File size: 198 Bytes
a4e6d8d
 
 
 
 
 
 
1
2
3
4
5
6
7
8
default_stage:
  default_modifiers:
    SmoothQuantModifier: {smoothing_strength: 0.8}
    QuantizationModifier:
      targets: [Linear]
      ignore: [lm_head, 're:.*mlp.gate$']
      scheme: W8A8