default_stage:
  default_modifiers:
    AWQModifier:
      config_groups:
        group_0:
          targets: [Linear]
          weights:
            num_bits: 4
            type: int
            symmetric: true
            group_size: 32
            strategy: group
            block_structure: null
            dynamic: false
            actorder: null
            scale_dtype: null
            zp_dtype: null
            observer: mse
            observer_kwargs: {}
          input_activations: null
          output_activations: null
          format: null
      targets: [Linear]
      ignore: [model.embed_tokens, 're:.*block_sparse_moe[.]e_score_correction_bias$', 're:.*block_sparse_moe[.]gate$',
        lm_head]
      bypass_divisibility_checks: false
      mappings:
      - smooth_layer: re:.*input_layernorm$
        balance_layers: ['re:.*q_proj$', 're:.*k_proj$', 're:.*v_proj$']
        activation_hook_target: null
        balance_exponent: 1
      - smooth_layer: re:.*post_attention_layernorm$
        balance_layers: ['re:.*block_sparse_moe[.]gate', 're:.*w1$', 're:.*w3$']
        activation_hook_target: null
        balance_exponent: 1
      - smooth_layer: re:.*w3$
        balance_layers: ['re:.*w2$']
        activation_hook_target: null
        balance_exponent: 1
      offload_device: !!python/object/apply:torch.device [cpu]
      duo_scaling: true
      n_grid: 20