quant_stage:
  quant_modifiers:
    GPTQModifier:
      sequential_update: false
      dampening_frac: 0.1
      ignore: [lm_head]
      scheme: W8A16
      targets: Linear