Comment on The Smol Training Playbook: The Secrets to Building World-Class LLMsparentComments−matusp10moThe number of runs you can afford are not enough to perform Bayesian optimization. Count how many different options they explored in the text and take a guess how many samples you need to start modeling the hyperparameter space.
Comments
The number of runs you can afford are not enough to perform Bayesian optimization. Count how many different options they explored in the text and take a guess how many samples you need to start modeling the hyperparameter space.