cal_kl_threshold¶
- paddle.fluid.contrib.slim.quantization.cal_kl_threshold. cal_kl_threshold ( hist, bin_width, bits ) [source]
-
Using the KL-divergenc method to get the more precise threshold.
- Parameters
-
hist (List) – The hist of the tensor.
bin_width (float) – The bin width for the hist.
bits (int) – The quantization bits.