Skip to content
ML Academy

DL, NLP and LLMs

Interview question

  • midlevel
  • 10 minsuggested time
  • 3ML engineer, Applied AI, Data scientist

Your neural network's training loss stays flat from the first epoch, and on a second run it becomes NaN after a few hundred steps. Walk me through how you debug it.

Tests: Systematic debugging of training: data and label checks, loss/output-layer mismatch, learning rate and numerical stability, sanity checks such as overfitting a tiny batch.

day 32day 33day 34

Sign in to practise: answers are saved before model answers are shown.