52.16 Evaluating and Debugging Generative Al


This note contains only the slides, no personal notes.



Comparing model outputs

Model Registry

It is a central system of records for the models.

Training & Finetuning LLMs

Training from scratch

  1. Long & expensive training runs
  2. Expensive & difficult evaluations
  3. Monitoring is critical
  4. Ability to restore training from a checkpoint


  1. Efficient methods being developed
  2. Expensive & difficult evaluations


Also Read

Thoughts 🤔 by Soumendra Kumar Sahoo is licensed under CC BY 4.0