recipe.yaml

default_stage:
  default_modifiers:
    QuantizationModifier:
      config_groups:
        group_0:
          targets: [Linear]
          weights:
            num_bits: 4
            type: !!python/object/apply:compressed_tensors.quantization.quant_args.QuantizationType [
              int]
            symmetric: true
            group_size: 32
            strategy: group
            block_structure: null
            dynamic: false
            actorder: null
            scale_dtype: null
            zp_dtype: null
            observer: memoryless_minmax
            observer_kwargs: {}
          input_activations: null
          output_activations: null
          format: null
      targets: [Linear]
      ignore: [lm_head, 're:.*embed.*', 're:.*vision_tower.*', 're:.*vision_embedder.*', 're:.*audio_tower.*',
        're:.*router.*']
      bypass_divisibility_checks: false