ncnn
ncnn copied to clipboard
Tencent
ncnn的softMax和torch.nn.functional softMax计算结果有偏差
图片是ncnn的softMax和pytorch的torch.nn.functional softMax进行计算的数据，计算的结果有偏差
ncnn 显示的是0，但torch.nn.functional 的SoftMax显示的是0.9？？
7767517 308 380 Input in0 0 1 in0 Convolution convrelu_0 1 1 in0 1 0=64 1=7 11=7 12=1 13=2 14=3 2=1 3=2 4=3 5=1 6=9408 9=1 Pooling maxpool2d_125 1 1 1 2 0=0 1=3 11=3 12=2 13=1 2=2 3=1 5=1 Split splitncnn_0 1 2 2 3 4 Convolution convrelu_1 1 1 4 5 0=64 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=4096 9=1 Convolution convrelu_2 1 1 5 6 0=64 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=36864 9=1 Convolution conv_43 1 1 6 7 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=16384 Convolution conv_42 1 1 3 8 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=16384 BinaryOp add_0 2 1 7 8 9 0=0 ReLU relu_129 1 1 9 10 Split splitncnn_1 1 2 10 11 12 Convolution convrelu_3 1 1 12 13 0=64 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=16384 9=1 Convolution convrelu_4 1 1 13 14 0=64 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=36864 9=1 Convolution conv_46 1 1 14 15 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=16384 BinaryOp add_1 2 1 15 11 16 0=0 ReLU relu_132 1 1 16 17 Split splitncnn_2 1 2 17 18 19 Convolution convrelu_5 1 1 19 20 0=64 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=16384 9=1 Convolution convrelu_6 1 1 20 21 0=64 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=36864 9=1 Convolution conv_49 1 1 21 22 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=16384 BinaryOp add_2 2 1 22 18 23 0=0 ReLU relu_135 1 1 23 24 Split splitncnn_3 1 2 24 25 26 MemoryData backbone.layers.1.0.conv2 0 1 27 0=128 MemoryData pnnx_unique_6 0 1 28 0=3 1=3 11=128 2=128 Convolution convrelu_7 1 1 26 29 0=128 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=32768 9=1 Split splitncnn_4 1 2 29 30 31 Convolution conv_51 1 1 31 32 0=27 1=3 11=3 12=1 13=2 14=1 2=1 3=2 4=1 5=1 6=31104 Slice chunk_0 1 3 32 33 34 35 -23300=3,-233,-233,-233 1=0 Concat cat_0 2 1 33 34 36 0=0 Sigmoid sigmoid_5 1 1 35 37 DCNv2 pnnx_35 5 1 30 28 27 36 37 38 10=1 11=1 12=1 13=1 5=3 6=3 7=2 8=2 9=1 BatchNorm bn_26 1 1 38 39 0=128 1=1.000000e-05 ReLU relu_137 1 1 39 40 Convolution conv_53 1 1 40 41 0=512 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=65536 Convolution conv_52 1 1 25 42 0=512 1=1 11=1 12=1 13=2 14=0 2=1 3=2 4=0 5=1 6=131072 BinaryOp add_3 2 1 41 42 43 0=0 ReLU relu_138 1 1 43 44 Split splitncnn_5 1 2 44 45 46 MemoryData backbone.layers.1.1.conv2 0 1 47 0=128 MemoryData pnnx_unique_9 0 1 48 0=3 1=3 11=128 2=128 Convolution convrelu_8 1 1 46 49 0=128 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=65536 9=1 Split splitncnn_6 1 2 49 50 51 Convolution conv_55 1 1 51 52 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=31104 Slice chunk_1 1 3 52 53 54 55 -23300=3,-233,-233,-233 1=0 Concat cat_1 2 1 53 54 56 0=0 Sigmoid sigmoid_6 1 1 55 57 DCNv2 pnnx_55 5 1 50 48 47 56 57 58 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_27 1 1 58 59 0=128 1=1.000000e-05 ReLU relu_140 1 1 59 60 Convolution conv_56 1 1 60 61 0=512 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=65536 BinaryOp add_4 2 1 61 45 62 0=0 ReLU relu_141 1 1 62 63 Split splitncnn_7 1 2 63 64 65 MemoryData backbone.layers.1.2.conv2 0 1 66 0=128 MemoryData pnnx_unique_12 0 1 67 0=3 1=3 11=128 2=128 Convolution convrelu_9 1 1 65 68 0=128 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=65536 9=1 Split splitncnn_8 1 2 68 69 70 Convolution conv_58 1 1 70 71 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=31104 Slice chunk_2 1 3 71 72 73 74 -23300=3,-233,-233,-233 1=0 Concat cat_2 2 1 72 73 75 0=0 Sigmoid sigmoid_7 1 1 74 76 DCNv2 pnnx_75 5 1 69 67 66 75 76 77 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_28 1 1 77 78 0=128 1=1.000000e-05 ReLU relu_143 1 1 78 79 Convolution conv_59 1 1 79 80 0=512 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=65536 BinaryOp add_5 2 1 80 64 81 0=0 ReLU relu_144 1 1 81 82 Split splitncnn_9 1 2 82 83 84 MemoryData backbone.layers.1.3.conv2 0 1 85 0=128 MemoryData pnnx_unique_15 0 1 86 0=3 1=3 11=128 2=128 Convolution convrelu_10 1 1 84 87 0=128 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=65536 9=1 Split splitncnn_10 1 2 87 88 89 Convolution conv_61 1 1 89 90 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=31104 Slice chunk_3 1 3 90 91 92 93 -23300=3,-233,-233,-233 1=0 Concat cat_3 2 1 91 92 94 0=0 Sigmoid sigmoid_8 1 1 93 95 DCNv2 pnnx_95 5 1 88 86 85 94 95 96 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_29 1 1 96 97 0=128 1=1.000000e-05 ReLU relu_146 1 1 97 98 Convolution conv_62 1 1 98 99 0=512 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=65536 BinaryOp add_6 2 1 99 83 100 0=0 ReLU relu_147 1 1 100 101 Split splitncnn_11 1 3 101 102 103 104 MemoryData backbone.layers.2.0.conv2 0 1 105 0=256 MemoryData pnnx_unique_18 0 1 106 0=3 1=3 11=256 2=256 Convolution convrelu_11 1 1 104 107 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=131072 9=1 Split splitncnn_12 1 2 107 108 109 Convolution conv_64 1 1 109 110 0=27 1=3 11=3 12=1 13=2 14=1 2=1 3=2 4=1 5=1 6=62208 Slice chunk_4 1 3 110 111 112 113 -23300=3,-233,-233,-233 1=0 Concat cat_4 2 1 111 112 114 0=0 Sigmoid sigmoid_9 1 1 113 115 DCNv2 pnnx_115 5 1 108 106 105 114 115 116 10=1 11=1 12=1 13=1 5=3 6=3 7=2 8=2 9=1 BatchNorm bn_30 1 1 116 117 0=256 1=1.000000e-05 ReLU relu_149 1 1 117 118 Convolution conv_66 1 1 118 119 0=1024 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 Convolution conv_65 1 1 103 120 0=1024 1=1 11=1 12=1 13=2 14=0 2=1 3=2 4=0 5=1 6=524288 BinaryOp add_7 2 1 119 120 121 0=0 ReLU relu_150 1 1 121 122 Split splitncnn_13 1 2 122 123 124 MemoryData backbone.layers.2.1.conv2 0 1 125 0=256 MemoryData pnnx_unique_21 0 1 126 0=3 1=3 11=256 2=256 Convolution convrelu_12 1 1 124 127 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 9=1 Split splitncnn_14 1 2 127 128 129 Convolution conv_68 1 1 129 130 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=62208 Slice chunk_5 1 3 130 131 132 133 -23300=3,-233,-233,-233 1=0 Concat cat_5 2 1 131 132 134 0=0 Sigmoid sigmoid_10 1 1 133 135 DCNv2 pnnx_135 5 1 128 126 125 134 135 136 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_31 1 1 136 137 0=256 1=1.000000e-05 ReLU relu_152 1 1 137 138 Convolution conv_69 1 1 138 139 0=1024 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 BinaryOp add_8 2 1 139 123 140 0=0 ReLU relu_153 1 1 140 141 Split splitncnn_15 1 2 141 142 143 MemoryData backbone.layers.2.2.conv2 0 1 144 0=256 MemoryData pnnx_unique_24 0 1 145 0=3 1=3 11=256 2=256 Convolution convrelu_13 1 1 143 146 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 9=1 Split splitncnn_16 1 2 146 147 148 Convolution conv_71 1 1 148 149 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=62208 Slice chunk_6 1 3 149 150 151 152 -23300=3,-233,-233,-233 1=0 Concat cat_6 2 1 150 151 153 0=0 Sigmoid sigmoid_11 1 1 152 154 DCNv2 pnnx_155 5 1 147 145 144 153 154 155 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_32 1 1 155 156 0=256 1=1.000000e-05 ReLU relu_155 1 1 156 157 Convolution conv_72 1 1 157 158 0=1024 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 BinaryOp add_9 2 1 158 142 159 0=0 ReLU relu_156 1 1 159 160 Split splitncnn_17 1 2 160 161 162 MemoryData backbone.layers.2.3.conv2 0 1 163 0=256 MemoryData pnnx_unique_27 0 1 164 0=3 1=3 11=256 2=256 Convolution convrelu_14 1 1 162 165 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 9=1 Split splitncnn_18 1 2 165 166 167 Convolution conv_74 1 1 167 168 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=62208 Slice chunk_7 1 3 168 169 170 171 -23300=3,-233,-233,-233 1=0 Concat cat_7 2 1 169 170 172 0=0 Sigmoid sigmoid_12 1 1 171 173 DCNv2 pnnx_175 5 1 166 164 163 172 173 174 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_33 1 1 174 175 0=256 1=1.000000e-05 ReLU relu_158 1 1 175 176 Convolution conv_75 1 1 176 177 0=1024 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 BinaryOp add_10 2 1 177 161 178 0=0 ReLU relu_159 1 1 178 179 Split splitncnn_19 1 2 179 180 181 MemoryData backbone.layers.2.4.conv2 0 1 182 0=256 MemoryData pnnx_unique_30 0 1 183 0=3 1=3 11=256 2=256 Convolution convrelu_15 1 1 181 184 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 9=1 Split splitncnn_20 1 2 184 185 186 Convolution conv_77 1 1 186 187 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=62208 Slice chunk_8 1 3 187 188 189 190 -23300=3,-233,-233,-233 1=0 Concat cat_8 2 1 188 189 191 0=0 Sigmoid sigmoid_13 1 1 190 192 DCNv2 pnnx_195 5 1 185 183 182 191 192 193 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_34 1 1 193 194 0=256 1=1.000000e-05 ReLU relu_161 1 1 194 195 Convolution conv_78 1 1 195 196 0=1024 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 BinaryOp add_11 2 1 196 180 197 0=0 ReLU relu_162 1 1 197 198 Split splitncnn_21 1 2 198 199 200 MemoryData backbone.layers.2.5.conv2 0 1 201 0=256 MemoryData pnnx_unique_33 0 1 202 0=3 1=3 11=256 2=256 Convolution convrelu_16 1 1 200 203 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 9=1 Split splitncnn_22 1 2 203 204 205 Convolution conv_80 1 1 205 206 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=62208 Slice chunk_9 1 3 206 207 208 209 -23300=3,-233,-233,-233 1=0 Concat cat_9 2 1 207 208 210 0=0 Sigmoid sigmoid_14 1 1 209 211 DCNv2 pnnx_215 5 1 204 202 201 210 211 212 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_35 1 1 212 213 0=256 1=1.000000e-05 ReLU relu_164 1 1 213 214 Convolution conv_81 1 1 214 215 0=1024 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 BinaryOp add_12 2 1 215 199 216 0=0 ReLU relu_165 1 1 216 217 Split splitncnn_23 1 3 217 218 219 220 MemoryData backbone.layers.3.0.conv2 0 1 221 0=512 MemoryData pnnx_unique_36 0 1 222 0=3 1=3 11=512 2=512 Convolution convrelu_17 1 1 220 223 0=512 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=524288 9=1 Split splitncnn_24 1 2 223 224 225 Convolution conv_83 1 1 225 226 0=27 1=3 11=3 12=1 13=2 14=1 2=1 3=2 4=1 5=1 6=124416 Slice chunk_10 1 3 226 227 228 229 -23300=3,-233,-233,-233 1=0 Concat cat_10 2 1 227 228 230 0=0 Sigmoid sigmoid_15 1 1 229 231 DCNv2 pnnx_235 5 1 224 222 221 230 231 232 10=1 11=1 12=1 13=1 5=3 6=3 7=2 8=2 9=1 BatchNorm bn_36 1 1 232 233 0=512 1=1.000000e-05 ReLU relu_167 1 1 233 234 Convolution conv_85 1 1 234 235 0=2048 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=1048576 Convolution conv_84 1 1 219 236 0=2048 1=1 11=1 12=1 13=2 14=0 2=1 3=2 4=0 5=1 6=2097152 BinaryOp add_13 2 1 235 236 237 0=0 ReLU relu_168 1 1 237 238 Split splitncnn_25 1 2 238 239 240 MemoryData backbone.layers.3.1.conv2 0 1 241 0=512 MemoryData pnnx_unique_39 0 1 242 0=3 1=3 11=512 2=512 Convolution convrelu_18 1 1 240 243 0=512 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=1048576 9=1 Split splitncnn_26 1 2 243 244 245 Convolution conv_87 1 1 245 246 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=124416 Slice chunk_11 1 3 246 247 248 249 -23300=3,-233,-233,-233 1=0 Concat cat_11 2 1 247 248 250 0=0 Sigmoid sigmoid_16 1 1 249 251 DCNv2 pnnx_255 5 1 244 242 241 250 251 252 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_37 1 1 252 253 0=512 1=1.000000e-05 ReLU relu_170 1 1 253 254 Convolution conv_88 1 1 254 255 0=2048 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=1048576 BinaryOp add_14 2 1 255 239 256 0=0 ReLU relu_171 1 1 256 257 Split splitncnn_27 1 2 257 258 259 MemoryData backbone.layers.3.2.conv2 0 1 260 0=512 MemoryData pnnx_unique_42 0 1 261 0=3 1=3 11=512 2=512 Convolution convrelu_19 1 1 259 262 0=512 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=1048576 9=1 Split splitncnn_28 1 2 262 263 264 Convolution conv_90 1 1 264 265 0=27 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=124416 Slice chunk_12 1 3 265 266 267 268 -23300=3,-233,-233,-233 1=0 Concat cat_12 2 1 266 267 269 0=0 Sigmoid sigmoid_17 1 1 268 270 DCNv2 pnnx_275 5 1 263 261 260 269 270 271 10=1 11=1 12=1 13=1 5=3 6=3 7=1 8=1 9=1 BatchNorm bn_38 1 1 271 272 0=512 1=1.000000e-05 ReLU relu_173 1 1 272 273 Convolution conv_91 1 1 273 274 0=2048 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=1048576 BinaryOp add_15 2 1 274 258 275 0=0 ReLU relu_174 1 1 275 276 Convolution conv_92 1 1 276 277 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=524288 Split splitncnn_29 1 2 277 278 279 Interp upsample_24 1 1 278 280 0=2 3=35 4=35 6=0 Convolution conv_93 1 1 218 281 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=262144 BinaryOp add_16 2 1 280 281 282 0=0 Split splitncnn_30 1 2 282 283 284 Interp upsample_25 1 1 283 285 0=2 3=69 4=69 6=0 Convolution conv_94 1 1 102 286 0=256 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=131072 BinaryOp add_17 2 1 285 286 287 0=0 Convolution convrelu_20 1 1 279 288 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Split splitncnn_31 1 2 288 289 290 Convolution convrelu_21 1 1 287 291 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Split splitncnn_32 1 2 291 292 293 Convolution convrelu_22 1 1 293 294 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Convolution convrelu_23 1 1 294 295 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Convolution convrelu_24 1 1 295 296 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Interp interpolate_0 1 1 296 297 0=2 1=2.000000e+00 2=2.000000e+00 6=0 ReLU relu_178 1 1 297 298 Convolution convrelu_25 1 1 292 299 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Split splitncnn_33 1 3 299 300 301 302 Convolution conv_106 1 1 302 303 0=36 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=82944 Convolution conv_107 1 1 301 304 0=729 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=1679616 Convolution conv_108 1 1 300 305 0=288 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=663552 Convolution convrelu_26 1 1 284 306 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Convolution convrelu_27 1 1 306 307 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Split splitncnn_34 1 3 307 308 309 310 Convolution conv_110 1 1 310 311 0=36 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=82944 Convolution conv_111 1 1 309 312 0=729 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=1679616 Convolution conv_112 1 1 308 313 0=288 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=663552 Convolution convrelu_28 1 1 289 314 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Split splitncnn_35 1 3 314 315 316 317 Convolution conv_114 1 1 317 318 0=36 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=82944 Convolution conv_115 1 1 316 319 0=729 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=1679616 Convolution conv_116 1 1 315 320 0=288 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=663552 Convolution conv_98 1 1 290 321 0=256 1=3 11=3 12=1 13=2 14=1 2=1 3=2 4=1 5=1 6=589824 Split splitncnn_36 1 2 321 322 323 Convolution convrelu_29 1 1 322 324 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Split splitncnn_37 1 3 324 325 326 327 Convolution conv_118 1 1 327 328 0=36 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=82944 Convolution conv_119 1 1 326 329 0=729 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=1679616 Convolution conv_120 1 1 325 330 0=288 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=663552 Convolution conv_99 1 1 323 331 0=256 1=3 11=3 12=1 13=2 14=1 2=1 3=2 4=1 5=1 6=589824 Convolution convrelu_30 1 1 331 332 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Split splitncnn_38 1 3 332 333 334 335 Convolution conv_122 1 1 335 336 0=36 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=82944 Convolution conv_123 1 1 334 337 0=729 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=1679616 Convolution conv_124 1 1 333 338 0=288 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=663552 Permute permute_202 1 1 304 339 0=3 Reshape reshape_195 1 1 339 340 0=81 1=-1 Permute permute_205 1 1 312 341 0=3 Reshape reshape_194 1 1 341 342 0=81 1=-1 Permute permute_208 1 1 319 343 0=3 Reshape reshape_193 1 1 343 344 0=81 1=-1 Permute permute_211 1 1 329 345 0=3 Reshape reshape_192 1 1 345 346 0=81 1=-1 Permute permute_214 1 1 337 347 0=3 Reshape reshape_191 1 1 347 348 0=81 1=-1 Concat cat_13 5 1 340 342 344 346 348 349 0=0 Permute permute_215 1 1 338 350 0=3 Reshape reshape_190 1 1 350 351 0=32 1=-1 TanH tanh_23 1 1 351 352 Permute permute_212 1 1 330 353 0=3 Reshape reshape_189 1 1 353 354 0=32 1=-1 TanH tanh_22 1 1 354 355 Permute permute_209 1 1 320 356 0=3 Reshape reshape_188 1 1 356 357 0=32 1=-1 TanH tanh_21 1 1 357 358 Permute permute_206 1 1 313 359 0=3 Reshape reshape_187 1 1 359 360 0=32 1=-1 TanH tanh_20 1 1 360 361 Permute permute_203 1 1 305 362 0=3 Reshape reshape_186 1 1 362 363 0=32 1=-1 TanH tanh_19 1 1 363 364 Concat cat_14 5 1 364 361 358 355 352 out2 0=0 Permute permute_201 1 1 303 366 0=3 Reshape reshape_200 1 1 366 367 0=4 1=-1 Permute permute_204 1 1 311 368 0=3 Reshape reshape_199 1 1 368 369 0=4 1=-1 Permute permute_207 1 1 318 370 0=3 Reshape reshape_198 1 1 370 371 0=4 1=-1 Permute permute_210 1 1 328 372 0=3 Reshape reshape_197 1 1 372 373 0=4 1=-1 Permute permute_213 1 1 336 374 0=3 Reshape reshape_196 1 1 374 375 0=4 1=-1 Concat cat_15 5 1 367 369 371 373 375 out0 0=0 Softmax softmax_18 1 1 349 out1 0=0 1=1 Convolution convrelu_31 1 1 298 378 0=256 1=3 11=3 12=1 13=1 14=1 2=1 3=1 4=1 5=1 6=589824 9=1 Convolution convrelu_32 1 1 378 out3 0=32 1=1 11=1 12=1 13=1 14=0 2=1 3=1 4=0 5=1 6=8192 9=1
以上是param参数 Softmax softmax_18 1 1 349 out1 0=0 1=1 这个Softmax计算有上述的问题