#模型性能Benchmark

#说明

• 测试条件：

  • 测试开发板：S100。

  • 测试核心数：单核。

  • 性能数据获取频率设置为：5分钟时间内性能参数的平均值。

  • Python版本：Python3.10。

  • 模型来源：OE包内 samples/ucp_tutorial/dnn/ai_benchmark/s100 路径下的模型。

  • 运行环境：Linux。

• 缩写说明：

  • C = 计算量，单位为GOPs（十亿次运算/秒）。此数据通过调用 hbm_perf 接口获得。

  • FPS = 每秒帧率。此数据在开发板多线程运行ai_benchmark示例包/script路径下各模型子文件夹的 fps.sh 脚本获取，包含后处理。

  • ITC = 推理耗时，单位为ms（毫秒）。此数据在开发板单线程运行ai_benchmark示例包/script路径下各模型子文件夹的 latency.sh 脚本获取，不含后处理。

  • TCPP = 后处理耗时，单位为ms（毫秒）。此数据在开发板单线程运行ai_benchmark示例包/script路径下各模型子文件夹的 latency.sh 脚本获取。

  • RV = 单次推理读取数据量，单位为mb（兆比特）。此数据通过调用 hbm_perf 接口获得。

  • WV = 单次推理写入数据量，单位为mb（兆比特）。此数据通过调用 hbm_perf 接口获得。

#模型主要性能数据

#模型全部性能数据

#MobileNetv1

• INPUT SIZE: 1x3x224x224
• C(GOPs): 1.14
• FPS: 4263.30
• ITC(ms): 0.527
• TCPP(ms): 0.034
• RV(mb): 4.56
• WV(mb): 0.0038
• Dataset: ImageNet
• ACCURACY: Top1: 0.7373(FLOAT)/0.7297(INT8)

#MobileNetv2

• INPUT SIZE: 1x3x224x224
• C(GOPs): 0.63
• FPS: 4277.30
• ITC(ms): 0.542
• TCPP(ms): 0.034
• RV(mb): 3.95
• WV(mb): 0.0038
• Dataset: ImageNet
• ACCURACY: Top1: 0.7217(FLOAT)/0.7144(INT8)

#ResNet50

• INPUT SIZE: 1x3x224x224
• C(GOPs): 7.72
• FPS: 1155.00
• ITC(ms): 1.218
• TCPP(ms): 0.034
• RV(mb): 26.08
• WV(mb): 0.0038
• Dataset: ImageNet
• ACCURACY: Top1: 0.7703(FLOAT)/0.7677(INT8)

#GoogleNet

• INPUT SIZE: 1x3x224x224
• C(GOPs): 3.00
• FPS: 2790.90
• ITC(ms): 0.702
• TCPP(ms): 0.034
• RV(mb): 7.09
• WV(mb): 0.0020
• Dataset: ImageNet
• ACCURACY: Top1: 0.7018(FLOAT)/0.6998(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/GoogleNet

#EfficientNet_Lite0

• INPUT SIZE: 1x224x224x3
• C(GOPs): 0.77
• FPS: 3900.80
• ITC(ms): 0.626
• TCPP(ms): 0.034
• RV(mb): 5.54
• WV(mb): 0.00000000
• Dataset: ImageNet
• ACCURACY: Top1: 0.7479(FLOAT)/0.7453(INT8)
• LINKS: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet/lite

#EfficientNet_Lite1

• INPUT SIZE: 1x240x240x3
• C(GOPs): 1.20
• FPS: 3200.60
• ITC(ms): 0.680
• TCPP(ms): 0.034
• RV(mb): 6.44
• WV(mb): 0.00000000
• Dataset: ImageNet
• ACCURACY: Top1: 0.7652(FLOAT)/0.7609(INT8)
• LINKS: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet/lite

#EfficientNet_Lite2

• INPUT SIZE: 1x260x260x3
• C(GOPs): 1.72
• FPS: 2477.50
• ITC(ms): 0.771
• TCPP(ms): 0.034
• RV(mb): 7.28
• WV(mb): 0.00000000
• Dataset: ImageNet
• ACCURACY: Top1: 0.7734(FLOAT)/0.7697(INT8)
• LINKS: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet/lite

#EfficientNet_Lite3

• INPUT SIZE: 1x280x280x3
• C(GOPs): 2.77
• FPS: 1865.80
• ITC(ms): 0.911
• TCPP(ms): 0.034
• RV(mb): 9.72
• WV(mb): 0.00000000
• Dataset: ImageNet
• ACCURACY: Top1: 0.7917(FLOAT)/0.7895(INT8)
• LINKS: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet/lite

#EfficientNet_Lite4

• INPUT SIZE: 1x300x300x3
• C(GOPs): 5.11
• FPS: 1297.10
• ITC(ms): 1.144
• TCPP(ms): 0.034
• RV(mb): 14.80
• WV(mb): 0.00000000
• Dataset: ImageNet
• ACCURACY: Top1: 0.8063(FLOAT)/0.8043(INT8)
• LINKS: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet/lite

#Vargconvnet

• INPUT SIZE: 1x3x224x224
• C(GOPs): 9.06
• FPS: 1408.70
• ITC(ms): 1.061
• TCPP(ms): 0.034
• RV(mb): 11.20
• WV(mb): 0.35
• Dataset: ImageNet
• ACCURACY: Top1: 0.7793(FLOAT)/0.7762(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/VargConvNet

#Efficientnasnet_m

• INPUT SIZE: 1x3x300x300
• C(GOPs): 4.53
• FPS: 1468.50
• ITC(ms): 1.029
• TCPP(ms): 0.034
• RV(mb): 13.93
• WV(mb): 0.0020
• Dataset: ImageNet
• ACCURACY: Top1: 0.7935(FLOAT)/0.7924(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/EfficientnasNet

#Efficientnasnet_s

• INPUT SIZE: 1x3x280x280
• C(GOPs): 1.44
• FPS: 3313.20
• ITC(ms): 0.643
• TCPP(ms): 0.034
• RV(mb): 5.61
• WV(mb): 0.0020
• Dataset: ImageNet
• ACCURACY: Top1: 0.7441(FLOAT)/0.7515(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/EfficientnasNet

#ResNet18

• INPUT SIZE: 1x3x224x224
• C(GOPs): 3.63
• FPS: 2553.80
• ITC(ms): 0.729
• TCPP(ms): 0.034
• RV(mb): 11.86
• WV(mb): 0.00000000
• Dataset: ImageNet
• ACCURACY: Top1: 0.7169(FLOAT)/0.7163(INT8)

#YOLOv2_Darknet19

• INPUT SIZE: 1x3x608x608
• C(GOPs): 62.94
• FPS: 226.19
• ITC(ms): 4.793
• TCPP(ms): 0.305
• RV(mb): 52.24
• WV(mb): 1.07
• Dataset: COCO
• ACCURACY: [IoU=0.50:0.95]= 0.2760(FLOAT)/0.2700(INT8)
• LINKS: https://pjreddie.com/darknet/yolo

#YOLOv3_Darknet53

• INPUT SIZE: 1x3x416x416
• C(GOPs): 65.86
• FPS: 212.55
• ITC(ms): 5.150
• TCPP(ms): 1.746
• RV(mb): 68.68
• WV(mb): 9.40
• Dataset: COCO
• ACCURACY: [IoU=0.50:0.95]= 0.3370(FLOAT)/0.3350(INT8)
• LINKS: https://github.com/ChenYingpeng/caffe-yolov3/

#YOLOv5x_v2.0

• INPUT SIZE: 1x3x672x672
• C(GOPs): 243.85
• FPS: 62.24
• ITC(ms): 16.593
• TCPP(ms): 5.907
• RV(mb): 138.56
• WV(mb): 48.43
• Dataset: COCO
• ACCURACY: [IoU=0.50:0.95]= 0.4810(FLOAT)/0.4670(INT8)
• LINKS: https://github.com/ultralytics/yolov5/releases/tag/v2.0

#SSD_MobileNetv1

• INPUT SIZE: 1x3x300x300
• C(GOPs): 2.30
• FPS: 3194.00
• ITC(ms): 0.727
• TCPP(ms): 0.198
• RV(mb): 6.24
• WV(mb): 0.20
• Dataset: VOC
• ACCURACY: mAP: 0.7345(FLOAT)/0.7269(INT8)
• LINKS: https://github.com/chuanqi305/MobileNet-SSD

#Centernet_resnet101

• INPUT SIZE: 1x3x512x512
• C(GOPs): 90.53
• FPS: 186.46
• ITC(ms): 5.781
• TCPP(ms): 0.991
• RV(mb): 59.17
• WV(mb): 11.63
• Dataset: COCO
• ACCURACY: [IoU=0.50:0.95]= 0.3420(FLOAT)/0.3270(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/Centernet

#YOLOv3_VargDarknet

• INPUT SIZE: 1x3x416x416
• C(GOPs): 42.82
• FPS: 293.51
• ITC(ms): 3.852
• TCPP(ms): 1.672
• RV(mb): 50.68
• WV(mb): 8.85
• Dataset: COCO
• ACCURACY: [IoU=0.50:0.95]= 0.3280(FLOAT)/0.3270(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/Yolov3_VargDarknet

#Deeplabv3plus_efficientnetb0

• INPUT SIZE: 1x3x1024x2048
• C(GOPs): 30.77
• FPS: 151.51
• ITC(ms): 7.016
• TCPP(ms): 0.314
• RV(mb): 21.99
• WV(mb): 11.76
• Dataset: Cityscapes
• ACCURACY: mIoU: 0.7630(FLOAT)/0.7569(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/DeeplabV3Plus

#Fastscnn_efficientnetb0

• INPUT SIZE: 1x3x1024x2048
• C(GOPs): 12.48
• FPS: 249.25
• ITC(ms): 4.422
• TCPP(ms): 0.315
• RV(mb): 11.77
• WV(mb): 9.55
• Dataset: Cityscapes
• ACCURACY: mIoU: 0.6997(FLOAT)/0.6914(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/FastSCNN

#Deeplabv3plus_efficientnetm1

• INPUT SIZE: 1x3x1024x2048
• C(GOPs): 77.04
• FPS: 92.00
• ITC(ms): 11.308
• TCPP(ms): 0.313
• RV(mb): 85.72
• WV(mb): 55.54
• Dataset: Cityscapes
• ACCURACY: mIoU: 0.7794(FLOAT)/0.7754(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/DeeplabV3Plus

#Deeplabv3plus_efficientnetm2

• INPUT SIZE: 1x3x1024x2048
• C(GOPs): 124.15
• FPS: 64.62
• ITC(ms): 15.921
• TCPP(ms): 0.312
• RV(mb): 154.10
• WV(mb): 89.10
• Dataset: Cityscapes
• ACCURACY: mIoU: 0.7882(FLOAT)/0.7853(INT8)
• LINKS: https://github.com/HorizonRobotics-Platform/ModelZoo/tree/master/DeeplabV3Plus

#Bev_gkt_mixvargenet_multitask

• INPUT SIZE: image: 6x3x512x960 points(0-8): 6x64x64x2
• C(GOPs): 207.16
• FPS: 68.40
• ITC(ms): 15.872
• TCPP(ms): 5.367
• RV(mb): 120.84
• WV(mb): 95.22
• Dataset: Nuscenes
• ACCURACY: NDS: 0.2810(FLOAT)/0.2798(INT8) MeanIOU: 0.4852(FLOAT)/0.4838(INT8) mAP: 0.1991(FLOAT)/0.1995(INT8)

#Bev_ipm_4d_efficientnetb0_multitask

• INPUT SIZE: image: 6x3x512x960 points: 6x128x128x2 prev_feat: 1x164x28x128 prev_point: 1x128x128x2
• C(GOPs): 53.58
• FPS: 111.92
• ITC(ms): 10.488
• TCPP(ms): 5.487
• RV(mb): 63.20
• WV(mb): 49.93
• Dataset: Nuscenes
• ACCURACY: NDS: 0.3721(FLOAT)/0.3725(INT8) MeanIOU: 0.5287(FLOAT)/0.5389(INT8) mAP: 0.2200(FLOAT)/0.2214(INT8)

#Bev_ipm_efficientnetb0_multitask

• INPUT SIZE: image: 6x3x512x960 points: 6x128x128x2
• C(GOPs): 52.97
• FPS: 115.12
• ITC(ms): 9.838
• TCPP(ms): 5.334
• RV(mb): 59.90
• WV(mb): 47.84
• Dataset: Nuscenes
• ACCURACY: NDS: 0.3055(FLOAT)/0.3032(INT8) MeanIOU: 0.5145(FLOAT)/0.5104(INT8) mAP: 0.2169(FLOAT)/0.2168(INT8)

#Bev_lss_efficientnetb0_multitask

• INPUT SIZE: image: 6x3x256x704 points(0&1): 10x128x128x2
• C(GOPs): 24.06
• FPS: 187.16
• ITC(ms): 6.494
• TCPP(ms): 5.417
• RV(mb): 25.43
• WV(mb): 18.99
• Dataset: Nuscenes
• ACCURACY: NDS: 0.3007(FLOAT)/0.2995(INT8) MeanIOU: 0.5180(FLOAT)/0.5148(INT8) mAP: 0.2062(FLOAT)/0.2042(INT8)

#Detr3d_efficientnetb3

• INPUT SIZE: coords(0-3): 6x4x256x2 image: 6x3x512x1408 masks: 1x4x256x24
• C(GOPs): 227.71
• FPS: 32.08
• ITC(ms): 31.877
• TCPP(ms): 1.122
• RV(mb): 333.71
• WV(mb): 175.58
• Dataset: Nuscenes
• ACCURACY: NDS: 0.3304(FLOAT)/0.3288(INT8) mAP: 0.2752(FLOAT)/0.2712(INT8)

#Petr_efficientnetb3

• INPUT SIZE: image: 6x3x512x1408 pos_embed: 1x96x44x256
• C(GOPs): 219.17
• FPS: 19.25
• ITC(ms): 52.643
• TCPP(ms): 1.140
• RV(mb): 260.88
• WV(mb): 144.23
• Dataset: Nuscenes
• ACCURACY: NDS: 0.3765(FLOAT)/0.3735(INT8) mAP: 0.3038(FLOAT)/0.2936(INT8)

#Bevformer_tiny_resnet50_detection

• INPUT SIZE: img: 6x3x480x800 prev_bev: 1x2500x256 prev_bev_ref: 1x50x50x2 queries_rebatch_grid: 6x20x32x2 restore_bev_grid: 1x100x50x2 reference_points_rebatch: 6x640x4x2 bev_pillar_counts: 1x2500x1
• C(GOPs): 387.29
• FPS: 31.17
• ITC(ms): 42.108
• TCPP(ms): 1.412
• RV(mb): 265.57
• WV(mb): 175.41
• Dataset: Nuscenes
• ACCURACY: NDS: 0.3713(FLOAT)/0.3679(INT8) mAP: 0.2673(FLOAT)/0.2614(INT8)

#Flashocc_henet_lss_occ3d_nuscenes

• INPUT SIZE: img: 6x3x512x960 points: 10x128x128x2 points_depth: 10x128x128x2
• C(GOPs): 126.75
• FPS: 96.28
• ITC(ms): 11.497
• TCPP(ms): 40.899
• RV(mb): 87.37
• WV(mb): 55.02
• Dataset: Nuscenes
• ACCURACY: mIoU: 0.3674(FLOAT)/0.3640(INT8)

#Horizon_swin_transformer

• INPUT SIZE: 1x3x224x224
• C(GOPs): 8.98
• FPS: 311.81
• ITC(ms): 3.569
• TCPP(ms): 0.035
• RV(mb): 46.27
• WV(mb): 6.82
• Dataset: ImageNet
• ACCURACY: Top1: 0.8024(FLOAT)/0.7955(INT8)

#Mixvargenet

• INPUT SIZE: 1x3x224x224
• C(GOPs): 2.07
• FPS: 4432.40
• ITC(ms): 0.534
• TCPP(ms): 0.034
• RV(mb): 2.51
• WV(mb): 0.0038
• Dataset: ImageNet
• ACCURACY: Top1: 0.7075(FLOAT)/0.7054(INT8)

#Vargnetv2

• INPUT SIZE: 1x3x224x224
• C(GOPs): 0.72
• FPS: 4027.10
• ITC(ms): 0.593
• TCPP(ms): 0.034
• RV(mb): 4.68
• WV(mb): 0.0038
• Dataset: ImageNet
• ACCURACY: Top1: 0.7342(FLOAT)/0.7316(INT8)

#Vit_small

• INPUT SIZE: 1x3x224x224
• C(GOPs): 9.20
• FPS: 547.19
• ITC(ms): 2.185
• TCPP(ms): 0.035
• RV(mb): 26.27
• WV(mb): 0.00000000
• Dataset: ImageNet
• ACCURACY: Top1: 0.7950(FLOAT)/0.7921(INT8)

#Centerpoint_pointpillar

• INPUT SIZE: points: 300000x5 voxel_feature: 1x5x20x40000 coors: 40000x4
• C(GOPs): 127.73
• FPS: 124.77
• ITC(ms): 16.555
• TCPP(ms): 14.028
• RV(mb): 51.20
• WV(mb): 26.87
• Dataset: Nuscenes
• ACCURACY: NDS: 0.5832(FLOAT)/0.5817(INT8) mAP: 0.4804(FLOAT)/0.4783(INT8)

#Detr_efficientnetb3

• INPUT SIZE: 1x3x800x1333
• C(GOPs): 67.39
• FPS: 52.73
• ITC(ms): 19.406
• TCPP(ms): 0.344
• RV(mb): 261.91
• WV(mb): 134.85
• Dataset: MS COCO
• ACCURACY: [IoU=0.50:0.95]= 0.3721(FLOAT)/0.3599(INT8)

#Detr_resnet50

• INPUT SIZE: 1x3x800x1333
• C(GOPs): 203.07
• FPS: 40.25
• ITC(ms): 25.385
• TCPP(ms): 0.343
• RV(mb): 376.39
• WV(mb): 240.01
• Dataset: MS COCO
• ACCURACY: [IoU=0.50:0.95]= 0.3569(FLOAT)/0.3164(INT8)

#FCOS3D_efficientnetb0

• INPUT SIZE: 1x3x512x896
• C(GOPs): 19.94
• FPS: 447.98
• ITC(ms): 3.346
• TCPP(ms): 2.745
• RV(mb): 11.23
• WV(mb): 4.17
• Dataset: nuscenes
• ACCURACY: NDS: 0.3061(FLOAT)/0.3029(INT8) mAP: 0.2133(FLOAT)/0.2064(INT8)

#Fcos_efficientnetb0

• INPUT SIZE: 1x3x512x512
• C(GOPs): 5.02
• FPS: 1079.30
• ITC(ms): 1.611
• TCPP(ms): 0.137
• RV(mb): 6.08
• WV(mb): 2.61
• Dataset: MS COCO
• ACCURACY: [IoU=0.50:0.95]= 0.3626(FLOAT)/0.3564(INT8)

#Ganet_mixvargenet

• INPUT SIZE: 1x3x320x800
• C(GOPs): 10.74
• FPS: 1514.30
• ITC(ms): 1.051
• TCPP(ms): 0.219
• RV(mb): 2.13
• WV(mb): 0.45
• Dataset: CuLane
• ACCURACY: F1Score: 0.7948(FLOAT)/0.7878(INT8)

#Keypoint_efficientnetb0

• INPUT SIZE: 1x3x128x128
• C(GOPs): 0.45
• FPS: 4289.70
• ITC(ms): 0.547
• TCPP(ms): 0.068
• RV(mb): 4.61
• WV(mb): 0.00000000
• Dataset: Carfusion
• ACCURACY: PCK(alpha=0.1): 0.9433(FLOAT)/0.9433(INT8)

#Pointpillars_kitti_car

• INPUT SIZE: 150000x4
• C(GOPs): 66.82
• FPS: 144.65
• ITC(ms): 33.240
• TCPP(ms): 0.539
• RV(mb): 70.72
• WV(mb): 30.65
• Dataset: Kitti3d
• ACCURACY: APDet= 0.7732(FLOAT)/0.7675(INT8)

#Deformable_detr_resnet50

• INPUT SIZE: 1x3x800x1333
• C(GOPs): 408.94
• FPS: 5.30
• ITC(ms): 190.060
• TCPP(ms): 15.533
• RV(mb): 3495.68
• WV(mb): 2485.77
• Dataset: MS COCO
• ACCURACY: [IoU=0.50:0.95]= 0.4414(FLOAT)/0.4202(INT8)

#Stereonetplus_mixvargenet

• INPUT SIZE: 2x3x544x960
• C(GOPs): 48.57
• FPS: 229.28
• ITC(ms): 4.853
• TCPP(ms): 1.970
• RV(mb): 27.85
• WV(mb): 25.86
• Dataset: SceneFlow
• ACCURACY: EPE: 1.1270(FLOAT)/1.1346(INT8)

#Centerpoint_mixvargnet_multitask

• INPUT SIZE: points: 300000x5 voxel_feature: 1x5x20x40000 coors: 40000x4
• C(GOPs): 51.45
• FPS: 180.01
• ITC(ms): 14.285
• TCPP(ms): 11.415
• RV(mb): 32.49
• WV(mb): 18.87
• Dataset: Nuscenes
• ACCURACY: NDS: 0.5809(FLOAT)/0.5751(INT8) MeanIOU: 0.9128(FLOAT)/0.9121(INT8) mAP: 0.4726(FLOAT)/0.4627(INT8)

#Unet_mobilenetv1

• INPUT SIZE: 1x3x1024x2048
• C(GOPs): 7.36
• FPS: 819.01
• ITC(ms): 1.709
• TCPP(ms): 0.148
• RV(mb): 12.46
• WV(mb): 7.59
• Dataset: Cityscapes
• ACCURACY: mIoU: 0.6802(FLOAT)/0.6758(INT8)

#Motr_efficientnetb3

• INPUT SIZE: image: 1x800x1422x3 track_query: 1x2x128x156 ref_points: 1x2x128x4 mask_query: 1x1x256x1
• C(GOPs): 64.43
• FPS: 74.28
• ITC(ms): 13.658
• TCPP(ms): 5.066
• RV(mb): 110.89
• WV(mb): 43.96
• Dataset: Mot17
• ACCURACY: MOTA: 0.5805(FLOAT)/0.5748(INT8)

#Densetnt_vectornet

• INPUT SIZE: goals_2d: 30x1x2048x2 goals_2d_mask: 30x1x2048x1 instance_mask: 30x1x96x1 lane_feat: 30x9x64x11 traj_feat: 30x19x32x9
• C(GOPs): 12.50
• FPS: 104.27
• ITC(ms): 10.427
• TCPP(ms): 2.306
• RV(mb): 53.07
• WV(mb): 33.36
• Dataset: Argoverse 1
• ACCURACY: minFDA: 1.2975(FLOAT)/1.3059(INT8)

#Maptroe_henet_tinym_bevformer

• INPUT SIZE: img: 6x3x480x800 osm_mask: 1x1x50x100 queries_rebatch_grid: 6x20x100x2 restore_bev_grid: 1x100x100x2 reference_points_rebatch: 6x2000x4x2 bev_pillar_counts: 1x5000x1
• C(GOPs): 134.57
• FPS: 75.31
• ITC(ms): 13.953
• TCPP(ms): 0.261
• RV(mb): 121.58
• WV(mb): 36.83
• Dataset: Nuscenes
• ACCURACY: mAP: 0.6633(FLOAT)/0.6569(INT8)

#Qcnet_oe

• INPUT SIZE: valid_mask: 1x30x10 valid_mask_a2a: 1x10x30x30 agent_type: 1x30x1 x_a_cur: 1x1x30x1,1x1x30x1,1x1x30x1,1x1x30x1 r_pl2a_cur: 1x1x30x80,1x1x30x80,1x1x30x80 r_t_cur: 1x1x30x6,1x1x30x6,1x1x30x6,1x1x30x6 r_a2a_cur: 1x1x30x30,1x1x30x30,1x1x30x30 x_a_mid_emb: 1x30x2x128 x_a: 1x30x6x128 pl_type,is_intersection: 1x80 r_pl2pl: 1x1x80x80,1x1x80x80,1x1x80x80 r_pt2pl: 1x1x80x50,1x1x80x50,1x1x80x50 mask_pl2pl: 1x80x80 magnitude,pt_type,side,mask: 1x80x50 mask_a2m: 1x30x30 mask_dst: 1x30x1 type_pl2pl: 1x80x80
• C(GOPs): 7.85
• FPS: 236.04
• ITC(ms): 6.299
• TCPP(ms): 0.827
• RV(mb): 37.89
• WV(mb): 18.03
• Dataset: Argoverse 2
• ACCURACY: hitrate: 0.8026(FLOAT)/0.7923(INT8)