Advice for document embeddings
#4
by
griffin
- opened
Should we provide an instruction or no? Thank you!
No need for an instruction for documents; Just put <|embed|>\n at the beginning
Can you give a quick example @Muennighoff trying to do document text classification into few categories and want to understand if embeddings using this model would be useful. The documents text are of variable lengths.
Does this example work for you: https://github.com/ContextualAI/gritlm?tab=readme-ov-file#inference ?