Hello i was wondering why im getting different results while doing inference with a model trained in alexeyAB network compared to my RP3 32Bit raspbian?
is it because the 32 bit/64 bit weight or
how the convolution is performed in each framework? <- NNPACK related
im getting for instance 90% for an class in AB
but 60% in DarknetNnpack
Hello i was wondering why im getting different results while doing inference with a model trained in alexeyAB network compared to my RP3 32Bit raspbian?
is it because the 32 bit/64 bit weight or how the convolution is performed in each framework? <- NNPACK related
im getting for instance 90% for an class in AB but 60% in DarknetNnpack