ebit Question

[digbangbang]

If I want to do the 8 bit training, how to set the ebit in

SWALP/models/quantizer.py

Line 28 in 99c1f4d

def block_quantize(data, bits, mode, ebit=8, small_block="FC", block_dim="B"):

or I default set the ebit=8? And what about I want to do the 4 bit training?