{ "last_module_idx": 66, "measurement": { "lm_head.linear": null, "model.layers.0.mlp": [ { "accuracy": 0.8833979054501182, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8865493222286827, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9054690787666723, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9109644952573275, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9428038659848665, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9470967493559185, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9550143543042635, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9705445327256855, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9729948137935839, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9709343094574777, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.974377442347376, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9851849690863961, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9872344551902068, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9920319679536318, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9924392559026417, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9940520944563966, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9978678502926701, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.0.self_attn": [ { "accuracy": 0.9061428183003476, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9119440568120856, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9246013540970652, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9507813579157779, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9518208848802667, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9543663671142176, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9655375637506184, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9689379554045828, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9735736344989977, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9749555587768555, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9767004047569475, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9789730373181795, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9800506422394201, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9826432654732152, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9882551141475376, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9912237909279371, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9904633282046569, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9946639059405578, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9974662685080579, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.1.mlp": [ { "accuracy": 0.8764496226059764, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8787798756047299, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8921024297413073, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8957637171996267, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9522795614443327, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9580257534980774, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9644574491601241, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9763479013192026, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9788192212581635, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9768708505128559, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9806702717354423, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9886257695524316, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9902395817794298, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9934452489802712, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9938387317877067, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9951152440748716, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9972506389021873, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.1.self_attn": [ { "accuracy": 0.9189570953971461, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9249023889240465, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9346624424583033, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9588889071815893, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9601798528119138, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9619040740163702, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9731086053346333, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9755017945640966, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9774597694999293, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9785979343088049, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9805926508025119, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9817305709186354, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9832864450780969, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9845957850155077, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9905143819357219, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9921343506951081, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9928441926052696, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9955450351300993, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.998071669551887, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.10.mlp": [ { "accuracy": 0.9518351209791083, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9537634692694011, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9605840664160878, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9625753697596098, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9765061798848604, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.978369657930575, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9814129989398154, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9878309902391935, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9889092672812311, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9880508843221163, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9894840787900122, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9938881318820151, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9947663251506654, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9967160244521341, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9969017186055058, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.997515566254917, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991075695541344, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.10.self_attn": [ { "accuracy": 0.9653690080893667, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.965691089630127, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9717134362772891, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9773214130025161, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9822988917953089, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9824923355328409, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9901737398222873, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9905403687765724, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9910937249660492, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9914847235930594, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9912591051114233, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9916563112484781, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9929716712550113, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.993554306265555, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9962110727241165, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967800018034483, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9974512306875304, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983878449389809, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.99932814036545, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.11.mlp": [ { "accuracy": 0.951210153730292, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9524993739630047, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9599142200068423, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9620853725232577, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9753246197575017, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9771912113616341, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9803592045056192, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9872800182355078, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9883195601011577, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9874349212960193, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9889328511137712, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9936118074937871, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9944861644977018, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9965642769085733, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967478490189502, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9973834519715685, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990986841485688, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.11.self_attn": [ { "accuracy": 0.9653168132430628, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9657643500127291, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.971744860473432, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9782044887542725, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9825379926907388, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.983062533955825, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9904710070083016, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9906949400901794, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9914694679410834, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9918687602407054, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9913141735290226, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9918192107426492, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9931149639581379, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9935906207875201, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9962336973924386, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967874427767176, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.997503642777079, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.99840759956523, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9993458122113034, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.12.mlp": [ { "accuracy": 0.9501487863691229, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9514782773820978, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9589815986783881, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9612010623279371, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9747250629098791, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9766584399499392, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9798747928518998, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9869411948480105, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9880141255102659, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9871192386275843, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.988657575688864, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9934468437966547, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9943495859441004, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9964701996037835, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9966618140277109, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9973098029823679, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990719607786128, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.12.self_attn": [ { "accuracy": 0.962122126629478, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9636631796234533, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.970582620093697, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9763547307566592, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9815055273081127, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9814373176348837, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9895573564265904, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9900157004594803, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9909286420596274, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9913396082426372, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.990728738276582, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9913379283327806, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9927804117140017, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9934827508895021, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961044666798491, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967294715737042, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972596299883566, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983903550984043, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992656127402657, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.13.mlp": [ { "accuracy": 0.9478320761730796, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9492013297582927, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.957160937158685, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9594999457660475, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9735689320062336, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9755778296997673, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9789921217843106, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9863500532351042, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9874578549673683, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9865337986695139, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9881284613358347, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9931439302469555, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9940835570818499, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9963046935828108, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9965086402861696, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9971918740162724, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9990284293889999, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.13.self_attn": [ { "accuracy": 0.9588139653205872, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9610627826891447, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9676920583373622, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.974586778565457, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9790860132167214, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9796632669473949, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9887781864718387, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9890355687392386, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9899547970608661, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9904449338975706, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9899103570925561, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9905682935526496, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9920201921149304, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9926407611683795, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9956356451699608, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9962765492106739, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970217880449797, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.998163261695912, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992227606769455, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.14.mlp": [ { "accuracy": 0.9452493504474038, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.946703829263386, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9550439746756303, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.957535655874955, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9722571529840168, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9743637034767553, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9779573506430576, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.985658115462253, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9868286346134386, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9858664462440893, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9875384169189554, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9928045374782462, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9937915319674894, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.996124939698922, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9963349492142075, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9970517960426054, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989762077794263, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.14.self_attn": [ { "accuracy": 0.9612591078406886, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9618651490462453, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9676608694227118, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9746541600478323, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9797675735072086, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9803143692644019, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9890310238850745, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9893600438770495, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9902204306502091, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9906804287119916, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9900057072702207, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9909199562511946, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9921029288517801, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9926199465990067, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9956644142144605, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9963718676253369, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971076953960093, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9981364623496407, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992397349131735, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.15.mlp": [ { "accuracy": 0.9434961143292879, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9449662158363744, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9534293883725217, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9560049803633439, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.971324368527061, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9735085870090284, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9771876837077894, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9851355913438296, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9863631662569547, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.985383579605504, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9871181869193127, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9925555986793417, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9935800103764785, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9959801121761924, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9962079536757971, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9969503922682059, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.99893693106347, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.15.self_attn": [ { "accuracy": 0.9578982403403834, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9591458063376577, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9665130439557528, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9728524465309946, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9793557069803539, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9795406601930919, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9886367524925032, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.988847163947005, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9898819139129237, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9901980091082422, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9895568411601218, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9904219131720694, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9921476833130184, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9927685582324078, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9957539027458743, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.996334549627806, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970236375535789, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9982015459161055, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992033743642663, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.16.mlp": [ { "accuracy": 0.940074130108482, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9416762715891788, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9507655689590856, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9535166746691653, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9697655408005965, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9720546634573686, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9759444610068673, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9843449827871824, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9856071064346715, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.984613056245603, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9864283282505838, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.992168938643054, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9932406819180438, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9957653723264995, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9960151899017786, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967924977995848, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9988779623649622, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.16.self_attn": [ { "accuracy": 0.9572994112968445, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9585663299811513, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9662721470782631, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9737129619247035, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9782490306778958, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9785667485312411, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9878096062886087, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9881608141096014, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9891921658264963, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9896519176269832, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9890354722738266, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9897911376074741, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.991554919826357, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9922837560114107, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9954103003991278, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9961209430506355, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968949203428469, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9980672093990602, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.999177787894089, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.17.mlp": [ { "accuracy": 0.936371131947166, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9380353563710263, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9476052083467182, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9505739902195177, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9678008619107699, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9702306766259042, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9744001266203428, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9833069584871593, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9846881502553037, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9836268597527554, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9855675979664451, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9916789092515644, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9928150012304908, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.995501635890258, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9957674808408085, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9966030046343803, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9988115776918436, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.17.self_attn": [ { "accuracy": 0.9618973888848957, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.962596667440314, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9684828394337704, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9742837563941353, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9802437001153043, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9802198159067255, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9895149179195103, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9897193759679794, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9906598658938157, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.991061134557975, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9902756465108771, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.991060022460787, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9922492762929515, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9928579157904575, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9958244746452883, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9964344752462286, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972770857183557, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9982319209528597, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.999286080465505, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.18.mlp": [ { "accuracy": 0.9337121499212164, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9353958242817929, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.945396109631187, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9485103582081041, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.966415800546345, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.968952411099484, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9733318153180575, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9825637544456282, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9840197422002491, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9829113044236836, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9849309811466619, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9913090679206347, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9924929840000052, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9953020764024634, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9955798646337107, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9964584496460462, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.998752083041166, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.18.self_attn": [ { "accuracy": 0.9553326525186238, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9564111922916613, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9652397789453205, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.972625826534472, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9768472690331308, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9773016791594656, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9872793076854003, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9876007374964262, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9885056458021465, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9890744678283993, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9886636702637923, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.989373474528915, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9911882328359705, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9920469632274226, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9952953814675933, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9959285596483632, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9966838010831883, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.99797315464208, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.999113926271859, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.19.mlp": [ { "accuracy": 0.9287568581731696, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9305941870338038, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9414686905710321, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.944832350078382, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9639288814444291, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9666543697056017, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9713816862357291, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9812690019607544, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9828557528947529, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9816419199893349, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9838190078735352, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.990663554323347, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9919416331931165, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9949527245603109, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9952491502228537, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9962032202043032, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.998648793191502, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.19.self_attn": [ { "accuracy": 0.9545814959626449, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9552125240627088, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9623956084251404, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9690233406267668, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9778326985083128, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9778528762491125, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9867739708800065, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9876795779717596, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9884522290606248, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9890736584600649, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9888461420410558, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9893198554453096, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9911929525827107, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9918372042869267, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9951641724297875, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9958915479089084, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9966877196964464, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9979191994980762, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991139736222593, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.2.mlp": [ { "accuracy": 0.906403754886828, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9087058180256894, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9229802018717715, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9270363857871607, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9531012836255526, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9567055608096876, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9627114189298529, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9760607951565793, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9780555179244593, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9761908901365179, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9790711857770619, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9879037753531807, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9895905507238287, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9935485453982102, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9938408579481276, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9949916848226598, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983031204656551, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.2.self_attn": [ { "accuracy": 0.9264937576494718, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9296025225990697, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9382351134952746, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9530888544885736, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9611787796020508, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9611662971346002, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9722768538876584, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9734736492759303, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.976217037753055, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9773596823215485, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9784197399490758, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9808602474237743, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9812775442474767, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9832452127808019, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9892440830406389, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9909604442747015, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9909296467116004, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9961062846215147, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9976122522432553, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.20.mlp": [ { "accuracy": 0.9253437958265606, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9272625320836118, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9385454340984947, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.941956269113641, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9622319246593275, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9650857009385762, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9699900840458117, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9804942482396176, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.982119553967526, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9807889806596857, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9830686418633712, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9902325209818388, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9915676501236463, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9947452756919359, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9950281883540907, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9959947027658161, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985928756037825, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.20.self_attn": [ { "accuracy": 0.9541984106365003, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9550033933237979, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9637175892528734, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9716743670011821, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9772808896867853, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9775168566327346, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9870412286959196, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9874651604577115, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9885112219735196, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9890514408287249, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9883564789044229, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9893327661250767, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9909383364413914, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.991669279180075, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9950001153506731, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9958353042602539, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9966338625863979, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.997876616685014, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9991252135093275, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.21.mlp": [ { "accuracy": 0.9276682326668187, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9294288283900211, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9400426400335211, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9432508255306044, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.963226926954169, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9659861106621592, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9706572545202155, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9810147065865366, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9826072313283619, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9812972043689928, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9834934441666854, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9905039542599728, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9917851325712705, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9949046451794473, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9951675334259084, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9960808436337271, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9986429006645554, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.21.self_attn": [ { "accuracy": 0.9642315130484731, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9642753130511234, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.971040499837775, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9762993015741047, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9817238434364921, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9817482562441575, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9894069119503623, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9897073303398333, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9907361574863133, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9910403693977156, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.99079761144362, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9914982295350024, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9930070803353661, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9936581759860641, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9962060129956195, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967903149755377, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972757953954371, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.998452457354257, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992606439872792, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.22.mlp": [ { "accuracy": 0.9279777752725702, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9296961395364058, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9399789383536891, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9430589801386783, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.963240319176724, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9660046320212514, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9705552741100913, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9810442202969601, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9826441924822958, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9812906478580675, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9834989105400286, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9904996988020445, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9917881480957332, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9949137737092218, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9951672510881173, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9960513138457349, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9986442771593207, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.22.self_attn": [ { "accuracy": 0.9617143241982711, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.963699830205817, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9720779563251295, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9770187170881974, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9805767677332226, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9818394733102698, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9893172924455843, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9892968151130175, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9903302075047242, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9907148315718299, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9903613815182134, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9912145812260477, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9931700464926267, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9936903213199816, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961389887489771, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967917978371444, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.997142815668332, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9984645143543419, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992247114056035, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.23.mlp": [ { "accuracy": 0.9255238332246479, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9272877793563039, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9376695156097412, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9407624821913869, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9619345602236296, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9647840230088485, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9693861415511683, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9803631399807177, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9820277785000048, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9806354971308457, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9829037581619463, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9901687996952158, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9914890182645697, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9947433189341897, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9949970794351477, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9958875940034264, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985979564095798, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.23.self_attn": [ { "accuracy": 0.9626138273038363, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9648216711847406, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9731827472385607, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9779894555869856, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9820159830545124, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9820698938871685, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.989438754947562, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9897195406650242, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9906899074190542, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9910514754684348, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9905741042212436, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9915082689962889, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9930116589132109, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.993509412988236, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.996172492833514, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9968623077790988, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.997128080968794, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985064657306985, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992498290774069, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.24.mlp": [ { "accuracy": 0.9212467607698942, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9231670592960558, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.933893272751256, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9370474815368652, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9596807172423915, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9627174640956678, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9674712890072873, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.979232099495436, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9809910953044891, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9794974342772835, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9819097613033495, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9895920353500467, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9909946714576922, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9944384537245098, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.994704448078808, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9956220188423207, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985199597124991, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.24.self_attn": [ { "accuracy": 0.9606357279576754, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9624003987563283, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9717728683823034, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9763468096130773, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9804538127623106, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.980465325870012, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9888276724438918, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9893668991954703, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9902816007011815, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9905090371244832, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9900878040414107, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9910269089435276, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9929490254113549, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9935898223989889, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9962094814369553, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9968029044960675, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971180285670256, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9984858303673958, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992265553752843, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.25.mlp": [ { "accuracy": 0.9167571444260447, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9188368320465088, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9298872947692871, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9331428189026683, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9573253079464561, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9605662509014732, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9654919881569712, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9780169584249195, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9798834527793684, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9782911429279729, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.980859177677255, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9889806501175228, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9904724414411344, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.994107664023575, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.994392018961279, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9953396661501182, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9984306590141434, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.25.self_attn": [ { "accuracy": 0.952006132979142, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9521688191514266, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9650355389243678, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9696922678696481, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9744412914702767, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9755054163305383, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9861737709296378, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9869813432818965, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9880577145438445, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9883641279057452, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9881206303834915, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9889862772665525, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9916186066050279, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9923278728598043, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9954895843800745, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9961644066007513, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9965454759566408, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9982289748364374, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9990709125995636, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.26.mlp": [ { "accuracy": 0.9118387510902003, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.914072312806782, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9256593867352134, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9290546304301212, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9547841360694483, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9582102267365706, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9633557827849137, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9766802756409896, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9786588706468281, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9769865070518694, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9797087578397048, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.988315612077713, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9898945046098608, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9937474723709258, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.994056704405107, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.995047581038977, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983423254207561, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.26.self_attn": [ { "accuracy": 0.9483140173711275, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9511216000506753, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9631358322344328, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.969289315374274, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9748452054826837, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9754121287872917, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9858876451065666, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9863269438869074, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9875074193665856, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9876219559656946, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9874061424481241, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.988514750411636, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9908839174007115, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9918494899021951, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9951638179390054, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.995887286176807, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9963556496720565, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9980561076418349, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9990569528584418, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.27.mlp": [ { "accuracy": 0.9067065276597676, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9091139090688605, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9211623041253341, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9246824603331716, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.952180310299522, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9558175959085163, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9611523214139437, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9753106471739317, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9774032062605807, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9756537299407156, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9785449959729847, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9876441932038257, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9893184045427724, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9933767263826571, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9937068694516232, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9947343700026211, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.998233098144594, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.27.self_attn": [ { "accuracy": 0.9472288834421259, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9503778909382067, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9627158202623066, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9691977281319467, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9743928344626176, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9748328328132629, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9853913203666085, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.985830274067427, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9875727421359012, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9879731347686366, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9869731507803264, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.988221141068559, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9907781332731247, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9916744051795257, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.995062208097232, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9958086037322095, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9963553010633117, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.997989161038085, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9989986771619633, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.28.mlp": [ { "accuracy": 0.9017543980949804, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9043277000126085, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9168854638149864, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9205885749114187, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9496063527308012, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9534426331520081, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9589901033200716, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9739271151392084, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9761003240158683, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9743126615097648, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9773654561293753, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9869457169582969, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9887213910880842, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9929757322135725, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9933463978139978, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9944138813175654, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9981174255280119, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.28.self_attn": [ { "accuracy": 0.9421535981328863, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9443678291220414, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9583936898331893, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9645682761543676, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9706611601929915, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9711715171211645, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.984223230888969, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9847645336075833, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.986216476089076, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9865827732964566, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9858118891716003, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9869524350291804, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9900114857836774, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9909965333185697, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9946307410535059, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9954824769183209, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9959495479339048, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9977831695424882, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9989374619173376, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.29.mlp": [ { "accuracy": 0.8901600210290206, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8930423887152421, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9060871852071661, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9100886395103053, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9434056282043457, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9477950209065488, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9537127864988226, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.969878403764022, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9722384339884708, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9705027090875726, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9738137565161052, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9846308921512804, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9866677757940794, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9915462261752078, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9920110035883752, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.993074296336425, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9974979456318053, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.29.self_attn": [ { "accuracy": 0.9446054446069818, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.943989910577473, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9569224244669864, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.964503664719431, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.969432953156923, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.970762174380453, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9847956205669203, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9851772879299364, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.986428301585348, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9868048131465912, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9862075573519656, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9871006529582175, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9898145583115125, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9906066907079596, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9944483857405814, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9953088964286604, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.995966112143115, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9977400042116642, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9989420157906256, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.3.mlp": [ { "accuracy": 0.8810023132123446, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8842012254815352, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9008108251973203, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9056044879712557, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9402067096609819, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9448014874207347, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9518775720345347, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9692110889836362, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9718122325445476, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9695792449148077, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9732086815332112, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9845147509323923, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9866599503316378, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9916855771290628, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9921087834395861, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9934800757389319, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9977965266688874, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.3.self_attn": [ { "accuracy": 0.9283590442255923, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9319525706140619, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9414493974886442, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.952218843133826, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9625428350348222, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9635152001129954, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9754158807428259, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9765823891288355, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9784304515311593, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9792907943851069, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9800566262320468, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9814682916591042, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9844972428522611, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9852622982702757, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9907740981955278, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9922184834354802, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9921624182086242, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9966378604110918, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978560881787225, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.30.mlp": [ { "accuracy": 0.8651304872412431, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8729492237693385, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.888343610261616, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8930633193568179, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9324447543997514, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9383946217988667, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9451172728287547, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.964156172777477, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9671395232802943, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9647766288958097, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9691412699849982, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9815701779566313, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9842618248964611, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.989918841343177, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9905360890062231, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9917823044877303, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9971717543115741, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.30.self_attn": [ { "accuracy": 0.9277198691117137, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9284116092481112, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9504072666168213, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9585866206570676, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9642577328180012, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9648442487967641, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9801756174940812, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9810172727233485, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9827500440572438, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9832937482156252, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9823576478581679, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9838747805670688, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9879682079741829, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9890972085689244, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9934980269325407, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.994496388262824, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9950310176140383, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.997452759625096, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9986734045179266, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.31.mlp": [ { "accuracy": 0.8858021560468172, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8892394241533781, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9024233253378617, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9066664231450934, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9415236335051687, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9464694198809172, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9523357874468753, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.969417719464553, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.972082395302622, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9700709989196376, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9739150185334055, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9847441268594641, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9869535541848132, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.991601616928452, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9921956297598387, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.993426260979552, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9976804405450821, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.31.self_attn": [ { "accuracy": 0.9131183436042384, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9148389163770174, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.938181714007729, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9435132302735981, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.958246045990994, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9618019336148312, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9806641544166365, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9826531959207434, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9842631879605745, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9842346043963182, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9832207215459723, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9850406850639143, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9888538318245035, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.989916393631383, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.993947720057086, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9951134818164926, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9954859630057686, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9978021436223858, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9987402710279352, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.4.mlp": [ { "accuracy": 0.8519805732526278, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8631281727238705, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8639167986418071, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8641822840038099, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9678132628139696, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9762651543868216, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9770694406409013, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9906146534179386, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9928400838061383, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9896006599852913, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9939715352497602, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9958139826592646, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9965232171510395, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968462568756781, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9980765324282018, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983677434685984, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989921454536287, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.4.self_attn": [ { "accuracy": 0.8876561114662572, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.8866384405838816, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9073285617326435, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9235967397689819, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9351385141673841, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9385938142475329, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9540242113565144, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9472981754102205, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9632508001829448, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9647938138560245, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.966027906066493, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.970478217852743, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9760876360692476, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.976091615463558, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9879508120448965, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9895959669037869, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.990504945579328, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9952645615527504, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.997202813037132, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.5.mlp": [ { "accuracy": 0.9757358620041295, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9763630032539368, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9798647212354761, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9809185846855766, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9877069937555414, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9886357486248016, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9901599985988516, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9936516414347448, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9941844324532308, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9937456446258646, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.994489250998748, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968198692720187, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9972570022862208, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9982886557516298, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983829858859903, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9986823023738045, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9995495483377262, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.5.self_attn": [ { "accuracy": 0.9807123808484328, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9816359516821409, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9856763262497751, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9874477308047446, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9909503946178838, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9906754187847439, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9949038703190652, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9950581714510918, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9954444644482512, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9955743547332915, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9952920094916695, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9955447900452112, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968030770358286, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9970302362191049, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9982590996905377, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.99853630805094, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9985816471866871, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9993904668739751, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9996129685375643, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.6.mlp": [ { "accuracy": 0.9691252865289387, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9699519998148868, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9744979566649387, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9758490276964087, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9843793040827701, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9855615167241347, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.987525158806851, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9919421774776358, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9926079582226904, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9920522045148047, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9929967569677454, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9959579870889061, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9965119228551262, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.997825481977902, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9979426803949633, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9983299639272062, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9994246754305143, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.6.self_attn": [ { "accuracy": 0.9781461860004225, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9796922457845587, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9832466370181033, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9859028584078738, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.989264292936576, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9896293909926164, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9939317377774339, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.994273834714764, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9946085591065256, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9948185494071559, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9946171739383748, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9949616707469288, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9960499260770647, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.996342560570491, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978399510054212, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9981849040640028, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.998381615861466, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991434619418884, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9995703589857409, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.7.mlp": [ { "accuracy": 0.963045336698231, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.964055924039138, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.969702002249266, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9713662894148576, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9813484126015714, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9827594741394645, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9851650172158292, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9903798330771295, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9911655532686334, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9905132337620384, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9916405960133201, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9951783148081679, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9958383452735449, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.997406441522272, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9975462340210614, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9980276374048308, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9993185387728246, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.7.self_attn": [ { "accuracy": 0.9710417703578347, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9719340236563432, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9801154230770311, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9837093086619126, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9862640233416307, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9865341782569885, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9925894980367861, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.992693388148358, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9932613835523003, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9934948957280109, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9930840111092517, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9936519335759314, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9951011762022972, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9955841894996794, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9972971515043786, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9977379900060201, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9978984217894705, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9989965945286187, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9994875376455878, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.8.mlp": [ { "accuracy": 0.9589479534249556, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9600331124506498, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9664161079808286, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9683069586753845, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9792133836369765, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9807999526199541, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.983545228054649, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9892675876617432, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9901598770367471, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9894102466733832, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9906823344920811, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9946064290247465, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9953571608976314, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9970960148463124, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.997252456843853, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9978059228313597, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9992308949953631, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.8.self_attn": [ { "accuracy": 0.9685747152880618, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9693091888176767, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9765589927372179, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9817066474964744, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9844459656037783, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9851048824034239, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9915924511457744, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9916974839411283, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9924215721456628, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9926326635636782, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9923245844088102, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9927853636051479, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9942713803366611, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9947665247477984, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9968729136805785, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9973232401044745, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9977287410905487, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9987522885203362, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9994118274434617, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.9.mlp": [ { "accuracy": 0.9565224929859764, "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": 395461696, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9577053126535917, "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": 409224256, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9644053390151576, "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": 457272320, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9663760442482798, "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": 512977920, "up_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.3, 0.7 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9780701869412473, "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": 578397280, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9797516935750058, "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": 592872448, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9826065097984514, "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": 637454432, "up_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "3": 32, "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9886709734013206, "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": 728741472, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9896080509612435, "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": 739476480, "up_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9888320494639246, "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": 751543392, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 128, "5": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9901736425726038, "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": 766018560, "up_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9943166335946635, "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": 924689504, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 128, "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.995102803173818, "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": 939164672, "up_proj": { "bits": [ 6, 5 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "5": 32, "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9969337619841099, "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": 1069524064, "up_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9971066113365324, "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": 1110384896, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9976930088902775, "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": 1209999616, "up_proj": { "bits": [ 8, 6 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "6": 128, "8": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9991896297586592, "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": 1415520512, "up_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.layers.9.self_attn": [ { "accuracy": 0.9606086392151683, "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": 89665536, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.05, 0.95 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9610989250634846, "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": 92221440, "v_proj": { "bits": [ 3, 2 ], "bits_prop": [ 0.25, 0.75 ], "group_size": { "2": 64, "3": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9706604104293021, "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": 95758848, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9752700454310367, "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": 112272384, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9808770198571054, "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": 132913152, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 128, "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9808334378819716, "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": 132980224, "v_proj": { "bits": [ 4, 3 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "3": 64, "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9896436568937803, "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": 169613312, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9897406563947075, "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": 169745920, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9903282876077452, "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": 171195392, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9906995045511346, "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": 173563904, "v_proj": { "bits": [ 4 ], "bits_prop": [ 1 ], "group_size": { "4": 32 }, "scale_bits": 4 } }, { "accuracy": 0.990563342445775, "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": 174923264, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 64, "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9909965607680773, "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": 175750144, "v_proj": { "bits": [ 5, 4 ], "bits_prop": [ 0.1, 0.9 ], "group_size": { "4": 32, "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9932090287145815, "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": 179253248, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 64 }, "scale_bits": 4 } }, { "accuracy": 0.9935899310206112, "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": 181592064, "v_proj": { "bits": [ 5 ], "bits_prop": [ 1 ], "group_size": { "5": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9961749844645199, "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": 220469248, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9967315536188451, "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": 223535104, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9971129602115405, "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": 253499392, "v_proj": { "bits": [ 6 ], "bits_prop": [ 1 ], "group_size": { "6": 128 }, "scale_bits": 4 } }, { "accuracy": 0.9985295140037411, "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": 265838592, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 32 }, "scale_bits": 4 } }, { "accuracy": 0.9992404109063117, "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": 337385472, "v_proj": { "bits": [ 8 ], "bits_prop": [ 1 ], "group_size": { "8": 128 }, "scale_bits": 4 } } ], "model.norm.norm": null } }