-
Notifications
You must be signed in to change notification settings - Fork 217
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Draft of Power Profiling #170 #232
Conversation
…dd estimates for Conv
…or submission, Convolutions todo
…mbers. code cleanup
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hi @mlewandowski0 ,
thank you very much for your extensive contribution! Everything looks fine to me.
Just as a comment: for testing everything, we could first put it to a new branch before merging it with the master branch. What do you think, @jeshraghian ?
@mlewandowski0 , some tests are failing:
Could you check again? Be sure that all tests are passing. |
I was wondering if there is a way to compare a network with SNN layers to an equivalent network without any spiking layers in terms of energy consumption or computational complexity? |
related to issue #170
Hi,
I have implemented an initial version of power profiling. It was based on the torchinfo (the display style is the same as torchinfo, however, it reports the number of synapses, neurons, spiking and all events with some power estimations, simillarly to KerasSpiking ). Currently, I implemented :
snntorch/energy_estimation/estimate_energy.py
]snntorch/energy_estimation/device_profile.py
andsnntorch/energy_estimation/device_profile_registry.py
]snntorch/energy_estimation/layer_parameter_event_calculator.py
]To see how in current shape it is used, and what it looks like please see the
tests/test_energy_estimation.py
file.Please let me know if you think this implementation is good enough and what I can improve to make it part of the repository.
Best Regards