How can we extract user Embeddings and use them directly in tf serving instead of indices for DeepFM?

[tanmaygovil1999-sharechat]

[massquantity]

u want to extract user embeddings from DeepFM? I don't know how to do that, since the paper didn't provide such method.

[tanmaygovil1999-sharechat]

Okay Okay

[tanmaygovil1999-sharechat]

And If I have a very huge dataset then how can I train DeepFM with split of that dataset ?

[massquantity]

Huge data that can't fit into the memory? It is possible to use this library, but you need to write all the data processing by yourself. So I think it's not worth it.