##SHAPE out_proj 27
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.45
    gpu__time_duration.sum                                                                   ms         1.42
    launch__occupancy_limit_registers                                                     block            2
    launch__occupancy_limit_shared_mem                                                    block            2
    launch__registers_per_thread                                                register/thread          201
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        21.89
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        48.83
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        62.47
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        16.32
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         4.60
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.19
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         1.67
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.39
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         0.94
##SHAPE out_proj 33
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         7.89
    gpu__time_duration.sum                                                                   ms         1.50
    launch__occupancy_limit_registers                                                     block            3
    launch__occupancy_limit_shared_mem                                                    block            2
    launch__registers_per_thread                                                register/thread          166
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        22.02
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        47.12
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        58.92
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        16.60
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         4.69
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.07
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         1.05
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.20
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.56
##SHAPE out_proj 30
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.57
    gpu__time_duration.sum                                                                   ms         1.39
    launch__occupancy_limit_registers                                                     block            3
    launch__occupancy_limit_shared_mem                                                    block            4
    launch__registers_per_thread                                                register/thread          168
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        23.82
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        51.20
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        63.69
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        24.10
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         6.34
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.44
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         2.13
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.60
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.57
##SHAPE out_proj 37
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         7.81
    gpu__time_duration.sum                                                                   ms         1.55
    launch__occupancy_limit_registers                                                     block            4
    launch__occupancy_limit_shared_mem                                                    block            3
    launch__registers_per_thread                                                register/thread          124
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        34.49
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        53.18
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        57.33
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        24.48
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         5.79
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.33
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         1.62
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.64
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.39
##SHAPE out_proj 32
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.78
    gpu__time_duration.sum                                                                   ms         1.37
    launch__occupancy_limit_registers                                                     block            2
    launch__occupancy_limit_shared_mem                                                    block            2
    launch__registers_per_thread                                                register/thread          121
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        36.98
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        53.80
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        64.76
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        32.64
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         7.75
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.90
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         2.66
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         1.00
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.32
##SHAPE out_proj 31
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.14
    gpu__time_duration.sum                                                                   ms         1.49
    launch__occupancy_limit_registers                                                     block            4
    launch__occupancy_limit_shared_mem                                                    block            4
    launch__registers_per_thread                                                register/thread          126
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        36.64
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        57.23
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        60.89
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        32.34
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         7.22
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.63
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         2.21
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.94
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.37
##SHAPE out_proj 34
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         7.01
    gpu__time_duration.sum                                                                   ms         1.71
    launch__occupancy_limit_registers                                                     block            4
    launch__occupancy_limit_shared_mem                                                    block            5
    launch__registers_per_thread                                                register/thread          117
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        30.35
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        76.17
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        51.80
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        32.33
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         6.11
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.45
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         1.86
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         1.25
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.00
##SHAPE gate_proj 27
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %        15.89
    gpu__time_duration.sum                                                                   ms         3.99
    launch__occupancy_limit_registers                                                     block            2
    launch__occupancy_limit_shared_mem                                                    block            2
    launch__registers_per_thread                                                register/thread          201
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        22.07
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        49.14
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        62.75
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        16.55
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         4.63
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.19
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         1.68
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.39
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         0.95
##SHAPE gate_proj 33
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %        14.74
    gpu__time_duration.sum                                                                   ms         4.27
    launch__occupancy_limit_registers                                                     block            3
    launch__occupancy_limit_shared_mem                                                    block            2
    launch__registers_per_thread                                                register/thread          166
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        22.03
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        47.06
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        58.74
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        16.62
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         4.70
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.07
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         1.05
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.21
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.55
##SHAPE gate_proj 30
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %        16.12
    gpu__time_duration.sum                                                                   ms         3.91
    launch__occupancy_limit_registers                                                     block            3
    launch__occupancy_limit_shared_mem                                                    block            4
    launch__registers_per_thread                                                register/thread          168
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        24.15
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        51.80
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        64.33
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        24.72
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         6.42
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.45
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         2.16
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.61
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.57
##SHAPE gate_proj 37
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %        14.87
    gpu__time_duration.sum                                                                   ms         4.34
    launch__occupancy_limit_registers                                                     block            4
    launch__occupancy_limit_shared_mem                                                    block            3
    launch__registers_per_thread                                                register/thread          124
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        34.86
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        53.71
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        57.77
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        24.82
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         5.81
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.32
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         1.64
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.65
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.40
##SHAPE gate_proj 32
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %        16.44
    gpu__time_duration.sum                                                                   ms         3.85
    launch__occupancy_limit_registers                                                     block            2
    launch__occupancy_limit_shared_mem                                                    block            2
    launch__registers_per_thread                                                register/thread          121
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        37.24
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        54.16
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        65.03
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        33.09
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         7.79
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.90
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         2.67
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         1.02
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.33
##SHAPE gate_proj 31
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %        16.61
    gpu__time_duration.sum                                                                   ms         4.05
    launch__occupancy_limit_registers                                                     block            4
    launch__occupancy_limit_shared_mem                                                    block            4
    launch__registers_per_thread                                                register/thread          126
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        37.45
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        58.46
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        62.07
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        33.01
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         7.20
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.58
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         2.28
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.97
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.37
##SHAPE gate_proj 34
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.46
    gpu__time_duration.sum                                                                   ms         4.83
    launch__occupancy_limit_registers                                                     block            4
    launch__occupancy_limit_shared_mem                                                    block            5
    launch__registers_per_thread                                                register/thread          117
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        30.61
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        76.72
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        52.12
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        33.01
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         6.19
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.46
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         1.88
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         1.27
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.01
