Quiz · 4 questions
🔧 Practical Understanding
Reading model cards, sizing GPUs, picking the right tool
Level 0Token
0 XP0/94 lessons0/10 achievements
0/120 XP to next level120 XP to go0% complete
Quiz
01How much GPU memory does a 70B-parameter model in FP16 need (weights only)?
Hint
2 bytes per parameter at FP16.
02What is the 'Chinchilla trap'?
Hint
It's about optimizing the wrong objective for production deployment.
03What throughput improvement does continuous batching typically provide?
Hint
It's the biggest single throughput multiplier in modern serving stacks.
04What should you try BEFORE fine-tuning a model?
Hint
The cheapest viable option always comes first.
Comments 0
🔔 Reply notifications (sign in)Sign in — Please sign in to comment.
No comments yet — be the first.