gnahzby

3 ms·

1. 1.▲
   Making Peace with LLM Non-Determinism(barryzhang.substack.com)

   3 points by gnahzby 3y ago|0 comments
2. 2.▲
   by gnahzby 3y ago

   This result really surprised us too, but it seems like we were not alone in observing this: https://medium.com/@gabrielgrinberg/openai-api-fine-tuned-gp... That being said, both accounts that we experimented on were pe
3. 3.▲
   Our humble attempt at “how much data do you need to fine-tune”(barryzhang.substack.com)

   149 points by gnahzby 3y ago|37 comments