flashocc和maptr模型输入问题 - 地平线开发者社区

背景：oe3.2.0版本的flashocc与maptr模型

问题：

1、flashocc中的input[12]和input[13]输入含义以及该如何基于工具联准备输入呢？

模型输入信息：

input[12]: name: points0 valid shape: (10,128,128,2) aligned byte size: 655360 tensor type: HB_DNN_TENSOR_TYPE_S16 quanti type: SCALE stride: (65536,512,4,2) scale data: (0.0078125)

input[13]: name: points1 valid shape: (10,128,128,2) aligned byte size: 655360 tensor type: HB_DNN_TENSOR_TYPE_S16 quanti type: SCALE stride: (65536,512,4,2) scale data: (0.015625)

2、maptr中的这些输入是什么意思，该如何准备呢？
input[12]: name: osm_mask valid shape: (1,1,50,100) aligned byte size: 6400 tensor type: HB_DNN_TENSOR_TYPE_S8 quanti type: SCALE stride: (6400,6400,128,1) scale data: (0.0078125)

input[13]: name: queries_rebatch_grid valid shape: (6,20,100,2) aligned byte size: 61440 tensor type: HB_DNN_TENSOR_TYPE_S16 quanti type: SCALE stride: (10240,512,4,2) scale data: (3.1135e-05,3.17642e-05) quantizeAxis: 3

input[14]: name: restore_bev_grid valid shape: (1,100,100,2) aligned byte size: 51200 tensor type: HB_DNN_TENSOR_TYPE_S16 quanti type: SCALE stride: (51200,512,4,2) scale data: (3.1135e-05,3.10314e-05) quantizeAxis: 3

input[15]: name: reference_points_rebatch valid shape: (6,2000,4,2) aligned byte size: 192000 tensor type: HB_DNN_TENSOR_TYPE_S16 quanti type: SCALE stride: (32000,16,4,2) scale data: (6.40879e-05)

input[16]: name: bev_pillar_counts valid shape: (1,5000,1) aligned byte size: 5120 tensor type: HB_DNN_TENSOR_TYPE_S8 quanti type: SCALE stride: (5000,1,1) scale data: (0.00784314)