quant_stage:
  quant_modifiers:
    QuantizationModifier:
      ignore: [lm_head, 're:.*gate$']
      config_groups:
        group0:
          targets: [Linear]
          weights: {num_bits: 8, type: float, symmetric: true, strategy: channel, observer: mse}
          input_activations: {num_bits: 8, type: float, symmetric: true, strategy: token,
            dynamic: true, observer: memoryless}