{ "last_module_idx": 58, "measurement": { "lm_head.linear": null, "model.layers.0.mlp": [ { "accuracy": 0.8551503733584755, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8598379837839227, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8839280605316162, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8917283007973119, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9287977281369661, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9343889198805156, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9452457553461978, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9619926120105543, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.965745957274186, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.963629170467979, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9682046206373918, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9814098195025796, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9840919438161349, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9896892332716992, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9905074277990743, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.992807644762491, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9970017828439411, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.0.self_attn": [ { "accuracy": 0.872567754042776, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8823489138954564, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9063533042606554, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9335484504699707, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9388456344604492, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9399561380085192, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9593330527606764, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9611869924946835, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9650190033410725, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9664848980150724, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.969404377435383, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9718834506837946, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9746822903030797, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9770710719259161, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9862766610948663, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9883006411163431, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9893960960601506, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9943787745739284, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.997219193922846, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.1.mlp": [ { "accuracy": 0.9227798173302099, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9245072540483976, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9414272935766923, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9494237962522005, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9599769774236178, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9634462500873365, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9726478739788658, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9781436810367986, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9806464832080038, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9793129312364679, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9820832249365354, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9893144792632053, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.990843296835297, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9938611407813273, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9944259383176502, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9963157204420943, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.997666702654801, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.1.self_attn": [ { "accuracy": 0.8716842500787032, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8809937552401894, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.896095545668351, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9361265584042198, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9379838077645553, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9363097015180086, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9615169042035153, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9595917086852224, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9630161272852045, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9646087508452565, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9678301873959994, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9693282936748705, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9717803691562853, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9736236804410031, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9846633217836681, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9871905391153536, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9892597198486328, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9933958210443196, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971564672887325, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.10.mlp": [ { "accuracy": 0.9838157656945681, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9842492640018463, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9865712617572985, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9871989347432789, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9918369329289386, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.992540097550342, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9934693508242306, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9958267031531585, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961585167207216, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9958061579810945, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9963585561827609, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978376903424138, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9981652302177328, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988227088592554, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9988736157354555, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990601021992532, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996357151111098, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.10.self_attn": [ { "accuracy": 0.9782408475875854, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.97889381176547, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9842756578796789, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9887617904888956, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.989045765839125, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9892924961290861, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9941115681278078, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9943847083731702, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9948016656072516, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.994959719479084, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9945887468363109, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9949968802301508, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961264984388101, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9964278952071541, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978683488933664, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9981991531033265, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9984593747281715, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991499045373577, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9995496110304406, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.11.mlp": [ { "accuracy": 0.9840074291354731, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9844504814398917, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.986640087868038, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9872575637541319, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9918207863443776, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9925837893235055, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.993482851668408, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9957813510769292, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961575583407754, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9957764129105368, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.996349739400964, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978157121099924, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9981549682193681, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988026874826142, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9988590331846162, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990416423074509, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996208244150406, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.11.self_attn": [ { "accuracy": 0.9758450169312326, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9766496027770796, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9823319080628847, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9876261878954736, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9877871651398508, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9880342146283702, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9933353952671352, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9936673331417536, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9941687732934952, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.994348433064787, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9939063548257476, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9943429341441706, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9955822707791078, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9959417126680675, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9975639960091365, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9979517712796989, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.998261318983216, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990371673514968, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9994957730860302, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.12.mlp": [ { "accuracy": 0.9851985981589869, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9856091549521998, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9876676970406583, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9882190415733739, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9924967312499097, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9931826269940326, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9940031072811076, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961641454382947, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9964833620347475, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961442527802367, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9966646064268915, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9980071759537646, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983170338367161, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9989094650863033, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989590642875746, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991196763554686, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996542973492882, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.12.self_attn": [ { "accuracy": 0.9790841890008826, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.979752579801961, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.984693535064396, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9887621198829851, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9895775733809722, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9898035079240799, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9943929789097685, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9946579356726847, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9950988767178435, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9952509289509371, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.994864499490512, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9952626126377206, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9963301679021433, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.996643944398353, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9979864543205813, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983100918562788, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9985283109310427, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992052433816226, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9995661783277204, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.13.mlp": [ { "accuracy": 0.9844254537632591, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9848623840432418, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9870634510328895, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9876844051637148, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9921610284792749, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9928113606415296, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9937138416265187, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.995988888175864, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9962923130706737, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9959810015402342, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9964975891144652, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9979355294061335, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9982373073304954, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988586830073282, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989193715155125, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990978593889036, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996355779125894, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.13.self_attn": [ { "accuracy": 0.9803181227884794, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9809341650260123, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9857936564244723, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9897391882381941, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9899517352643766, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9901490470296458, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9944556994657767, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9947093011517274, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9951247204291193, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9952619028718848, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9949321586050486, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.995287472088086, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9964269097698363, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9967059571492044, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9980490211593477, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983487158621612, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9985729956901387, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999235192901994, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.999580882979851, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.14.mlp": [ { "accuracy": 0.9860772380703374, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9864616848920521, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.988486102537105, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9890383348653191, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9929908623820857, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9935745534143949, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9943721451257405, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9964153056866244, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9966894094096986, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9964065340004469, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9968566672974511, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9981502472962204, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9984157432458902, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.998980385692496, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990302899754361, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991914158883063, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996801521629095, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.14.self_attn": [ { "accuracy": 0.9808931099741083, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9816874751919195, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9858728600175757, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9899958207419044, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9904109843467411, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.990623729793649, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9947962055080816, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9950441847506323, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9954036787936562, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9955612538676513, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9952593339901221, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9955744202199736, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9964462087342614, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9967483633050793, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9980571330770066, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983599695720171, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9986266044801787, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992132610396335, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9995927409219899, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.15.mlp": [ { "accuracy": 0.9872514699634752, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9875573390408566, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9894543941083708, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9899823234269494, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9935253325261568, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9940673481476935, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.994847507853257, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9966758528822347, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9969518439550149, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9966739067121556, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9971038620722922, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9982857168897202, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985376583706391, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9990565751336122, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991013692005685, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992538363623776, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996969536387998, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.15.self_attn": [ { "accuracy": 0.9822864014851419, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9828851521015167, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9875697313170684, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9905985664380225, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9909059773934515, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9913167922120345, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9952786458950293, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9955412549407858, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9958674178311699, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9959786640186059, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9957386517995283, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9960264955696306, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9969936319087681, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.997231237394245, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9983334584455741, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985943976789713, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9987670683154934, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999341957369133, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9996272477468378, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.16.mlp": [ { "accuracy": 0.9884794526978543, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9887415947098481, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9904860658080954, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9909568727016449, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9941757697808115, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.99466872215271, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9953735451164999, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970413307218176, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972739992173094, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970191783810917, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9973973669111729, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9984631037437602, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9986839089542627, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991605790626061, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991910622798299, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9993222646513268, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999732013963359, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.16.self_attn": [ { "accuracy": 0.9845266687242609, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9852389865799954, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9888263023213336, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9917292673336832, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9921190456340188, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9922679168613333, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9956558695749232, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9958921847374815, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9962282533708372, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9963297730213717, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.99615617763055, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9964118666554752, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972286549837965, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9974699218414331, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9984759858956462, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9987174326455907, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988737561200794, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9993985281570962, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9996593244569866, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.17.mlp": [ { "accuracy": 0.9881611466407776, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9884267567019713, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9902278443700389, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9907390883094386, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9939795273699259, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9944807487098795, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9952401412945044, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9969025864020774, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971705531995547, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9969145704648996, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9973048944222299, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9984102654025743, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9986379989667943, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991235346778443, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991631697079069, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9993047387290158, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9997114079612258, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.17.self_attn": [ { "accuracy": 0.9868619551784114, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9874944663361499, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9905772130740317, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9931664866836447, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.993445387404216, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9935436429161775, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9962665662169456, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9963943852405799, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9967346026709205, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968211507718814, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9967380522897369, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9969685842332087, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9976283431445298, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9978121583791155, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9986910971959955, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9988964315699903, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.999023636803031, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9994761177191609, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9996938497612351, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.18.mlp": [ { "accuracy": 0.9877698821456808, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9880565922511252, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9898872359802848, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9903955953685861, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9937738831105986, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9942932611233309, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9950670980309185, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9967925948532004, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970843413550603, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968092321957412, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9972194725353467, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9983610083398066, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985975476079866, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991023656176893, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991419379060206, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992885078842703, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9997144157164976, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.18.self_attn": [ { "accuracy": 0.9850055653797952, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9857045270894703, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9888458487234617, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9921295242874246, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9925100317126826, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9925613873883298, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.995758886400022, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9959327237386453, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9962544829437607, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9963795809369338, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9963000118732452, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.996529536811929, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971952028572559, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9974272345241747, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9984546367471155, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9986998365309677, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988977092465288, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9993740300108728, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9996593999921491, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.19.mlp": [ { "accuracy": 0.9878457988563337, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9881312509900645, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9899234097254904, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9904444805885616, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9937729404160851, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9942968566166727, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9950884384544272, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9967622229535329, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970697799795553, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968044544128996, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9972149412098684, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9983469957583829, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985894182402837, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9990823503191534, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991343711551867, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992839920481569, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996979553497544, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.19.self_attn": [ { "accuracy": 0.9870024786183709, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9874806192360426, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.990115360994088, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9929768247039694, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9933908071957136, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9934790255992036, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9962999267797721, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9964384844428614, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9967451713194972, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968544301625929, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9967365570758518, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9969614326561752, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9975394873242629, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9977418229376015, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9986436971904415, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9988557288521215, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9990233029974135, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9994365264612594, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.999688397284205, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.2.mlp": [ { "accuracy": 0.9624620487815455, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9633448437640542, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9692370452378926, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9709538786034835, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.980786825481214, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9824791880030381, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.985015680915431, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9901386679787385, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9910378197306081, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9901510422167025, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9914552436063164, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9949607307973661, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9957300997094104, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972940724539129, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9974171909454622, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9979157134106285, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992557957180237, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.2.self_attn": [ { "accuracy": 0.920105915320547, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9233990280251754, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9336702133479872, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9558533492841219, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9598596535230938, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9602910625307184, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9775633247275102, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9780966906171096, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9794871838469255, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9806128724625236, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9797062591502541, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.980877736681386, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9830918139532993, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9844322094791814, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9907696717663815, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9921560765881288, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9941235423872345, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9959898865536639, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.998433816981943, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.20.mlp": [ { "accuracy": 0.9882435320239318, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9884525733558756, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9903728828618401, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9909962909786325, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9938744730071017, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9943580694104496, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9952598371003804, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9967716253901783, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970964920756064, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968357850846491, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9972348070066226, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9983589673708928, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985889717936516, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9990816362398235, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991303706835759, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9993006718394003, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996910171191159, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.20.self_attn": [ { "accuracy": 0.9880510325494566, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9886099272652676, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9912934616992348, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9933161657107504, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9940949200015319, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9942927337006519, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9968043975532055, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.996915723736349, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9971775285115367, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972400545681778, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971456496339095, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9973595612927487, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9979407830457938, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9980998117672769, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988739415022888, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999032008706739, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991641511258326, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9995308922799794, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9997237485245263, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.21.mlp": [ { "accuracy": 0.9889092790453058, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9891268830550345, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9906887223846034, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9911591449850484, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9942973961955622, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9947555959224701, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9954592530664644, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970220146994841, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9973085189336225, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970671989415821, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9974336373178583, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9984828099412354, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9986945220122212, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991612760840278, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992005822591876, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999331483617425, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9997318196600598, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.21.self_attn": [ { "accuracy": 0.9886603708329954, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9890149620018507, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.991603957195031, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9934144020080566, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9941671020106265, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9943420985027364, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9967307781702593, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.996932818309257, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.997235206789092, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972997552862293, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971721854649092, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9973740744355478, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9979516197192041, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9981748379374805, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988899899548606, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990691023045465, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991530051552936, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9995557101639477, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9997266357960669, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.22.mlp": [ { "accuracy": 0.9878587103203723, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9880865396637666, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9898668684457478, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9904373939100065, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9936181295074915, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9942058297364336, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9950308380158324, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.996663072783696, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970109939183059, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.996730888360425, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9971575803662601, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9983055079846006, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985479470342398, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9990521883101839, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990981550593125, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999251782011829, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996845270016868, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.22.self_attn": [ { "accuracy": 0.987489214068965, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.98800380920109, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9908945599668905, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9927581854556736, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9935872009710262, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9938549877781617, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9966346845030785, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.996798393169516, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9970981323797452, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971636902344855, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970775285833761, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9972780533134937, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978501818289882, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9980428681561822, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988139357025686, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999004468713936, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991040400376445, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9995059594021816, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9997093204261833, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.23.mlp": [ { "accuracy": 0.9877935561694597, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9880442211502477, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9897527859399193, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9902731384101667, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9937465288921407, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9942311828857974, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9949941713559, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9967446976194256, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970434800182518, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9967908892584475, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9971816190763524, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9983437810289232, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985723364117899, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9990862375615459, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999130383722092, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992778285553581, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999712869034786, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.23.self_attn": [ { "accuracy": 0.9890999229330766, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9892960080974981, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9918333480232641, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9934967830777168, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9944275770532457, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9946118828497434, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9970541968941689, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9972345029052935, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.997476660891583, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9975678595273119, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9974062497678556, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9975824354118422, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9981173461205081, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9982891937619761, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9989589455684549, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991327565359441, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992510232780325, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999565220290893, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9997523254469821, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.24.mlp": [ { "accuracy": 0.9869130771411093, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9871772151244315, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9890457893672743, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9896249904444343, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9932923709091387, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9938037473904459, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9946420749551371, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9965031888139876, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968286364486343, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9965584744748316, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9969771949476317, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9982273786475784, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9984716979884788, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9990249164402485, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990722114514363, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992362166705885, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996983803514587, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.24.self_attn": [ { "accuracy": 0.9876369175158048, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.988058560772946, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9912115555060538, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9932340634496588, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9940412432739609, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9940707648271009, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9967256204078072, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9969071673327371, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9972198415351542, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9973158883421045, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971599500430258, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.997366879527506, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978398390506443, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9980452556751276, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988073611533955, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989989194038668, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991374419708001, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9995032237646612, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9997180630452931, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.25.mlp": [ { "accuracy": 0.984401495833146, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9847345446285448, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9869901749648546, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9877111370626249, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9919808514808354, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9926147209970575, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9936485016032269, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9957964569330215, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961956633548987, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9958734061372908, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9963858974607367, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978759090759253, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9981734150726544, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988248649200326, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.998888432587448, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990926372181428, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999632934083868, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.25.self_attn": [ { "accuracy": 0.9861161850000683, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9874627347055235, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9899361760992753, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9916268406729949, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9933682789928034, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9935254297758404, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9968159471295381, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9969706029484147, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9972899673800719, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9973326723434424, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9969360334308524, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9973960028667199, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9979121904624136, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9981071821560985, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988646061208687, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990246498859242, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991867571677032, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9994974652991483, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9997197875851079, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.26.mlp": [ { "accuracy": 0.9821977301647788, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.982567664824034, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.985359284438585, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9862501966325861, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9908842044441324, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9915934095257207, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9928525317656366, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9952326805183762, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.99568102822492, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9953280824579691, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9959003109680978, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9975953278572935, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9979281217644089, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9986692514074477, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9987462873717672, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.998997029407244, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9995868456500926, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.26.self_attn": [ { "accuracy": 0.9864152560108587, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9873253405094147, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9892045816308573, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9913245894406971, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.992800835716097, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9932177074645695, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.99687042793161, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9970677408732866, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9972945945827585, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9974050980649496, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972751203336214, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.997481620037242, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978277704825527, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9979777285142949, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9987857451564387, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989825961621184, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991769066178485, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9994609995107901, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9997193390797627, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.27.mlp": [ { "accuracy": 0.9708153605461121, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9716063079081083, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9764654824608251, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9777683443144748, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9859788966806311, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9871147784747576, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9889889688868272, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9927437046640798, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9933309672694457, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9927847730486017, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9936648038657088, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9962746591160172, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967933889282378, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9979418531844491, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9980357065796852, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9984118604150257, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9993405833742336, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.27.self_attn": [ { "accuracy": 0.987609841321644, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9895006386857284, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9918793628090307, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9940779154237948, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9949983966193701, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9950214338145758, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9971739492918316, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9972446533410173, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9975259300125273, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9976478840567564, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9975954486351264, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9977991943688769, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9980154221779421, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9981751730175394, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9989014021856221, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990740526271494, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992363317624519, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.999497045811854, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9997305549975288, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.3.mlp": [ { "accuracy": 0.9646574572512978, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9654105431155154, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9707881431830557, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9723484327918605, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9818792986242395, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9834145683991281, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9857454582264549, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.990659267494553, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9915256751211066, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9907073974609375, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9919057778621975, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9952424780318612, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9959593399574882, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9974518045783043, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9975659333561596, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9980401051671881, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992944179102778, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.3.self_attn": [ { "accuracy": 0.9298326153504222, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9322245560194317, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9460864757236681, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.963779025956204, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9649699832263746, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9652976832891765, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9810874462127686, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9816383656702543, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9828697035187169, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9835442398723803, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9823834456895527, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9834861865169123, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9864833888254667, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9874823485550127, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9926338925173408, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9937157160357425, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9950872440087167, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9969699778839162, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9986865834185952, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.4.mlp": [ { "accuracy": 0.9716180594343888, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9723043284918133, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9766769189583627, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9779153654449865, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9855069580831026, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9867823453325975, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9886340679306733, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9925479998714045, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9932303946269186, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9925618736367476, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9935428480568685, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961832871562556, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967676671712022, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9979460772715116, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9980368961236978, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9984082744309777, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9994190726919394, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.4.self_attn": [ { "accuracy": 0.9547656303957889, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9563453009254054, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9644939460252461, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9769298763651597, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9773167481547907, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9775890990307456, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9878329179788891, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9881612360477448, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9889677904154125, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9894385000592784, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9886170829597273, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9893262370636589, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9910532016503183, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9917344074500235, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9951126214705015, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9958435728361732, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968074124894644, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.997987865225265, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991304723447875, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.5.mlp": [ { "accuracy": 0.9760921330828416, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9767390207240456, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9803921894023293, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9814008364551946, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9879077492575896, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9889541448731172, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.990451864506069, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9938110309211832, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9943258154549097, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9938000989587683, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9946067415569958, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968148128766763, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9972990414029673, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9982763156294823, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983528611299238, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9986526262211172, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9995002830891233, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.5.self_attn": [ { "accuracy": 0.9602992189557928, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9617434677324797, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9695961098921927, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9796011855727748, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9800690741915452, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9803391930304075, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9892902021345339, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9896078917541002, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9902833260987934, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9906253814697266, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9900189473440772, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9906363150006846, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9923716375702306, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9929382408920088, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9958474181200329, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9964568603195643, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972100936268505, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983058128702013, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992333326096597, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.6.mlp": [ { "accuracy": 0.9779243077102461, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9785429992173847, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9819116561036361, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9828144390332071, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9888942092657089, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9898454546928406, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.991205989530212, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9943228105181142, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9947853821672892, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9942919807998758, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9950435589018621, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970686459227612, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9975149782472535, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9984170034723846, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.998484868654295, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9987567684760219, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9995304263736072, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.6.self_attn": [ { "accuracy": 0.9641681376256441, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9654102388181185, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9728116738168817, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9817448493681455, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9821139777961531, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.982375797472502, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9904281693069559, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9907899044061962, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9913757286573711, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9916823749479494, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9910833208184493, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9916438342709291, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9932228028774261, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.993744850942963, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.996317198794139, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9968590587377548, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9975071580786454, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9984972941640177, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9993076513178254, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.7.mlp": [ { "accuracy": 0.9787260905692452, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9793132512192977, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9826088939842424, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9835301433738909, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9892186891091498, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9901834726333618, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9915386657965811, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9944608674237603, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9949303022340724, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9944626626215483, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9951955370212856, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971510177772296, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9975840021905146, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9984375530559766, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985202493048028, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9988066905031079, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9995229864296945, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.7.self_attn": [ { "accuracy": 0.9654641716103805, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9666816215766103, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.973678287706877, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9823033088131955, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9826606763036627, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9829449794794384, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9907187058737403, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9910472262846796, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9916567810271916, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.991963552016961, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9913681094583712, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9919197739739167, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9934239430647147, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9939320816805488, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9964174243964647, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9969475116384657, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9975836214266325, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985395490534996, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9993256548615662, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.8.mlp": [ { "accuracy": 0.984861682904394, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9853437762511404, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9874602297418996, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9880280204509434, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9924239343718478, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9930739543939892, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9939045078660312, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961380813466875, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.996427637181784, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961161727183744, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9966213205143025, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9979970937496737, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983002686578977, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9989028972034392, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989546392309038, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991227283485626, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996535473160053, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.8.self_attn": [ { "accuracy": 0.9696946332329198, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.97079021366019, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9769786141420665, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9844727861253839, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9848995145998503, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9851816186779424, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9918018752022794, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9920754864027626, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9926364915935617, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9928825050592422, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9924738062055487, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9929672205134442, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9942889405708564, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9947150549606273, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968587352257026, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9973250033431932, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978325582648578, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9986629969391384, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.999381556616802, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.9.mlp": [ { "accuracy": 0.9841586618047011, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 169638736, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9846234101998178, "down_proj": { "bits": [ 6, 3, 2 ], "bits_prop": [ 0.05, 0.2, 0.75 ], "group_size": { "2": 64, "3": 64, "6": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 175405904, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.986896876441805, "down_proj": { "bits": [ 5, 3 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "3": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 195969536, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.987496303100335, "down_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 219857408, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9920921074716669, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 128, "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 248018656, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9927389041373604, "down_proj": { "bits": [ 8, 4, 3 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "3": 32, "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 254268416, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9936209085740542, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 }, "total_bits": 273355232, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9959667105423776, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 312411232, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9962593916999666, "down_proj": { "bits": [ 8, 4 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "4": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 317019136, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.995943482769163, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 128, "5": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 322238176, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9964520362646956, "down_proj": { "bits": [ 8, 5, 4 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "4": 32, "5": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 328487936, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9979090782764711, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 128, "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 396457696, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9982115189103704, "down_proj": { "bits": [ 8, 6, 5 ], "bits_prop": [ 0.05, 0.1, 0.85 ], "group_size": { "5": 32, "6": 32, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 402707456, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9988583456724882, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "6": 128, "8": 32 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 458487520, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989076986124641, "down_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.15, 0.85 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 476255488, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990909835813861, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 }, "total_bits": 518722816, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9996486816702312, "down_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "gate_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 606803200, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.9.self_attn": [ { "accuracy": 0.9750761468159524, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 53879808, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9759480843418523, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 55387136, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9815508425235748, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 58040192, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9871720177562613, "k_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 }, "total_bits": 67377920, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9873091264774925, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79827456, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9875987634847039, "k_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 }, "total_bits": 79861376, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.993054811107485, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101847552, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9933789490084899, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 }, "total_bits": 101947008, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9938766383811047, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 }, "total_bits": 102837504, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9940880003728365, "k_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 }, "total_bits": 104220672, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9936637729406357, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 }, "total_bits": 105027200, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9940908429654021, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 105458944, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9953291874967123, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 }, "total_bits": 108078080, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9957175678328464, "k_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 }, "total_bits": 109660160, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9974392181948611, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 }, "total_bits": 132911616, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9978406284200517, "k_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 }, "total_bits": 134825984, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9981928636368952, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 }, "total_bits": 152179200, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989664460483351, "k_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "o_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "q_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 }, "total_bits": 160843776, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9994824748780382, "k_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "o_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "q_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 }, "total_bits": 202510848, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.norm.norm": null } }