Benchmark your device
Why it matters Everything in this course runs on your own computer. Knowing how fast it is tells you which lessons will train live and which will show a run that was trained in advance.
The test below multiplies two 200 by 200 matrices fifty times and measures how long it takes. That is 50 times 200 times 200 times 200 times 2, about 800 million FLOPs. Dividing FLOPs by seconds gives your device's speed in FLOPs per second.
FLOP: one floating-point multiply or add. A matrix multiply of an (n, k) by a (k, m) costs about 2 times n times k times m FLOPs.
Typical results in this in-browser Python: a phone or an older laptop does a few hundred million FLOPs per second, a recent laptop a few billion. Later lessons scale their live runs to your tier.
Words and symbols in this lesson
- FLOP
- one multiply or add; how the course measures the cost of a run
Run the benchmark. Then set flops_per_matmul to the FLOPs of one 200 by 200 times 200 by 200 multiply, using the formula above.
Spotted a problem in this lesson? Report it · Ask about this lesson
This exercise opens soon
Finish the lesson before it and this one unlocks. The whole course is free, no account needed, and you can read this lesson right now.
This browser is blocking cookies for learn.floati.life, so lessons you finished here cannot open in order. Allow cookies for this site and reload.