| 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: ['re:.*embed_tokens', 're:.*linear_attn[.]conv1d', 're:.*linear_attn[.]in_proj_a', | |
| 're:.*linear_attn[.]in_proj_b', 're:model[.]visual.*', 're:mtp.*', lm_head] | |
| bypass_divisibility_checks: false | |
| mappings: | |
| - smooth_layer: re:model.*layers[.](3|7|11|15|19|23|27|31)[.]input_layernorm | |
| balance_layers: ['re:model.*layers[.](3|7|11|15|19|23|27|31)[.]self_attn[.]q_proj', | |
| 're:model.*layers[.](3|7|11|15|19|23|27|31)[.]self_attn[.]k_proj', 're:model.*layers[.](3|7|11|15|19|23|27|31)[.]self_attn[.]v_proj'] | |
| activation_hook_target: null | |
| - smooth_layer: re:model.*layers[.](3|7|11|15|19|23|27|31)[.]self_attn[.]v_proj | |
| balance_layers: ['re:model.*layers[.](3|7|11|15|19|23|27|31)[.]self_attn[.]o_proj'] | |
| activation_hook_target: null | |
| - smooth_layer: re:model.*layers[.](0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|28|29|30)[.]input_layernorm | |
| balance_layers: ['re:model.*layers[.](0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|28|29|30)[.]linear_attn[.]in_proj_qkv', | |
| 're:model.*layers[.](0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|28|29|30)[.]linear_attn[.]in_proj_z', | |
| 're:model.*layers[.](0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|28|29|30)[.]linear_attn[.]in_proj_b', | |
| 're:model.*layers[.](0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|28|29|30)[.]linear_attn[.]in_proj_a'] | |
| activation_hook_target: null | |
| - smooth_layer: re:model.*post_attention_layernorm | |
| balance_layers: ['re:model.*mlp[.]gate_proj', 're:model.*mlp[.]up_proj'] | |
| activation_hook_target: null | |
| offload_device: !!python/object/apply:torch.device [cuda] | |
| duo_scaling: true | |
| n_grid: 20 | |