Why the LogME is implemented not in torch but in numpy?

[Kyfafyd]

RT.
Maybe with CUDA, LogME costs less time? Or if this is the problem setting assuming the transferability evaluation can not access GPU? (but the feature extraction part needs GPU?)

[youkaichao]

You are welcome to open a pull request with CUDA implementation.

[Kyfafyd]

Thanks for response!
I would like to further learn if using torch implementation for comparing efficiency is fair?

[youkaichao]

You can compare torch naive implementation and torch accelerated implementation using our proposed algorithm of fixed point iteration.