After a needed, fantastic beach vacation with family and good food, I jumped straight back into my goal of lifelong learning and earned NVIDIA's Building LLM Applications With Prompt Engineering certificate. There was a guided workshop where our instructor walked us through the Jupyter notebook for the certificate program, showing us the various examples they present before the final assessment. At the end of the all-day workshop, he walked through one possible solution, which, due to cloud issues, ended up not working on my instance. So now we're in the world where you can have the exact same code and it not work [Narrator: Always has been]. Thankfully, due to having to sit down and actually learn what Pydantic was instead of just hearing about it and nodding along like I knew what they were talking about, I was able to swap a JsonOutputParser for a PydanticOutputParser, since the LLM wasn't following the prompt closely enough for the looser parser to handle. This meant I had to do another workaround downstream to get the data back in the expected formats for the rest of the assessment, but after the main workshop I was able to get a stable cloud instance and complete it! I also met some great folks involved in robotics, so it was a great way to end my vacation and made coming back to Mississippi a little easier.
Leave a comment