Support some best-effort methods for some of the following: clustering, topic modeling, similarity calculation, and/or visualization (via ex. UMAP). These would be helper functions that operate directly on gobbli EmbeddingOutput.
Motivation
Would save some time for people looking to use embeddings in downstream tasks. Currently we don't do much besides pass along the embeddings.
Feature
Support some best-effort methods for some of the following: clustering, topic modeling, similarity calculation, and/or visualization (via ex. UMAP). These would be helper functions that operate directly on gobbli EmbeddingOutput.
Motivation
Would save some time for people looking to use embeddings in downstream tasks. Currently we don't do much besides pass along the embeddings.
Additional Details