I've found the flops_table_maker.py file . How can I use it to get the GFLOPs of my tuned networks? As I can see, the python script runs the cnet_ana mode, but I integrated the auto-tuning to the run_cnet mode.
usually, FLOPs are considered to be a property of the network, based on assuming 'normal' implementations of convolutions (as opposed to FFT/Winograd/etc).
I've found the flops_table_maker.py file . How can I use it to get the GFLOPs of my tuned networks? As I can see, the python script runs the cnet_ana mode, but I integrated the auto-tuning to the run_cnet mode.