow .yml.txt](https://github.com/langgenius/dify/files/14955579/Customer.Review.Analysis.Workflow.yml.txt) @jiacheo I noticed that you have an extra blank space in this category, resulting in conditions that do not...

[bug] Question Classifier working wrong in the workflow process

> > > @crazywoola The DSL file here: [Customer Review Analysis Workflow .yml.txt](https://github.com/langgenius/dify/files/14955579/Customer.Review.Analysis.Workflow.yml.txt) > > > > > > @jiacheo I noticed that you have an extra blank space in...

[bug] Question Classifier working wrong in the workflow process

> I would close this issue now. > > Btw, I found this template from the cloud version( cloud.dify.ai ) also have the same 'blank space' behind the 'Negetive review'...

fix: duplicate text filtering

> ``` > exists_duplicate_node = self.text_exists(doc_id) > ``` > > it is used for check node is exist in vector db , the logic can't be removed. @YidaHu @JohnJyong I...

question ：文档向量化这个可以自己手动实现么？

> 现有公司级数据500G+，需要使用这个功能，请问如何手动实现这个向量化，然后并加载感谢。利用emebdding模块把数据向量化，修改vectorstores模块把向量存储到milvus中，向量检索直接从milvus中查询，最后把匹配出的文本作为上下文和问题一起添加到prompt中给LLM。这样应该能实现你的功能吧

Fine-tuning huge loss value

I also have the same situation, the loss is very large, and I am also using V100. May I ask if there is any solution? thx

Fine-tuning huge loss value

> I also have the same situation, the loss is very large, and I am also using V100. May I ask if there is any solution? thx The problem was...

知识库对长文的知识相关度匹配不太理想有何优化方向

我也遇到类似问题，求解决方案