2
$\begingroup$

I have a question regarding computing resource cost for training and testing of machine learning or deep learning algorithms. Any suggestion how can I calculate or extract resource cost like memory overhead and cpu overhead for t instance in time? If there is any formula that I can use or extract it from the computer I am using to train and test my algorithm

Any suggestion will be really helpful

$\endgroup$
1
  • $\begingroup$ Not really a formula, but Pytorch has a module named torchsummary that does this kind of calculation and display the result, I don't know exactly how it is done tho, but can help you check if your results are correct. The librayr is based on this article apprently, which can probably help answering your question $\endgroup$
    – Ubikuity
    Jun 25, 2021 at 8:41

1 Answer 1

1
$\begingroup$

The Floating Point Operations (FLOP) or Events Per Cycle (EPC) are representative measures of CPU activity.

Libraries like python-papi can extract many CPU events (low level or high level): http://www.bnikolic.co.uk/blog/python/flops/2019/09/27/python-counting-events.html

https://flozz.github.io/pypapi/papi_high.html

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.