Is it possible to generate semantic embeddings?
#12
by
michael-newsrx-com
- opened
Is there a way to get vector embeddings for cosine similarity ranking and KNN search?
Given that Falcon models are autogressive decoder-only models, they will produce embeddings that are subpar compared to state-of-the-art embedding models (e.g., E5 or Instructor)
FalconLLM
changed discussion status to
closed