##SHAPE out_proj 27
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.43
    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                               %        48.84
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        62.48
    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         0.95
##SHAPE out_proj 50
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.33
    gpu__time_duration.sum                                                                   ms         1.44
    launch__occupancy_limit_registers                                                     block            2
    launch__occupancy_limit_shared_mem                                                    block            2
    launch__registers_per_thread                                                register/thread          198
    sm__inst_executed_pipe_lsu.avg.pct_of_peak_sustained_active                               %        21.85
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        48.67
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        62.36
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        16.32
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         4.61
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.12
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         1.65
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.37
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.06
##SHAPE out_proj 30
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.60
    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                               %        23.92
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        51.41
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        63.96
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        24.09
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         6.32
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.43
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         2.12
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.58
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.58
##SHAPE out_proj 51
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.50
    gpu__time_duration.sum                                                                   ms         1.40
    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                               %        23.67
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        51.12
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        63.29
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        24.17
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         6.38
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.36
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         2.15
    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         1.66
##SHAPE out_proj 32
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.77
    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                               %        36.94
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        53.75
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        64.70
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        32.64
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         7.76
    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 52
    dram__throughput.avg.pct_of_peak_sustained_elapsed                                        %         8.49
    gpu__time_duration.sum                                                                   ms         1.41
    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                               %        35.78
    sm__pipe_alu_cycles_active.avg.pct_of_peak_sustained_active                               %        52.78
    sm__pipe_tensor_op_imma_cycles_active.avg.pct_of_peak_sustained_active                    %        62.66
    sm__warps_active.avg.pct_of_peak_sustained_active                                         %        32.62
    smsp__average_warp_latency_per_inst_issued.ratio                                      cycle         7.95
    smsp__average_warps_issue_stalled_barrier_per_issue_active.ratio                       inst         0.95
    smsp__average_warps_issue_stalled_math_pipe_throttle_per_issue_active.ratio            inst         2.63
    smsp__average_warps_issue_stalled_not_selected_per_issue_active.ratio                  inst         0.95
    smsp__average_warps_issue_stalled_wait_per_issue_active.ratio                          inst         1.39
