I'm using CLIP immich-app/XLM-Roberta-Large-ViT-H-14__frozen_laion5b_s13b_b90k
The model is multilingual. There are clearly similar words in different languages. As far as I know, the model itself determines the language with which it is used, but how to explicitly set the language so that the model produces an accurate result, and how does the algorithm for searching for a result for a language work?
a small question addition. Let's say I have a powerful server and I can allow the heaviest possible model to get the most accurate results, then what kind of clip model will be for multilingual search?
Here is a git discussion about use language for searching
https://github.com/immich-app/immich/pull/13567
I'm sorry, but I somehow misplaced the post.
.