On rv1126b, the Resize op in unet fallback to cpu #374
Description
D RKNN: [09:38:25.031] --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
D RKNN: [09:38:25.031] Network Layer Information Table
D RKNN: [09:38:25.031] --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
D RKNN: [09:38:25.031] ID OpType DataType Target InputShape OutputShape Cycles(DDR/NPU/Total) SparseRatio RW(KB) FullName
D RKNN: [09:38:25.031] --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
D RKNN: [09:38:25.031] 0 InputOperator INT8 CPU \ (1,3,128,256) 0/0/0 0 InputOperator:input
D RKNN: [09:38:25.031] 1 ConvRelu INT8 NPU (1,3,128,256),(16,3,3,3),(16) (1,16,128,256) 83513/294912/294912 0% 96 Conv:Conv_0
D RKNN: [09:38:25.031] 2 ConvRelu INT8 NPU (1,16,128,256),(16,16,3,3),(16) (1,16,128,256) 140809/294912/294912 0% 514 Conv:Conv_2
D RKNN: [09:38:25.031] 3 MaxPool INT8 NPU (1,16,128,256) (1,16,64,128) 0/0/0 512 MaxPool:MaxPool_4
D RKNN: [09:38:25.031] 4 ConvRelu INT8 NPU (1,16,64,128),(32,16,3,3),(32) (1,32,64,128) 53327/73728/73728 0% 132 Conv:Conv_5
D RKNN: [09:38:25.031] 5 ConvRelu INT8 NPU (1,32,64,128),(32,32,3,3),(32) (1,32,64,128) 71502/73728/73728 0% 265 Conv:Conv_7
D RKNN: [09:38:25.031] 6 MaxPool INT8 NPU (1,32,64,128) (1,32,32,64) 0/0/0 256 MaxPool:MaxPool_9
D RKNN: [09:38:25.031] 7 ConvRelu INT8 NPU (1,32,32,64),(64,32,3,3),(64) (1,64,32,64) 28876/36864/36864 0% 82 Conv:Conv_10
D RKNN: [09:38:25.031] 8 ConvRelu INT8 NPU (1,64,32,64),(64,64,3,3),(64) (1,64,32,64) 40124/73728/73728 0% 164 Conv:Conv_12
D RKNN: [09:38:25.031] 9 MaxPool INT8 NPU (1,64,32,64) (1,64,16,32) 0/0/0 128 MaxPool:MaxPool_14
D RKNN: [09:38:25.031] 10 ConvRelu INT8 NPU (1,64,16,32),(128,64,3,3),(128) (1,128,16,32) 23183/36864/36864 0% 105 Conv:Conv_15
D RKNN: [09:38:25.031] 11 ConvRelu INT8 NPU (1,128,16,32),(128,128,3,3),(128) (1,128,16,32) 37449/73728/73728 0% 209 Conv:Conv_17
D RKNN: [09:38:25.031] 12 MaxPool INT8 NPU (1,128,16,32) (1,128,8,16) 0/0/0 64 MaxPool:MaxPool_19
D RKNN: [09:38:25.031] 13 ConvRelu INT8 NPU (1,128,8,16),(128,128,3,3),(128) (1,128,8,16) 24280/18432/24280 0% 161 Conv:Conv_20
D RKNN: [09:38:25.031] 14 ConvRelu INT8 NPU (1,128,8,16),(128,128,3,3),(128) (1,128,8,16) 24280/18432/24280 0% 161 Conv:Conv_22
D RKNN: [09:38:25.031] 15 Resize INT8 CPU (1,128,8,16),(0),(4) (1,128,16,32) 0/0/0 16 Resize:Resize_25
D RKNN: [09:38:25.031] 16 Concat INT8 NPU (1,128,16,32),(1,128,16,32) (1,256,16,32) 0/0/0 128 Concat:Concat_26
D RKNN: [09:38:25.031] 17 ConvRelu INT8 NPU (1,256,16,32),(128,256,3,3),(128) (1,128,16,32) 65981/147456/147456 0% 417 Conv:Conv_27
D RKNN: [09:38:25.031] 18 ConvRelu INT8 NPU (1,128,16,32),(64,128,3,3),(64) (1,64,16,32) 23114/36864/36864 0% 136 Conv:Conv_29
D RKNN: [09:38:25.031] 19 Resize INT8 CPU (1,64,16,32),(0),(4) (1,64,32,64) 0/0/0 32 Resize:Resize_32
D RKNN: [09:38:25.031] 20 Concat INT8 NPU (1,64,32,64),(1,64,32,64) (1,128,32,64) 0/0/0 256 Concat:Concat_33
D RKNN: [09:38:25.031] 21 ConvRelu INT8 NPU (1,128,32,64),(64,128,3,3),(64) (1,64,32,64) 62620/147456/147456 0% 328 Conv:Conv_34
D RKNN: [09:38:25.031] 22 ConvRelu INT8 NPU (1,64,32,64),(32,64,3,3),(32) (1,32,32,64) 28841/36864/36864 0% 146 Conv:Conv_36
D RKNN: [09:38:25.031] 23 Resize INT8 CPU (1,32,32,64),(0),(4) (1,32,64,128) 0/0/0 64 Resize:Resize_39
D RKNN: [09:38:25.031] 24 Concat INT8 NPU (1,32,64,128),(1,32,64,128) (1,64,64,128) 0/0/0 512 Concat:Concat_40
D RKNN: [09:38:25.031] 25 ConvRelu INT8 NPU (1,64,64,128),(32,64,3,3),(32) (1,32,64,128) 107853/147456/147456 0% 530 Conv:Conv_41
D RKNN: [09:38:25.031] 26 ConvRelu INT8 NPU (1,32,64,128),(16,32,3,3),(16) (1,16,64,128) 53327/73728/73728 0% 260 Conv:Conv_43
D RKNN: [09:38:25.031] 27 Resize INT8 CPU (1,16,64,128),(0),(4) (1,16,128,256) 0/0/0 128 Resize:Resize_46
D RKNN: [09:38:25.031] 28 Concat INT8 NPU (1,16,128,256),(1,16,128,256) (1,32,128,256) 0/0/0 1024 Concat:Concat_47
D RKNN: [09:38:25.031] 29 ConvRelu INT8 NPU (1,32,128,256),(16,32,3,3),(16) (1,16,128,256) 211351/294912/294912 0% 1028 Conv:Conv_48
D RKNN: [09:38:25.031] 30 ConvRelu INT8 NPU (1,16,128,256),(16,16,3,3),(16) (1,16,128,256) 140809/294912/294912 0% 514 Conv:Conv_50
D RKNN: [09:38:25.031] 31 Conv INT8 NPU (1,16,128,256),(2,16,1,1),(2) (1,2,128,256) 210738/131072/210738 0% 512 Conv:Conv_52
D RKNN: [09:38:25.031] 32 OutputOperator INT8 CPU (1,2,128,256) \ 0/0/0 1024 OutputOperator:output
D RKNN: [09:38:25.031] --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
According to 05_RKNN_Compiler_Support_Operator_List_V2.3.2.pdf page 223
I've changed nearest_mode from floor to round_prefer_floor.