Skip to main content

1-bit adam, mp=1, pp=2, ZeRO-1, 12 layers, 1 node

Created on February 24|Last edited on February 24

Section 1


5001k1.5k2kStep5101520
05001k1.5k2kStep2468101214
5001k1.5k2kStep1e+92e+93e+94e+9
Run set
8



Run set
8