Lightning AI Studios: Never set up a local environment again →

Log in or create a free account to track your progress and access additional course materials  

Unit 9 – Exercises

Exercise 1: Evaluating Mixed-Precision Performance

I recommend trying out the different mixed-precision choices on your GPU (if you have one):

  • regular float32 ("32-true")

  • regular float16 ("16-true")

  • regular float64 ("64-true")

  • mixed precision with float16 ("16-mixed")

  • mixed-precision with bfloat16 ("bf16-mixed")

For this, you can modify the precision argument of the Trainer in the DistillBert model from lecture 9.1here.

Please also fell free to share your results in the discussion here — I’d be interested to hear what you find on various GPU architectures!

Log in or create a free account to access:

  • Quizzes
  • Completion badges
  • Progress tracking
  • Additional downloadable content
  • Additional AI education resources
  • Notifications when new units are released
  • Free cloud computing credits
Watch Video 1

Unit 9 Exercises