recipe.yaml

quant_stage:
  quant_modifiers:
    GPTQModifier:
      ignore: [language_model.lm_head, 're:vision_tower.*', 're:multi_modal_projector.*']
      sequential_targets: [MistralDecoderLayer]
      dampening_frac: 0.01
      config_groups:
        group0:
          targets: [Linear]
          weights: {num_bits: 4, type: int, strategy: group, group_size: 128, symmetric: true,
            actorder: weight, observer: minmax}