How to access the entities in NER.v2 in spacy-llma from logging or directly as they come from GPT-3 without using .ents ?

[innocent-charles]

i want to access and take the entities from logging or directly from gpt-3 without using .ents .

Example: consider the labels of PERSON and LOCATION . The text such as "The woman Paris was walking around in Paris"

1. from LLM logging
   DEBUG:spacy_llm:LLM response for doc: The woman Paris was walking around in Paris
   LOCATION: Paris
   PERSON: Paris

2. but from .ents
   LOCATION: Paris
   LOCATION: Paris

So, that means there is problem when taking response from LLM to Doc ...

now, i want a way to access the entities directly from LLMs or from logging maybe , i don't want to use .ents

[rmitsch]

Hi @innocent-charles, this looks like a bug in the spacy-llm NER task. We'll look into that as soon as possible.
As for the LLM response, the best way to access this is via your logs (you can define your logger to write to file).

[rmitsch]

I can confirm this is a bug. I'll update this thread once we implemented a fix and released a patched version of spacy-llm.

[svlandeg]

(Duplicate of #12812)