| { |
| "bits": [ |
| 2, |
| 4, |
| 8 |
| ], |
| "group_size": 128, |
| "sym": true, |
| "data_type": "int", |
| "iters": 10, |
| "nsamples": 512, |
| "autoround_version": "0.9.2", |
| "dynamic": { |
| "+:.*model\\.layers\\.0\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.0\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.0\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.0\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.0\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.0\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.0\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.1\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.1\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.1\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.1\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.1\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.1\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.1\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.2\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.2\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.2\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.2\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.2\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.2\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.2\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.3\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.3\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.3\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.3\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.3\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.3\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.3\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.4\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.4\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.4\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.4\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.4\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.4\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.4\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.5\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.5\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.5\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.5\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.5\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.5\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.5\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.6\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.6\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.6\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.6\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.6\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.6\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.6\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.7\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.7\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.7\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.7\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.7\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.7\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.7\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.8\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.8\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.8\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.8\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.8\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.8\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.8\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.9\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.9\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.9\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.9\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.9\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.9\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.9\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.10\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.10\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.10\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.10\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.10\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.10\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.10\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.11\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.11\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.11\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.11\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.11\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.11\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.11\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.12\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.12\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.12\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.12\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.12\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.12\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.12\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.13\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.13\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.13\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.13\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.13\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.13\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.13\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.14\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.14\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.14\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.14\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.14\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.14\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.14\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.15\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.15\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.15\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.15\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.15\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.15\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.15\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.16\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.16\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.16\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.16\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.16\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.16\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.16\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.17\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.17\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.17\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.17\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.17\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.17\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.17\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.18\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.18\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.18\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.18\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.18\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.18\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.18\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.19\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.19\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.19\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.19\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.19\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.19\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.19\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.20\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.20\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.20\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.20\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.20\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.20\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.20\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.21\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.21\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.21\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.21\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.21\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.21\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.21\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.22\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.22\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.22\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.22\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.22\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.22\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.22\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.23\\.self_attn\\.q_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.23\\.self_attn\\.k_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.23\\.self_attn\\.v_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.23\\.self_attn\\.o_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.23\\.mlp\\.gate_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.23\\.mlp\\.up_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| }, |
| "+:.*model\\.layers\\.23\\.mlp\\.down_proj.*": { |
| "bits": 8, |
| "group_size": 128, |
| "sym": true |
| } |
| }, |
| "lm_head": false, |
| "provider": "auto-round", |
| "quant_method": "gptq", |
| "desc_act": false, |
| "true_sequential": false, |
| "damp_percent": 0.01 |
| } |
