default_stage:
  default_modifiers:
    QuantizationModifier:
      targets: ['re:.*self_attn\.(q|k|v|o)_proj$', 're:.*layers\.(1|2|3|10|11|12|13|14|15|19|20|23)\.mlp\.(gate|up|down)_proj$']
      ignore: ['re:.*language_model.*layers\.(?!(1|2|3|10|11|12|13|14|15|19|20|23)\.)\d+\.mlp\.(gate|up|down)_proj$',
        're:.*vision_tower.*', 're:.*embed_vision.*', 're:.*audio.*', 're:.*embed_tokens.*',
        're:.*per_layer.*']
      scheme: FP8_DYNAMIC
      kv_cache_scheme:
        num_bits: 8
        type: float
        symmetric: true
        group_size: null
        strategy: tensor
        block_structure: null
        dynamic: false
        actorder: null
        scale_dtype: null
        zp_dtype: null
        observer: static_minmax
        observer_kwargs: {}
      bypass_divisibility_checks: false