Open DWHNicholas opened 3 months ago
I already know that theram_tag_list_4585_llm_tag_descriptions.json
text is *233835 based on 458551**, which should be the embedding of the description corresponding to the tag. Is there any other document that describes how to generate this embedding or what model is used to generate the embedding?
I think I already know how to do it
@DWHNicholas Could you please explain briefly how?
In the fourth step of model training preparation, you need to download a
ram_plus_tag_embedding_class_4585_des_51.pth
file. Looking at the code, this file seems to be used for tag embedding. I read it in the code and saw that the shape is ([233835, 512]). I would like to know how this file is formed and how to form such a file when preparing my own data set for processing.