search

the-full-stack / fsdl-text-recognizer-2022

Source of the FSDL 2022 labs, which are at https://github.com/full-stack-deep-learning/fsdl-text-recognizer-2022-labs
https://fullstackdeeplearning.com/course
MIT License
81 stars 26 forks source link

approx_conv_multiplications @ lab02 #76

Closed magomagic closed 1 year ago

magomagic commented 1 year ago

Hi ! I am reviewing the lab notebook and watching the course videos with great enjoyment. I have a question regarding the function approx_conv_multiplications(). It is not clear to me why num_input_channels is used in the calculation. It seems to me this dimension is already counted in num_kernel_elements(multiplication of all 3 dimensions of the kernel).

I suggest changing : mutliplications_per_kernel = num_spatial_applications num_kernel_elements num_input_channels to: mutliplications_per_kernel = num_spatial_applications * num_kernel_elements

BR, magomagic

charlesfrye commented 1 year ago

you're right! thanks for letting us know.