default_stage: default_modifiers: QuantizationModifier: config_groups: group_0: targets: ['re:.*self_attn\.(q|k|v|o)_proj$', 're:.*linear_attn\.(in_proj_qkv|in_proj_z|out_proj)$', 're:.*lm_head', 're:.*layers\.(56|57|58|59|60|61|62|63)\.mlp\.(gate|up|down)_proj$'] weights: num_bits: 8 type: float symmetric: true group_size: null strategy: channel block_structure: null dynamic: false actorder: null scale_dtype: null zp_dtype: null observer: memoryless_minmax observer_kwargs: {} input_activations: num_bits: 8 type: float symmetric: true group_size: null strategy: token block_structure: null dynamic: true actorder: null scale_dtype: null zp_dtype: null observer: null observer_kwargs: {} output_activations: null format: null group_1: targets: ['re:.*mlp\.(gate|up|down)_proj$'] weights: num_bits: 4 type: float symmetric: true group_size: 16 strategy: tensor_group block_structure: null dynamic: false actorder: null scale_dtype: null zp_dtype: null observer: memoryless_minmax observer_kwargs: {} input_activations: num_bits: 4 type: float symmetric: true group_size: 16 strategy: tensor_group block_structure: null dynamic: local actorder: null scale_dtype: null zp_dtype: null observer: minmax observer_kwargs: {} output_activations: null format: null targets: [Linear] ignore: ['re:.*visual\..*'] bypass_divisibility_checks: false