recipe.yaml

quant_stage:
  quant_modifiers:
    GPTQModifier:
      ignore: [lm_head]
      config_groups:
        group_0:
          weights: {num_bits: 4, strategy: group, group_size: 128}
          targets: [Linear]