FYI- after re-reading this, I realized that the last prompt actually had errors in the calculation of the score by using 1 instead of 2 for the hashtags. Sometimes rules are too complex for GPT-3 turbo 3.5... Regardless, the basic approach of leveraging the system to help debug is still useful.
Would love to know what you think about this post. My plan is to try to release another post in a week or two digging into the ways real-world transformers work - but meant to provide a fundamental understanding through examples, not just some deep math or formal diagrams....
FYI- after re-reading this, I realized that the last prompt actually had errors in the calculation of the score by using 1 instead of 2 for the hashtags. Sometimes rules are too complex for GPT-3 turbo 3.5... Regardless, the basic approach of leveraging the system to help debug is still useful.
Would love to know what you think about this post. My plan is to try to release another post in a week or two digging into the ways real-world transformers work - but meant to provide a fundamental understanding through examples, not just some deep math or formal diagrams....