[Question]: How do machines without GPU use embedding models to parse documents?

[Nuclear6]

Describe your problem

I built the ragflow project on a machine with only CPU. After running it, I found that the uploaded document could be parsed successfully using Tongyi Qianwen's embedding model. Is the embedding model calculated locally or remotely?

Task has been received.
Start to parse.
Extract Q&A: 642. 38 failure, line: 22,34,53...
Finished slicing files(642). Start to embedding the content.
Finished embedding(366.5834937430918)! Start to build index!
Done!

PS: I haven’t applied for Tongyi Qianwen’s API key either.

[KevinHuSh]

It's locally. We add the model to the docker image.

[Nuclear6]

You mean, the embedding model can be run in cpu? According to my previous experience, the model needs to make some changes and adaptations to run in the CPU.