| quant_stage: | |
| quant_modifiers: | |
| QuantizationModifier: | |
| config_groups: | |
| group_0: | |
| targets: ['re:.*mlp.gate_proj.*', 're:.*mlp.up_proj.*', 're:.*self_attn.k_proj.*', | |
| 're:.*self_attn.o_proj.*', 're:.*self_attn.q_proj.*', 're:.*self_attn.v_proj.*'] | |
| weights: | |
| num_bits: 4 | |
| type: float | |
| symmetric: true | |
| group_size: 16 | |
| strategy: tensor_group | |
| block_structure: null | |
| dynamic: false | |
| actorder: null | |
| observer: 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 | |
| observer: minmax | |
| observer_kwargs: {} | |
| output_activations: null | |
| format: null | |
| group_1: | |
| targets: ['re:.*mlp.down_proj.*'] | |
| weights: | |
| num_bits: 8 | |
| type: float | |
| symmetric: true | |
| group_size: null | |
| strategy: channel | |
| block_structure: null | |
| dynamic: false | |
| actorder: null | |
| observer: minmax | |
| observer_kwargs: {} | |
| input_activations: | |
| num_bits: 8 | |
| type: float | |
| symmetric: true | |
| group_size: null | |
| strategy: token | |
| block_structure: null | |
| dynamic: true | |
| actorder: null | |
| observer: null | |
| observer_kwargs: {} | |
| output_activations: null | |
| format: null | |
| targets: [Linear] | |
| ignore: [lm_head] | |