diff --git "a/quant_strategy.json" "b/quant_strategy.json" new file mode 100644--- /dev/null +++ "b/quant_strategy.json" @@ -0,0 +1,5188 @@ +{ + "measurement": { + "model.layers.0": { + "accuracy": 0.9509583879262209, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.1": { + "accuracy": 0.8756404705345631, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.2": { + "accuracy": 0.5743598639965057, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.3": { + "accuracy": 0.833496299572289, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.4": { + "accuracy": 0.9792194842593744, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.5": { + "accuracy": 0.9784856701735407, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.6": { + "accuracy": 0.97777493391186, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.7": { + "accuracy": 0.9753335256827995, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.8": { + "accuracy": 0.9742659898474813, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.9": { + "accuracy": 0.9749326118035242, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.10": { + "accuracy": 0.9758923348272219, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.11": { + "accuracy": 0.9758641880471259, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.12": { + "accuracy": 0.9747861284995452, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.13": { + "accuracy": 0.9731597697827965, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.14": { + "accuracy": 0.9731135030742735, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.15": { + "accuracy": 0.9733343166299164, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.16": { + "accuracy": 0.9733786483993754, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.17": { + "accuracy": 0.9731195298954844, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.18": { + "accuracy": 0.9687351906904951, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.19": { + "accuracy": 0.9699929938651621, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.20": { + "accuracy": 0.9673012403072789, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.21": { + "accuracy": 0.9683234440162778, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.22": { + "accuracy": 0.9679232456255704, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.23": { + "accuracy": 0.9674058206146583, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.24": { + "accuracy": 0.9668316674651578, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.25": { + "accuracy": 0.9673871672712266, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.26": { + "accuracy": 0.9681136056315154, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.27": { + "accuracy": 0.9704282367601991, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.28": { + "accuracy": 0.9702738138148561, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.29": { + "accuracy": 0.9700905686477199, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.30": { + "accuracy": 0.9706572184804827, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.31": { + "accuracy": 0.9724596897140145, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.32": { + "accuracy": 0.9683455248596147, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.33": { + "accuracy": 0.9689904929837212, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.34": { + "accuracy": 0.9688115782337263, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.35": { + "accuracy": 0.9675803781719878, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.36": { + "accuracy": 0.9661333947442472, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.37": { + "accuracy": 0.9613336799666286, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.38": { + "accuracy": 0.959481391357258, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.39": { + "accuracy": 0.9561584184411913, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.40": { + "accuracy": 0.9494191433768719, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.41": { + "accuracy": 0.9478681846521795, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.42": { + "accuracy": 0.9439280114602298, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.43": { + "accuracy": 0.944461886305362, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.44": { + "accuracy": 0.9452667955774814, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.45": { + "accuracy": 0.9463562490418553, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.46": { + "accuracy": 0.9431603618431836, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + }, + "model.layers.47": { + "accuracy": 0.9348422677721828, + "total_bits": 2541146112.0, + "q_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "k_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "v_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "o_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_gate": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_gate_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_up_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + }, + "moe_expert_down_proj": { + "group_size": { + "4": 128 + }, + "bits": [ + 4 + ], + "bits_prop": [ + 1 + ], + "scale_bits": 4, + "scale_groups:": 32 + } + } + } +} \ No newline at end of file