
Symmetric INT8 quantization (Geron Ch 19, morie.fn grq8)
Source:R/geron_train_native.R
morie_geron_int8_quantization.Rds = max|x|/qmax, q = round(x/s) clipped; zero maps to zero exactly.
Examples
V <- c(1, 2, 3, 4, 5, 6, 7, 8)
morie_geron_int8_quantization(V)
#> $q
#> [1] 16 32 48 64 79 95 111 127
#>
#> $scale
#> [1] 0.06299213
#>
#> $dequantized
#> [1] 1.007874 2.015748 3.023622 4.031496 4.976378 5.984252 6.992126 8.000000
#>
#> $max_abs_error
#> [1] 0.03149606
#>
#> $snr_db
#> [1] 48.73785
#>
#> $bits
#> [1] 8
#>
#> $estimate
#> [1] 16 32 48 64 79 95 111 127
#>
#> $n
#> [1] 8
#>