对yolov5模型中输出层的scale data有些疑问

问题1：地平线提供的yolov5模型中输出层的scale data有很长一串浮点数，官方的预训练模型转成后就只有一个，这里的scale data是训练还是什么阶段确定的呢？（我看代码这里和后处理相关，需要用到scale data，再通过DequantiScale函数计算score）

问题2：输出层的scale data和hbDNNQuantiScale中的scale data是一回事吗？
例如：
output[2]: name: 1312 valid shape: (1,21,21,255,) aligned shape: (1,21,21,256,) aligned byte size: 451584 tensor type: HB_DNN_TENSOR_TYPE_S32 tensor layout: HB_DNN_LAYOUT_NONE quanti type: SCALE stride: (451584,21504,1024,4,) scale data: 9.90397e-05,9.30927e-05,0.000121531,0.000153553,0.000157213,0.000240775,0.000115051,0.000179019,0.00010979,0.000108036,0.000123514,0.000116957,0.000116423,9.72861e-05,9.46176e-05,0.000115661,0.000103309,0.000113297,0.000123056,0.00013678,0.00011345,0.000135713,0.0001104,0.000102776,9.83535e-05,9.33977e-05,9.17204e-05,0.000128393,0.000112763,0.00015889,0.000126335,0.000174139,0.000138229,0.000139525,0.000116652,9.2864e-05,9.85822e-05,0.000140211,8.10463e-05,9.43126e-05,0.000106588,0.000114288,0.000103462,0.000145014,0.000161788,0.000134493,0.000175969,0.000129003,0.000136246,0.000119549,0.000184203,9.95734e-05,0.00011223,0.000112382,0.000115127,0.000111467,0.000119549,0.000101937,0.000105292,0.000125877,0.000121226,0.000174902,0.00013373,0.000143261,0.00012237,0.000189845,0.000121531,0.000137237,0.00012786,0.000130681,0.000102013,0.000164685,0.000118482,0.000120845,0.000126258,8.98143e-05,0.000158128,0.000124352,0.000147988,0.000133959,0.000140287,8.03602e-05,0.000128241,6.1795e-05,8.41723e-05,0.000106207,0.000112001,8.60784e-05,8.219e-05,0.000187253,0.000225374,0.000107655,0.000166057,0.000107121,0.000109942,0.000119778,0.000118634,0.000116271,9.39314e-05,8.85182e-05,0.000114746,0.000100717,0.000109714,0.000116957,0.000126716,0.000109104,0.000127173,0.000107045,9.98021e-05,9.599e-05,9.24828e-05,9.0043e-05,0.000126335,0.000112001,0.000139449,0.000118787,0.00016133,0.000120159,0.00013312,0.000106588,9.01193e-05,9.75911e-05,0.000120311,7.7806e-05,8.77557e-05,0.000105978,0.00011162,0.000102242,0.000130757,0.000149131,0.000126182,0.000166667,0.000122828,0.000121608,0.000104377,0.000178256,9.46938e-05,0.000102699,0.000108113,0.000102166,0.000104377,0.000107426,9.63712e-05,9.79723e-05,0.000120693,0.000117567,0.000174597,0.000132663,0.000125267,0.000120845,0.000173377,0.000119625,0.00013617,0.000122065,0.000112153,9.24828e-05,0.000147683,0.000107731,0.000115966,0.000125267,8.44773e-05,0.00015058,0.000125572,0.000136246,0.000124276,0.000129156,7.71198e-05,0.00012237,5.98508e-05,7.42988e-05,0.000134645,0.000142498,8.07414e-05,7.38414e-05,0.000251754,0.000198079,0.000101479,0.000150427,0.000102547,0.000106664,0.000118787,0.000118558,0.000113221,8.99668e-05,7.40701e-05,0.000105292,9.22541e-05,9.73623e-05,0.00011284,0.000112306,0.000100793,0.000112382,0.000102699,9.11867e-05,9.04242e-05,9.07292e-05,8.40961e-05,0.000118177,0.000105825,0.000114365,0.000112001,0.000129537,8.9738e-05,0.00012176,9.17204e-05,8.10463e-05,8.88994e-05,9.41601e-05,7.12491e-05,7.31552e-05,8.95093e-05,0.000100565,9.46176e-05,0.000108189,0.000114136,0.000101174,0.000133654,0.00010735,0.000100717,8.53922e-05,0.000163008,8.66883e-05,8.85182e-05,9.77436e-05,8.43248e-05,9.37027e-05,8.35624e-05,8.54684e-05,9.04242e-05,0.000110934,0.000109333,0.000166515,0.000127631,0.000106588,0.000114288,0.000162398,0.000112153,0.000131214,0.000115279,9.51513e-05,7.91403e-05,0.000128241,9.27878e-05,0.000107503,0.000118939,7.17828e-05,0.000138076,0.000120769,0.00011993,0.000109866,0.000107274,6.90762e-05,0.000112306,5.09303e-05,6.03464e-05, quantizeAxis: 3 zero_point data: ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, op_type: HB_DNN_OUTPUT_OPERATOR_TYPE_UNKNOWN 比如说想这个ouput[2]中的scale data就是 typedef struct { int32_t scaleLen; float *scaleData; int32_t zeroPointLen; int32_t *zeroPointData; } hbDNNQuantiScale; 中的scaleData吗？