prepare_for_tokenization (Instruct template) is used for supervised training but not for inference (example)

McGill-NLP / llm2vec

Code for 'LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders'

https://mcgill-nlp.github.io/llm2vec/

MIT License

1.17k stars 88 forks source link

Closed bzantium closed 3 months ago

bzantium commented 3 months ago

Thanks for sharing code! I have some question:

following the code, instruct template is used for training but not used for inference (examples). don't need to use template for inference?
I am also curious how you use Instruction (including whether you use template or not) when evaluating on MTEB. If possible, can you share code to evaluate?