{ "measurement": { "model.layers.0": { "accuracy": 0.8987722396850586, "total_bits": 507787520, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "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": [ 0.1, 0.9 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "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.1": { "accuracy": 0.9002890586853027, "total_bits": 478689536, "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.2": { "accuracy": 0.9608855843544006, "total_bits": 456407296, "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": 0.9555550813674927, "total_bits": 456407296, "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": 0.9499313831329346, "total_bits": 456407296, "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": 0.94085693359375, "total_bits": 456407296, "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": 0.9383460283279419, "total_bits": 456407296, "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": 0.9287691116333008, "total_bits": 456407296, "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": 0.9277379512786865, "total_bits": 456407296, "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": 0.9248785972595215, "total_bits": 456407296, "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": 0.9232504367828369, "total_bits": 456407296, "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": 0.9223508834838867, "total_bits": 456407296, "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": 0.9154980182647705, "total_bits": 456407296, "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": 0.9131753444671631, "total_bits": 456407296, "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": 0.9053680896759033, "total_bits": 456407296, "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": 0.9041793346405029, "total_bits": 465844480, "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": [ 0.01, 0.99 ], "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": [ 0.1, 0.9 ], "scale_bits": 4 } }, "model.layers.16": { "accuracy": 0.9069504737854004, "total_bits": 478689536, "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.17": { "accuracy": 0.9034838676452637, "total_bits": 478689536, "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.18": { "accuracy": 0.9072866439819336, "total_bits": 507787520, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "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": [ 0.1, 0.9 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "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.19": { "accuracy": 0.9008588790893555, "total_bits": 478689536, "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.20": { "accuracy": 0.9053101539611816, "total_bits": 478689536, "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.21": { "accuracy": 0.9090385437011719, "total_bits": 478689536, "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.22": { "accuracy": 0.9069957733154297, "total_bits": 465844480, "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": [ 0.01, 0.99 ], "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": [ 0.1, 0.9 ], "scale_bits": 4 } }, "model.layers.23": { "accuracy": 0.9098095893859863, "total_bits": 478689536, "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.24": { "accuracy": 0.9100494384765625, "total_bits": 478689536, "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.25": { "accuracy": 0.9088339805603027, "total_bits": 478689536, "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.26": { "accuracy": 0.9072427749633789, "total_bits": 478689536, "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.27": { "accuracy": 0.9096179008483887, "total_bits": 478689536, "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.28": { "accuracy": 0.9028897285461426, "total_bits": 478689536, "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": 0.902318000793457, "total_bits": 513030400, "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": [ 0.4, 0.6 ], "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": [ 0.4, 0.6 ], "scale_bits": 4 } }, "model.layers.30": { "accuracy": 0.9191799163818359, "total_bits": 607664384, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "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": [ 0.1, 0.9 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128 }, "bits": [ 4 ], "bits_prop": [ 1.0 ], "scale_bits": 4 } }, "model.layers.31": { "accuracy": 0.9125514030456543, "total_bits": 607664384, "q_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "k_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "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": [ 0.1, 0.9 ], "scale_bits": 4 }, "up_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "gate_proj": { "group_size": { "4": 128, "2": 128 }, "bits": [ 4, 2 ], "bits_prop": [ 0.1, 0.9 ], "scale_bits": 4 }, "down_proj": { "group_size": { "4": 128 }, "bits": [ 4 ], "bits_prop": [ 1.0 ], "scale_bits": 4 } } } }