question about pre-computed LUT

[hiyijian]

hi @jiaxiang-wu ,
You teach us to quantize fully-connected layer weights via codebooks. But There is not any instruction in the paper for how the input S(m) could be quantized, So I am not every clear that how the LUT could be pre-computed. Would you pls to explain this more?

[jiaxiang-wu]

Dear @hiyijian ,

We do not quantize the input S(m). Instead, for a new-coming input S, we compute its corresponding look-up table (LUT) in the run-time and then use it for the approximate layer response computation. Please refer to Line 806-818 and 993-999 for the LUT computation in the convolutional and fully-connected layers, respectively.

Best,

[hiyijian]

Cool. Thank you.