{ "measurement": { "model.layers.0": { "accuracy": 0.8409156799316406, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.1": { "accuracy": 0.8796482086181641, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.2": { "accuracy": 0.9096565246582031, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.3": { "accuracy": 0.9143798351287842, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.4": { "accuracy": 0.8918952941894531, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.5": { "accuracy": 0.8808212280273438, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.6": { "accuracy": 0.86273193359375, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.7": { "accuracy": 0.8384037017822266, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.8": { "accuracy": 0.8897716999053955, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.9": { "accuracy": 0.917323112487793, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.10": { "accuracy": 0.9123013019561768, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.11": { "accuracy": 0.9101545810699463, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.12": { "accuracy": 0.9074265956878662, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.13": { "accuracy": 0.9024324417114258, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.14": { "accuracy": 0.9005353450775146, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.15": { "accuracy": 0.8948688507080078, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.16": { "accuracy": 0.8869843482971191, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.17": { "accuracy": 0.8833765983581543, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.18": { "accuracy": 0.8767848014831543, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.19": { "accuracy": 0.8756699562072754, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.20": { "accuracy": 0.8780655860900879, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.21": { "accuracy": 0.8770933151245117, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.22": { "accuracy": 0.882850170135498, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.23": { "accuracy": 0.8855595588684082, "total_bits": 435991584, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 } }, "model.layers.24": { "accuracy": 0.8869576454162598, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.25": { "accuracy": 0.8871769905090332, "total_bits": 435991584, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 } }, "model.layers.26": { "accuracy": 0.893949031829834, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.27": { "accuracy": 0.8951325416564941, "total_bits": 424064032, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.05, 0.95 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.01, 0.99 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 } }, "model.layers.28": { "accuracy": 0.9013614654541016, "total_bits": 446084128, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.2, 0.8 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.02, 0.98 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.2, 0.8 ], "scale_bits": 4 } }, "model.layers.29": { "accuracy": 0.9262962341308594, "total_bits": 578335776, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.5, 0.5 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.5, 0.5 ], "scale_bits": 4 } }, "model.layers.30": { "accuracy": 0.9261157512664795, "total_bits": 578335776, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.5, 0.5 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.35, 0.65 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.5, 0.5 ], "scale_bits": 4 } }, "model.layers.31": { "accuracy": 0.9445695877075195, "total_bits": 745583648, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.75, 0.25 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.75, 0.25 ], "scale_bits": 4 }, "v_proj": { "group_size": { "4": 128 }, "bits": [ 4 ], "bits_prop": [ 1.0 ], "scale_bits": 4 }, "o_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.75, 0.25 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.75, 0.25 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.75, 0.25 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128 }, "bits": [ 4 ], "bits_prop": [ 1.0 ], "scale_bits": 4 } } } }