I have some questions about some concepts, and they are not well solved through search engines and gpt, so I hope you can explain them to me. I don't understand "nn layout" and "nt layout" of neural network memory layout. There are two explanations, but I don't know which one is correct. One explanation is that "NN" stands for "numerator numerator",in NN layout, the layout of input and weight tensors are both "numerator". "Numerator" means that the channels in the tensor are placed in the first dimension. ; NT stands for "numerator denominator", In NT layout, the layout of the input tensor is "numerator", while the layout of the weight tensor is "denominator". This means that the channels of the input tensor are still in the first dimension, while the channels of the weight tensor are in the second dimensions" Another explanation is "nn stands for column-major layout and nt stands for row-major layout."
Hope someone can give me an accurate explanation, thank you!