Used for benchmarking CPU and GPU.
Start with fib(35) and you can increase in smaller steps. ~5
You can use either pytorch or tensorflow. Pytorch is a newer implementation in this git and is the preferred method.
Start with a smaller matrix_size e.g. 10,000 and increase from there. Also start with moderate iterations (e.g. 100) and increase only if necessary.
You can find the results for Geekbench ML scripts here. The module can be downloaded from this link.