I want to estimate the memory bandwidth of my neural network. In order to do this, I need to know the FLOPS required for an inference. Does this number depend on the library that I am using (e.g. ComputeLibrary, OpenBLAS)? In this case, how can I estimate it?
If you're using Keras, you could just use the patch in this pull request: https://github.com/fchollet/keras/pull/6203
Then just call print_summary() and you'll see both the flops per layer and the total.
Even if not using Keras, it may be worth it to recreate your nets in Keras just so you can get the flops counts. find more at SLC