Marat Dukhan | 4cea232 | 2021-03-09 09:35:36 -0800 | [diff] [blame] | 1 | // Copyright 2020 Google LLC |
| 2 | // |
| 3 | // This source code is licensed under the BSD-style license found in the |
| 4 | // LICENSE file in the root directory of this source tree. |
| 5 | |
| 6 | #include <xnnpack.h> |
| 7 | |
| 8 | #include <array> |
| 9 | #include <algorithm> |
| 10 | #include <functional> |
| 11 | #include <iostream> |
| 12 | #include <limits> |
| 13 | #include <random> |
| 14 | |
| 15 | #include "models/models.h" |
| 16 | |
| 17 | namespace models { |
| 18 | |
| 19 | ExecutionPlan FP32SparseMobileNetV3Large(float sparsity, pthreadpool_t threadpool) { |
| 20 | alignas(16) static std::array<float, 150528> v0; |
| 21 | alignas(16) static std::array<float, 200704> v1; |
| 22 | alignas(16) static std::array<float, 200704> v2; |
| 23 | alignas(16) static std::array<float, 200704> v3; |
| 24 | alignas(16) static std::array<float, 200704> v4; |
| 25 | alignas(16) static std::array<float, 200704> v5; |
| 26 | alignas(16) static std::array<float, 802816> v6; |
| 27 | alignas(16) static std::array<float, 200704> v7; |
| 28 | alignas(16) static std::array<float, 75264> v8; |
| 29 | alignas(16) static std::array<float, 225792> v9; |
| 30 | alignas(16) static std::array<float, 225792> v10; |
| 31 | alignas(16) static std::array<float, 75264> v11; |
| 32 | alignas(16) static std::array<float, 75264> v12; |
| 33 | alignas(16) static std::array<float, 225792> v13; |
| 34 | alignas(16) static std::array<float, 56448> v14; |
| 35 | alignas(16) static std::array<float, 72> v15; |
| 36 | alignas(16) static std::array<float, 24> v16; |
| 37 | alignas(16) static std::array<float, 72> v17; |
| 38 | alignas(16) static std::array<float, 56448> v18; |
| 39 | alignas(16) static std::array<float, 31360> v19; |
| 40 | alignas(16) static std::array<float, 94080> v20; |
| 41 | alignas(16) static std::array<float, 94080> v21; |
| 42 | alignas(16) static std::array<float, 120> v22; |
| 43 | alignas(16) static std::array<float, 32> v23; |
| 44 | alignas(16) static std::array<float, 120> v24; |
| 45 | alignas(16) static std::array<float, 94080> v25; |
| 46 | alignas(16) static std::array<float, 31360> v26; |
| 47 | alignas(16) static std::array<float, 31360> v27; |
| 48 | alignas(16) static std::array<float, 94080> v28; |
| 49 | alignas(16) static std::array<float, 94080> v29; |
| 50 | alignas(16) static std::array<float, 120> v30; |
| 51 | alignas(16) static std::array<float, 32> v31; |
| 52 | alignas(16) static std::array<float, 120> v32; |
| 53 | alignas(16) static std::array<float, 94080> v33; |
| 54 | alignas(16) static std::array<float, 31360> v34; |
| 55 | alignas(16) static std::array<float, 31360> v35; |
| 56 | alignas(16) static std::array<float, 188160> v36; |
| 57 | alignas(16) static std::array<float, 188160> v37; |
| 58 | alignas(16) static std::array<float, 47040> v38; |
| 59 | alignas(16) static std::array<float, 47040> v39; |
| 60 | alignas(16) static std::array<float, 15680> v40; |
| 61 | alignas(16) static std::array<float, 39200> v41; |
| 62 | alignas(16) static std::array<float, 39200> v42; |
| 63 | alignas(16) static std::array<float, 39200> v43; |
| 64 | alignas(16) static std::array<float, 39200> v44; |
| 65 | alignas(16) static std::array<float, 15680> v45; |
| 66 | alignas(16) static std::array<float, 15680> v46; |
| 67 | alignas(16) static std::array<float, 36064> v47; |
| 68 | alignas(16) static std::array<float, 36064> v48; |
| 69 | alignas(16) static std::array<float, 36064> v49; |
| 70 | alignas(16) static std::array<float, 36064> v50; |
| 71 | alignas(16) static std::array<float, 15680> v51; |
| 72 | alignas(16) static std::array<float, 15680> v52; |
| 73 | alignas(16) static std::array<float, 36064> v53; |
| 74 | alignas(16) static std::array<float, 36064> v54; |
| 75 | alignas(16) static std::array<float, 36064> v55; |
| 76 | alignas(16) static std::array<float, 36064> v56; |
| 77 | alignas(16) static std::array<float, 15680> v57; |
| 78 | alignas(16) static std::array<float, 15680> v58; |
| 79 | alignas(16) static std::array<float, 94080> v59; |
| 80 | alignas(16) static std::array<float, 94080> v60; |
| 81 | alignas(16) static std::array<float, 94080> v61; |
| 82 | alignas(16) static std::array<float, 94080> v62; |
| 83 | alignas(16) static std::array<float, 480> v63; |
| 84 | alignas(16) static std::array<float, 120> v64; |
| 85 | alignas(16) static std::array<float, 480> v65; |
| 86 | alignas(16) static std::array<float, 94080> v66; |
| 87 | alignas(16) static std::array<float, 21952> v67; |
| 88 | alignas(16) static std::array<float, 131712> v68; |
| 89 | alignas(16) static std::array<float, 131712> v69; |
| 90 | alignas(16) static std::array<float, 131712> v70; |
| 91 | alignas(16) static std::array<float, 131712> v71; |
| 92 | alignas(16) static std::array<float, 672> v72; |
| 93 | alignas(16) static std::array<float, 168> v73; |
| 94 | alignas(16) static std::array<float, 672> v74; |
| 95 | alignas(16) static std::array<float, 131712> v75; |
| 96 | alignas(16) static std::array<float, 21952> v76; |
| 97 | alignas(16) static std::array<float, 21952> v77; |
| 98 | alignas(16) static std::array<float, 131712> v78; |
| 99 | alignas(16) static std::array<float, 131712> v79; |
| 100 | alignas(16) static std::array<float, 32928> v80; |
| 101 | alignas(16) static std::array<float, 32928> v81; |
| 102 | alignas(16) static std::array<float, 672> v82; |
| 103 | alignas(16) static std::array<float, 168> v83; |
| 104 | alignas(16) static std::array<float, 672> v84; |
| 105 | alignas(16) static std::array<float, 32928> v85; |
| 106 | alignas(16) static std::array<float, 7840> v86; |
| 107 | alignas(16) static std::array<float, 47040> v87; |
| 108 | alignas(16) static std::array<float, 47040> v88; |
| 109 | alignas(16) static std::array<float, 47040> v89; |
| 110 | alignas(16) static std::array<float, 47040> v90; |
| 111 | alignas(16) static std::array<float, 960> v91; |
| 112 | alignas(16) static std::array<float, 240> v92; |
| 113 | alignas(16) static std::array<float, 960> v93; |
| 114 | alignas(16) static std::array<float, 47040> v94; |
| 115 | alignas(16) static std::array<float, 7840> v95; |
| 116 | alignas(16) static std::array<float, 7840> v96; |
| 117 | alignas(16) static std::array<float, 47040> v97; |
| 118 | alignas(16) static std::array<float, 47040> v98; |
| 119 | alignas(16) static std::array<float, 47040> v99; |
| 120 | alignas(16) static std::array<float, 47040> v100; |
| 121 | alignas(16) static std::array<float, 960> v101; |
| 122 | alignas(16) static std::array<float, 240> v102; |
| 123 | alignas(16) static std::array<float, 960> v103; |
| 124 | alignas(16) static std::array<float, 47040> v104; |
| 125 | alignas(16) static std::array<float, 7840> v105; |
| 126 | alignas(16) static std::array<float, 7840> v106; |
| 127 | alignas(16) static std::array<float, 47040> v107; |
| 128 | alignas(16) static std::array<float, 47040> v108; |
| 129 | alignas(16) static std::array<float, 960> v109; |
| 130 | alignas(16) static std::array<float, 1280> v110; |
| 131 | alignas(16) static std::array<float, 1280> v111; |
| 132 | alignas(16) static std::array<float, 1280> v112; |
| 133 | alignas(16) static std::array<float, 1001> v113; |
| 134 | alignas(16) static std::array<float, 432> w114; |
| 135 | alignas(16) static std::array<float, 16> w115; |
| 136 | alignas(16) static std::array<float, 144> w116; |
| 137 | alignas(16) static std::array<float, 16> w117; |
| 138 | alignas(16) static std::array<float, 256> w118; |
| 139 | alignas(16) static std::array<float, 16> w119; |
| 140 | alignas(16) static std::array<float, 1024> w120; |
| 141 | alignas(16) static std::array<float, 64> w121; |
| 142 | alignas(16) static std::array<float, 576> w122; |
| 143 | alignas(16) static std::array<float, 64> w123; |
| 144 | alignas(16) static std::array<float, 1536> w124; |
| 145 | alignas(16) static std::array<float, 24> w125; |
| 146 | alignas(16) static std::array<float, 1728> w126; |
| 147 | alignas(16) static std::array<float, 72> w127; |
| 148 | alignas(16) static std::array<float, 648> w128; |
| 149 | alignas(16) static std::array<float, 72> w129; |
| 150 | alignas(16) static std::array<float, 1728> w130; |
| 151 | alignas(16) static std::array<float, 24> w131; |
| 152 | alignas(16) static std::array<float, 1728> w132; |
| 153 | alignas(16) static std::array<float, 72> w133; |
| 154 | alignas(16) static std::array<float, 1800> w134; |
| 155 | alignas(16) static std::array<float, 72> w135; |
| 156 | alignas(16) static std::array<float, 1728> w136; |
| 157 | alignas(16) static std::array<float, 24> w137; |
| 158 | alignas(16) static std::array<float, 1728> w138; |
| 159 | alignas(16) static std::array<float, 72> w139; |
| 160 | alignas(16) static std::array<float, 2880> w140; |
| 161 | alignas(16) static std::array<float, 40> w141; |
| 162 | alignas(16) static std::array<float, 4800> w142; |
| 163 | alignas(16) static std::array<float, 120> w143; |
| 164 | alignas(16) static std::array<float, 3000> w144; |
| 165 | alignas(16) static std::array<float, 120> w145; |
| 166 | alignas(16) static std::array<float, 3840> w146; |
| 167 | alignas(16) static std::array<float, 32> w147; |
| 168 | alignas(16) static std::array<float, 3840> w148; |
| 169 | alignas(16) static std::array<float, 120> w149; |
| 170 | alignas(16) static std::array<float, 4800> w150; |
| 171 | alignas(16) static std::array<float, 40> w151; |
| 172 | alignas(16) static std::array<float, 4800> w152; |
| 173 | alignas(16) static std::array<float, 120> w153; |
| 174 | alignas(16) static std::array<float, 3000> w154; |
| 175 | alignas(16) static std::array<float, 120> w155; |
| 176 | alignas(16) static std::array<float, 3840> w156; |
| 177 | alignas(16) static std::array<float, 32> w157; |
| 178 | alignas(16) static std::array<float, 3840> w158; |
| 179 | alignas(16) static std::array<float, 120> w159; |
| 180 | alignas(16) static std::array<float, 4800> w160; |
| 181 | alignas(16) static std::array<float, 40> w161; |
| 182 | alignas(16) static std::array<float, 9600> w162; |
| 183 | alignas(16) static std::array<float, 240> w163; |
| 184 | alignas(16) static std::array<float, 2160> w164; |
| 185 | alignas(16) static std::array<float, 240> w165; |
| 186 | alignas(16) static std::array<float, 19200> w166; |
| 187 | alignas(16) static std::array<float, 80> w167; |
| 188 | alignas(16) static std::array<float, 16000> w168; |
| 189 | alignas(16) static std::array<float, 200> w169; |
| 190 | alignas(16) static std::array<float, 1800> w170; |
| 191 | alignas(16) static std::array<float, 200> w171; |
| 192 | alignas(16) static std::array<float, 16000> w172; |
| 193 | alignas(16) static std::array<float, 80> w173; |
| 194 | alignas(16) static std::array<float, 14720> w174; |
| 195 | alignas(16) static std::array<float, 184> w175; |
| 196 | alignas(16) static std::array<float, 1656> w176; |
| 197 | alignas(16) static std::array<float, 184> w177; |
| 198 | alignas(16) static std::array<float, 14720> w178; |
| 199 | alignas(16) static std::array<float, 80> w179; |
| 200 | alignas(16) static std::array<float, 14720> w180; |
| 201 | alignas(16) static std::array<float, 184> w181; |
| 202 | alignas(16) static std::array<float, 1656> w182; |
| 203 | alignas(16) static std::array<float, 184> w183; |
| 204 | alignas(16) static std::array<float, 14720> w184; |
| 205 | alignas(16) static std::array<float, 80> w185; |
| 206 | alignas(16) static std::array<float, 38400> w186; |
| 207 | alignas(16) static std::array<float, 480> w187; |
| 208 | alignas(16) static std::array<float, 4320> w188; |
| 209 | alignas(16) static std::array<float, 480> w189; |
| 210 | alignas(16) static std::array<float, 57600> w190; |
| 211 | alignas(16) static std::array<float, 120> w191; |
| 212 | alignas(16) static std::array<float, 57600> w192; |
| 213 | alignas(16) static std::array<float, 480> w193; |
| 214 | alignas(16) static std::array<float, 53760> w194; |
| 215 | alignas(16) static std::array<float, 112> w195; |
| 216 | alignas(16) static std::array<float, 75264> w196; |
| 217 | alignas(16) static std::array<float, 672> w197; |
| 218 | alignas(16) static std::array<float, 6048> w198; |
| 219 | alignas(16) static std::array<float, 672> w199; |
| 220 | alignas(16) static std::array<float, 112896> w200; |
| 221 | alignas(16) static std::array<float, 168> w201; |
| 222 | alignas(16) static std::array<float, 112896> w202; |
| 223 | alignas(16) static std::array<float, 672> w203; |
| 224 | alignas(16) static std::array<float, 75264> w204; |
| 225 | alignas(16) static std::array<float, 112> w205; |
| 226 | alignas(16) static std::array<float, 75264> w206; |
| 227 | alignas(16) static std::array<float, 672> w207; |
| 228 | alignas(16) static std::array<float, 16800> w208; |
| 229 | alignas(16) static std::array<float, 672> w209; |
| 230 | alignas(16) static std::array<float, 112896> w210; |
| 231 | alignas(16) static std::array<float, 168> w211; |
| 232 | alignas(16) static std::array<float, 112896> w212; |
| 233 | alignas(16) static std::array<float, 672> w213; |
| 234 | alignas(16) static std::array<float, 107520> w214; |
| 235 | alignas(16) static std::array<float, 160> w215; |
| 236 | alignas(16) static std::array<float, 153600> w216; |
| 237 | alignas(16) static std::array<float, 960> w217; |
| 238 | alignas(16) static std::array<float, 24000> w218; |
| 239 | alignas(16) static std::array<float, 960> w219; |
| 240 | alignas(16) static std::array<float, 230400> w220; |
| 241 | alignas(16) static std::array<float, 240> w221; |
| 242 | alignas(16) static std::array<float, 230400> w222; |
| 243 | alignas(16) static std::array<float, 960> w223; |
| 244 | alignas(16) static std::array<float, 153600> w224; |
| 245 | alignas(16) static std::array<float, 160> w225; |
| 246 | alignas(16) static std::array<float, 153600> w226; |
| 247 | alignas(16) static std::array<float, 960> w227; |
| 248 | alignas(16) static std::array<float, 24000> w228; |
| 249 | alignas(16) static std::array<float, 960> w229; |
| 250 | alignas(16) static std::array<float, 230400> w230; |
| 251 | alignas(16) static std::array<float, 240> w231; |
| 252 | alignas(16) static std::array<float, 230400> w232; |
| 253 | alignas(16) static std::array<float, 960> w233; |
| 254 | alignas(16) static std::array<float, 153600> w234; |
| 255 | alignas(16) static std::array<float, 160> w235; |
| 256 | alignas(16) static std::array<float, 153600> w236; |
| 257 | alignas(16) static std::array<float, 960> w237; |
| 258 | alignas(16) static std::array<float, 1228800> w238; |
| 259 | alignas(16) static std::array<float, 1280> w239; |
| 260 | alignas(16) static std::array<float, 1281280> w240; |
| 261 | alignas(16) static std::array<float, 1001> w241; |
| 262 | |
| 263 | std::random_device random_device; |
| 264 | auto rng = std::mt19937(random_device()); |
| 265 | auto f32rng = std::bind(std::uniform_real_distribution<float>(-1.0f, +1.0f), std::ref(rng)); |
| 266 | std::generate(v0.begin(), v0.end(), std::ref(f32rng)); |
| 267 | std::generate(v1.begin(), v1.end(), std::ref(f32rng)); |
| 268 | std::generate(v2.begin(), v2.end(), std::ref(f32rng)); |
| 269 | std::generate(v3.begin(), v3.end(), std::ref(f32rng)); |
| 270 | std::generate(v4.begin(), v4.end(), std::ref(f32rng)); |
| 271 | std::generate(v5.begin(), v5.end(), std::ref(f32rng)); |
| 272 | std::generate(v6.begin(), v6.end(), std::ref(f32rng)); |
| 273 | std::generate(v7.begin(), v7.end(), std::ref(f32rng)); |
| 274 | std::generate(v8.begin(), v8.end(), std::ref(f32rng)); |
| 275 | std::generate(v9.begin(), v9.end(), std::ref(f32rng)); |
| 276 | std::generate(v10.begin(), v10.end(), std::ref(f32rng)); |
| 277 | std::generate(v11.begin(), v11.end(), std::ref(f32rng)); |
| 278 | std::generate(v12.begin(), v12.end(), std::ref(f32rng)); |
| 279 | std::generate(v13.begin(), v13.end(), std::ref(f32rng)); |
| 280 | std::generate(v14.begin(), v14.end(), std::ref(f32rng)); |
| 281 | std::generate(v15.begin(), v15.end(), std::ref(f32rng)); |
| 282 | std::generate(v16.begin(), v16.end(), std::ref(f32rng)); |
| 283 | std::generate(v17.begin(), v17.end(), std::ref(f32rng)); |
| 284 | std::generate(v18.begin(), v18.end(), std::ref(f32rng)); |
| 285 | std::generate(v19.begin(), v19.end(), std::ref(f32rng)); |
| 286 | std::generate(v20.begin(), v20.end(), std::ref(f32rng)); |
| 287 | std::generate(v21.begin(), v21.end(), std::ref(f32rng)); |
| 288 | std::generate(v22.begin(), v22.end(), std::ref(f32rng)); |
| 289 | std::generate(v23.begin(), v23.end(), std::ref(f32rng)); |
| 290 | std::generate(v24.begin(), v24.end(), std::ref(f32rng)); |
| 291 | std::generate(v25.begin(), v25.end(), std::ref(f32rng)); |
| 292 | std::generate(v26.begin(), v26.end(), std::ref(f32rng)); |
| 293 | std::generate(v27.begin(), v27.end(), std::ref(f32rng)); |
| 294 | std::generate(v28.begin(), v28.end(), std::ref(f32rng)); |
| 295 | std::generate(v29.begin(), v29.end(), std::ref(f32rng)); |
| 296 | std::generate(v30.begin(), v30.end(), std::ref(f32rng)); |
| 297 | std::generate(v31.begin(), v31.end(), std::ref(f32rng)); |
| 298 | std::generate(v32.begin(), v32.end(), std::ref(f32rng)); |
| 299 | std::generate(v33.begin(), v33.end(), std::ref(f32rng)); |
| 300 | std::generate(v34.begin(), v34.end(), std::ref(f32rng)); |
| 301 | std::generate(v35.begin(), v35.end(), std::ref(f32rng)); |
| 302 | std::generate(v36.begin(), v36.end(), std::ref(f32rng)); |
| 303 | std::generate(v37.begin(), v37.end(), std::ref(f32rng)); |
| 304 | std::generate(v38.begin(), v38.end(), std::ref(f32rng)); |
| 305 | std::generate(v39.begin(), v39.end(), std::ref(f32rng)); |
| 306 | std::generate(v40.begin(), v40.end(), std::ref(f32rng)); |
| 307 | std::generate(v41.begin(), v41.end(), std::ref(f32rng)); |
| 308 | std::generate(v42.begin(), v42.end(), std::ref(f32rng)); |
| 309 | std::generate(v43.begin(), v43.end(), std::ref(f32rng)); |
| 310 | std::generate(v44.begin(), v44.end(), std::ref(f32rng)); |
| 311 | std::generate(v45.begin(), v45.end(), std::ref(f32rng)); |
| 312 | std::generate(v46.begin(), v46.end(), std::ref(f32rng)); |
| 313 | std::generate(v47.begin(), v47.end(), std::ref(f32rng)); |
| 314 | std::generate(v48.begin(), v48.end(), std::ref(f32rng)); |
| 315 | std::generate(v49.begin(), v49.end(), std::ref(f32rng)); |
| 316 | std::generate(v50.begin(), v50.end(), std::ref(f32rng)); |
| 317 | std::generate(v51.begin(), v51.end(), std::ref(f32rng)); |
| 318 | std::generate(v52.begin(), v52.end(), std::ref(f32rng)); |
| 319 | std::generate(v53.begin(), v53.end(), std::ref(f32rng)); |
| 320 | std::generate(v54.begin(), v54.end(), std::ref(f32rng)); |
| 321 | std::generate(v55.begin(), v55.end(), std::ref(f32rng)); |
| 322 | std::generate(v56.begin(), v56.end(), std::ref(f32rng)); |
| 323 | std::generate(v57.begin(), v57.end(), std::ref(f32rng)); |
| 324 | std::generate(v58.begin(), v58.end(), std::ref(f32rng)); |
| 325 | std::generate(v59.begin(), v59.end(), std::ref(f32rng)); |
| 326 | std::generate(v60.begin(), v60.end(), std::ref(f32rng)); |
| 327 | std::generate(v61.begin(), v61.end(), std::ref(f32rng)); |
| 328 | std::generate(v62.begin(), v62.end(), std::ref(f32rng)); |
| 329 | std::generate(v63.begin(), v63.end(), std::ref(f32rng)); |
| 330 | std::generate(v64.begin(), v64.end(), std::ref(f32rng)); |
| 331 | std::generate(v65.begin(), v65.end(), std::ref(f32rng)); |
| 332 | std::generate(v66.begin(), v66.end(), std::ref(f32rng)); |
| 333 | std::generate(v67.begin(), v67.end(), std::ref(f32rng)); |
| 334 | std::generate(v68.begin(), v68.end(), std::ref(f32rng)); |
| 335 | std::generate(v69.begin(), v69.end(), std::ref(f32rng)); |
| 336 | std::generate(v70.begin(), v70.end(), std::ref(f32rng)); |
| 337 | std::generate(v71.begin(), v71.end(), std::ref(f32rng)); |
| 338 | std::generate(v72.begin(), v72.end(), std::ref(f32rng)); |
| 339 | std::generate(v73.begin(), v73.end(), std::ref(f32rng)); |
| 340 | std::generate(v74.begin(), v74.end(), std::ref(f32rng)); |
| 341 | std::generate(v75.begin(), v75.end(), std::ref(f32rng)); |
| 342 | std::generate(v76.begin(), v76.end(), std::ref(f32rng)); |
| 343 | std::generate(v77.begin(), v77.end(), std::ref(f32rng)); |
| 344 | std::generate(v78.begin(), v78.end(), std::ref(f32rng)); |
| 345 | std::generate(v79.begin(), v79.end(), std::ref(f32rng)); |
| 346 | std::generate(v80.begin(), v80.end(), std::ref(f32rng)); |
| 347 | std::generate(v81.begin(), v81.end(), std::ref(f32rng)); |
| 348 | std::generate(v82.begin(), v82.end(), std::ref(f32rng)); |
| 349 | std::generate(v83.begin(), v83.end(), std::ref(f32rng)); |
| 350 | std::generate(v84.begin(), v84.end(), std::ref(f32rng)); |
| 351 | std::generate(v85.begin(), v85.end(), std::ref(f32rng)); |
| 352 | std::generate(v86.begin(), v86.end(), std::ref(f32rng)); |
| 353 | std::generate(v87.begin(), v87.end(), std::ref(f32rng)); |
| 354 | std::generate(v88.begin(), v88.end(), std::ref(f32rng)); |
| 355 | std::generate(v89.begin(), v89.end(), std::ref(f32rng)); |
| 356 | std::generate(v90.begin(), v90.end(), std::ref(f32rng)); |
| 357 | std::generate(v91.begin(), v91.end(), std::ref(f32rng)); |
| 358 | std::generate(v92.begin(), v92.end(), std::ref(f32rng)); |
| 359 | std::generate(v93.begin(), v93.end(), std::ref(f32rng)); |
| 360 | std::generate(v94.begin(), v94.end(), std::ref(f32rng)); |
| 361 | std::generate(v95.begin(), v95.end(), std::ref(f32rng)); |
| 362 | std::generate(v96.begin(), v96.end(), std::ref(f32rng)); |
| 363 | std::generate(v97.begin(), v97.end(), std::ref(f32rng)); |
| 364 | std::generate(v98.begin(), v98.end(), std::ref(f32rng)); |
| 365 | std::generate(v99.begin(), v99.end(), std::ref(f32rng)); |
| 366 | std::generate(v100.begin(), v100.end(), std::ref(f32rng)); |
| 367 | std::generate(v101.begin(), v101.end(), std::ref(f32rng)); |
| 368 | std::generate(v102.begin(), v102.end(), std::ref(f32rng)); |
| 369 | std::generate(v103.begin(), v103.end(), std::ref(f32rng)); |
| 370 | std::generate(v104.begin(), v104.end(), std::ref(f32rng)); |
| 371 | std::generate(v105.begin(), v105.end(), std::ref(f32rng)); |
| 372 | std::generate(v106.begin(), v106.end(), std::ref(f32rng)); |
| 373 | std::generate(v107.begin(), v107.end(), std::ref(f32rng)); |
| 374 | std::generate(v108.begin(), v108.end(), std::ref(f32rng)); |
| 375 | std::generate(v109.begin(), v109.end(), std::ref(f32rng)); |
| 376 | std::generate(v110.begin(), v110.end(), std::ref(f32rng)); |
| 377 | std::generate(v111.begin(), v111.end(), std::ref(f32rng)); |
| 378 | std::generate(v112.begin(), v112.end(), std::ref(f32rng)); |
| 379 | std::generate(v113.begin(), v113.end(), std::ref(f32rng)); |
| 380 | std::generate(w114.begin(), w114.end(), std::ref(f32rng)); |
| 381 | std::generate(w115.begin(), w115.end(), std::ref(f32rng)); |
| 382 | std::generate(w116.begin(), w116.end(), std::ref(f32rng)); |
| 383 | std::generate(w117.begin(), w117.end(), std::ref(f32rng)); |
| 384 | std::fill(w118.begin(), w118.end(), 0.0f); |
| 385 | std::generate(w118.begin(), w118.end() - size_t(sparsity * w118.size()), std::ref(f32rng)); |
| 386 | std::shuffle(w118.begin(), w118.end(), rng); |
| 387 | std::generate(w119.begin(), w119.end(), std::ref(f32rng)); |
| 388 | std::fill(w120.begin(), w120.end(), 0.0f); |
| 389 | std::generate(w120.begin(), w120.end() - size_t(sparsity * w120.size()), std::ref(f32rng)); |
| 390 | std::shuffle(w120.begin(), w120.end(), rng); |
| 391 | std::generate(w121.begin(), w121.end(), std::ref(f32rng)); |
| 392 | std::generate(w122.begin(), w122.end(), std::ref(f32rng)); |
| 393 | std::generate(w123.begin(), w123.end(), std::ref(f32rng)); |
| 394 | std::fill(w124.begin(), w124.end(), 0.0f); |
| 395 | std::generate(w124.begin(), w124.end() - size_t(sparsity * w124.size()), std::ref(f32rng)); |
| 396 | std::shuffle(w124.begin(), w124.end(), rng); |
| 397 | std::generate(w125.begin(), w125.end(), std::ref(f32rng)); |
| 398 | std::fill(w126.begin(), w126.end(), 0.0f); |
| 399 | std::generate(w126.begin(), w126.end() - size_t(sparsity * w126.size()), std::ref(f32rng)); |
| 400 | std::shuffle(w126.begin(), w126.end(), rng); |
| 401 | std::generate(w127.begin(), w127.end(), std::ref(f32rng)); |
| 402 | std::generate(w128.begin(), w128.end(), std::ref(f32rng)); |
| 403 | std::generate(w129.begin(), w129.end(), std::ref(f32rng)); |
| 404 | std::fill(w130.begin(), w130.end(), 0.0f); |
| 405 | std::generate(w130.begin(), w130.end() - size_t(sparsity * w130.size()), std::ref(f32rng)); |
| 406 | std::shuffle(w130.begin(), w130.end(), rng); |
| 407 | std::generate(w131.begin(), w131.end(), std::ref(f32rng)); |
| 408 | std::fill(w132.begin(), w132.end(), 0.0f); |
| 409 | std::generate(w132.begin(), w132.end() - size_t(sparsity * w132.size()), std::ref(f32rng)); |
| 410 | std::shuffle(w132.begin(), w132.end(), rng); |
| 411 | std::generate(w133.begin(), w133.end(), std::ref(f32rng)); |
| 412 | std::generate(w134.begin(), w134.end(), std::ref(f32rng)); |
| 413 | std::generate(w135.begin(), w135.end(), std::ref(f32rng)); |
| 414 | std::fill(w136.begin(), w136.end(), 0.0f); |
| 415 | std::generate(w136.begin(), w136.end() - size_t(sparsity * w136.size()), std::ref(f32rng)); |
| 416 | std::shuffle(w136.begin(), w136.end(), rng); |
| 417 | std::generate(w137.begin(), w137.end(), std::ref(f32rng)); |
| 418 | std::fill(w138.begin(), w138.end(), 0.0f); |
| 419 | std::generate(w138.begin(), w138.end() - size_t(sparsity * w138.size()), std::ref(f32rng)); |
| 420 | std::shuffle(w138.begin(), w138.end(), rng); |
| 421 | std::generate(w139.begin(), w139.end(), std::ref(f32rng)); |
| 422 | std::fill(w140.begin(), w140.end(), 0.0f); |
| 423 | std::generate(w140.begin(), w140.end() - size_t(sparsity * w140.size()), std::ref(f32rng)); |
| 424 | std::shuffle(w140.begin(), w140.end(), rng); |
| 425 | std::generate(w141.begin(), w141.end(), std::ref(f32rng)); |
| 426 | std::fill(w142.begin(), w142.end(), 0.0f); |
| 427 | std::generate(w142.begin(), w142.end() - size_t(sparsity * w142.size()), std::ref(f32rng)); |
| 428 | std::shuffle(w142.begin(), w142.end(), rng); |
| 429 | std::generate(w143.begin(), w143.end(), std::ref(f32rng)); |
| 430 | std::generate(w144.begin(), w144.end(), std::ref(f32rng)); |
| 431 | std::generate(w145.begin(), w145.end(), std::ref(f32rng)); |
| 432 | std::fill(w146.begin(), w146.end(), 0.0f); |
| 433 | std::generate(w146.begin(), w146.end() - size_t(sparsity * w146.size()), std::ref(f32rng)); |
| 434 | std::shuffle(w146.begin(), w146.end(), rng); |
| 435 | std::generate(w147.begin(), w147.end(), std::ref(f32rng)); |
| 436 | std::fill(w148.begin(), w148.end(), 0.0f); |
| 437 | std::generate(w148.begin(), w148.end() - size_t(sparsity * w148.size()), std::ref(f32rng)); |
| 438 | std::shuffle(w148.begin(), w148.end(), rng); |
| 439 | std::generate(w149.begin(), w149.end(), std::ref(f32rng)); |
| 440 | std::fill(w150.begin(), w150.end(), 0.0f); |
| 441 | std::generate(w150.begin(), w150.end() - size_t(sparsity * w150.size()), std::ref(f32rng)); |
| 442 | std::shuffle(w150.begin(), w150.end(), rng); |
| 443 | std::generate(w151.begin(), w151.end(), std::ref(f32rng)); |
| 444 | std::fill(w152.begin(), w152.end(), 0.0f); |
| 445 | std::generate(w152.begin(), w152.end() - size_t(sparsity * w152.size()), std::ref(f32rng)); |
| 446 | std::shuffle(w152.begin(), w152.end(), rng); |
| 447 | std::generate(w153.begin(), w153.end(), std::ref(f32rng)); |
| 448 | std::generate(w154.begin(), w154.end(), std::ref(f32rng)); |
| 449 | std::generate(w155.begin(), w155.end(), std::ref(f32rng)); |
| 450 | std::fill(w156.begin(), w156.end(), 0.0f); |
| 451 | std::generate(w156.begin(), w156.end() - size_t(sparsity * w156.size()), std::ref(f32rng)); |
| 452 | std::shuffle(w156.begin(), w156.end(), rng); |
| 453 | std::generate(w157.begin(), w157.end(), std::ref(f32rng)); |
| 454 | std::fill(w158.begin(), w158.end(), 0.0f); |
| 455 | std::generate(w158.begin(), w158.end() - size_t(sparsity * w158.size()), std::ref(f32rng)); |
| 456 | std::shuffle(w158.begin(), w158.end(), rng); |
| 457 | std::generate(w159.begin(), w159.end(), std::ref(f32rng)); |
| 458 | std::fill(w160.begin(), w160.end(), 0.0f); |
| 459 | std::generate(w160.begin(), w160.end() - size_t(sparsity * w160.size()), std::ref(f32rng)); |
| 460 | std::shuffle(w160.begin(), w160.end(), rng); |
| 461 | std::generate(w161.begin(), w161.end(), std::ref(f32rng)); |
| 462 | std::fill(w162.begin(), w162.end(), 0.0f); |
| 463 | std::generate(w162.begin(), w162.end() - size_t(sparsity * w162.size()), std::ref(f32rng)); |
| 464 | std::shuffle(w162.begin(), w162.end(), rng); |
| 465 | std::generate(w163.begin(), w163.end(), std::ref(f32rng)); |
| 466 | std::generate(w164.begin(), w164.end(), std::ref(f32rng)); |
| 467 | std::generate(w165.begin(), w165.end(), std::ref(f32rng)); |
| 468 | std::fill(w166.begin(), w166.end(), 0.0f); |
| 469 | std::generate(w166.begin(), w166.end() - size_t(sparsity * w166.size()), std::ref(f32rng)); |
| 470 | std::shuffle(w166.begin(), w166.end(), rng); |
| 471 | std::generate(w167.begin(), w167.end(), std::ref(f32rng)); |
| 472 | std::fill(w168.begin(), w168.end(), 0.0f); |
| 473 | std::generate(w168.begin(), w168.end() - size_t(sparsity * w168.size()), std::ref(f32rng)); |
| 474 | std::shuffle(w168.begin(), w168.end(), rng); |
| 475 | std::generate(w169.begin(), w169.end(), std::ref(f32rng)); |
| 476 | std::generate(w170.begin(), w170.end(), std::ref(f32rng)); |
| 477 | std::generate(w171.begin(), w171.end(), std::ref(f32rng)); |
| 478 | std::fill(w172.begin(), w172.end(), 0.0f); |
| 479 | std::generate(w172.begin(), w172.end() - size_t(sparsity * w172.size()), std::ref(f32rng)); |
| 480 | std::shuffle(w172.begin(), w172.end(), rng); |
| 481 | std::generate(w173.begin(), w173.end(), std::ref(f32rng)); |
| 482 | std::fill(w174.begin(), w174.end(), 0.0f); |
| 483 | std::generate(w174.begin(), w174.end() - size_t(sparsity * w174.size()), std::ref(f32rng)); |
| 484 | std::shuffle(w174.begin(), w174.end(), rng); |
| 485 | std::generate(w175.begin(), w175.end(), std::ref(f32rng)); |
| 486 | std::generate(w176.begin(), w176.end(), std::ref(f32rng)); |
| 487 | std::generate(w177.begin(), w177.end(), std::ref(f32rng)); |
| 488 | std::fill(w178.begin(), w178.end(), 0.0f); |
| 489 | std::generate(w178.begin(), w178.end() - size_t(sparsity * w178.size()), std::ref(f32rng)); |
| 490 | std::shuffle(w178.begin(), w178.end(), rng); |
| 491 | std::generate(w179.begin(), w179.end(), std::ref(f32rng)); |
| 492 | std::fill(w180.begin(), w180.end(), 0.0f); |
| 493 | std::generate(w180.begin(), w180.end() - size_t(sparsity * w180.size()), std::ref(f32rng)); |
| 494 | std::shuffle(w180.begin(), w180.end(), rng); |
| 495 | std::generate(w181.begin(), w181.end(), std::ref(f32rng)); |
| 496 | std::generate(w182.begin(), w182.end(), std::ref(f32rng)); |
| 497 | std::generate(w183.begin(), w183.end(), std::ref(f32rng)); |
| 498 | std::fill(w184.begin(), w184.end(), 0.0f); |
| 499 | std::generate(w184.begin(), w184.end() - size_t(sparsity * w184.size()), std::ref(f32rng)); |
| 500 | std::shuffle(w184.begin(), w184.end(), rng); |
| 501 | std::generate(w185.begin(), w185.end(), std::ref(f32rng)); |
| 502 | std::fill(w186.begin(), w186.end(), 0.0f); |
| 503 | std::generate(w186.begin(), w186.end() - size_t(sparsity * w186.size()), std::ref(f32rng)); |
| 504 | std::shuffle(w186.begin(), w186.end(), rng); |
| 505 | std::generate(w187.begin(), w187.end(), std::ref(f32rng)); |
| 506 | std::generate(w188.begin(), w188.end(), std::ref(f32rng)); |
| 507 | std::generate(w189.begin(), w189.end(), std::ref(f32rng)); |
| 508 | std::fill(w190.begin(), w190.end(), 0.0f); |
| 509 | std::generate(w190.begin(), w190.end() - size_t(sparsity * w190.size()), std::ref(f32rng)); |
| 510 | std::shuffle(w190.begin(), w190.end(), rng); |
| 511 | std::generate(w191.begin(), w191.end(), std::ref(f32rng)); |
| 512 | std::fill(w192.begin(), w192.end(), 0.0f); |
| 513 | std::generate(w192.begin(), w192.end() - size_t(sparsity * w192.size()), std::ref(f32rng)); |
| 514 | std::shuffle(w192.begin(), w192.end(), rng); |
| 515 | std::generate(w193.begin(), w193.end(), std::ref(f32rng)); |
| 516 | std::fill(w194.begin(), w194.end(), 0.0f); |
| 517 | std::generate(w194.begin(), w194.end() - size_t(sparsity * w194.size()), std::ref(f32rng)); |
| 518 | std::shuffle(w194.begin(), w194.end(), rng); |
| 519 | std::generate(w195.begin(), w195.end(), std::ref(f32rng)); |
| 520 | std::fill(w196.begin(), w196.end(), 0.0f); |
| 521 | std::generate(w196.begin(), w196.end() - size_t(sparsity * w196.size()), std::ref(f32rng)); |
| 522 | std::shuffle(w196.begin(), w196.end(), rng); |
| 523 | std::generate(w197.begin(), w197.end(), std::ref(f32rng)); |
| 524 | std::generate(w198.begin(), w198.end(), std::ref(f32rng)); |
| 525 | std::generate(w199.begin(), w199.end(), std::ref(f32rng)); |
| 526 | std::fill(w200.begin(), w200.end(), 0.0f); |
| 527 | std::generate(w200.begin(), w200.end() - size_t(sparsity * w200.size()), std::ref(f32rng)); |
| 528 | std::shuffle(w200.begin(), w200.end(), rng); |
| 529 | std::generate(w201.begin(), w201.end(), std::ref(f32rng)); |
| 530 | std::fill(w202.begin(), w202.end(), 0.0f); |
| 531 | std::generate(w202.begin(), w202.end() - size_t(sparsity * w202.size()), std::ref(f32rng)); |
| 532 | std::shuffle(w202.begin(), w202.end(), rng); |
| 533 | std::generate(w203.begin(), w203.end(), std::ref(f32rng)); |
| 534 | std::fill(w204.begin(), w204.end(), 0.0f); |
| 535 | std::generate(w204.begin(), w204.end() - size_t(sparsity * w204.size()), std::ref(f32rng)); |
| 536 | std::shuffle(w204.begin(), w204.end(), rng); |
| 537 | std::generate(w205.begin(), w205.end(), std::ref(f32rng)); |
| 538 | std::fill(w206.begin(), w206.end(), 0.0f); |
| 539 | std::generate(w206.begin(), w206.end() - size_t(sparsity * w206.size()), std::ref(f32rng)); |
| 540 | std::shuffle(w206.begin(), w206.end(), rng); |
| 541 | std::generate(w207.begin(), w207.end(), std::ref(f32rng)); |
| 542 | std::generate(w208.begin(), w208.end(), std::ref(f32rng)); |
| 543 | std::generate(w209.begin(), w209.end(), std::ref(f32rng)); |
| 544 | std::fill(w210.begin(), w210.end(), 0.0f); |
| 545 | std::generate(w210.begin(), w210.end() - size_t(sparsity * w210.size()), std::ref(f32rng)); |
| 546 | std::shuffle(w210.begin(), w210.end(), rng); |
| 547 | std::generate(w211.begin(), w211.end(), std::ref(f32rng)); |
| 548 | std::fill(w212.begin(), w212.end(), 0.0f); |
| 549 | std::generate(w212.begin(), w212.end() - size_t(sparsity * w212.size()), std::ref(f32rng)); |
| 550 | std::shuffle(w212.begin(), w212.end(), rng); |
| 551 | std::generate(w213.begin(), w213.end(), std::ref(f32rng)); |
| 552 | std::fill(w214.begin(), w214.end(), 0.0f); |
| 553 | std::generate(w214.begin(), w214.end() - size_t(sparsity * w214.size()), std::ref(f32rng)); |
| 554 | std::shuffle(w214.begin(), w214.end(), rng); |
| 555 | std::generate(w215.begin(), w215.end(), std::ref(f32rng)); |
| 556 | std::fill(w216.begin(), w216.end(), 0.0f); |
| 557 | std::generate(w216.begin(), w216.end() - size_t(sparsity * w216.size()), std::ref(f32rng)); |
| 558 | std::shuffle(w216.begin(), w216.end(), rng); |
| 559 | std::generate(w217.begin(), w217.end(), std::ref(f32rng)); |
| 560 | std::generate(w218.begin(), w218.end(), std::ref(f32rng)); |
| 561 | std::generate(w219.begin(), w219.end(), std::ref(f32rng)); |
| 562 | std::fill(w220.begin(), w220.end(), 0.0f); |
| 563 | std::generate(w220.begin(), w220.end() - size_t(sparsity * w220.size()), std::ref(f32rng)); |
| 564 | std::shuffle(w220.begin(), w220.end(), rng); |
| 565 | std::generate(w221.begin(), w221.end(), std::ref(f32rng)); |
| 566 | std::fill(w222.begin(), w222.end(), 0.0f); |
| 567 | std::generate(w222.begin(), w222.end() - size_t(sparsity * w222.size()), std::ref(f32rng)); |
| 568 | std::shuffle(w222.begin(), w222.end(), rng); |
| 569 | std::generate(w223.begin(), w223.end(), std::ref(f32rng)); |
| 570 | std::fill(w224.begin(), w224.end(), 0.0f); |
| 571 | std::generate(w224.begin(), w224.end() - size_t(sparsity * w224.size()), std::ref(f32rng)); |
| 572 | std::shuffle(w224.begin(), w224.end(), rng); |
| 573 | std::generate(w225.begin(), w225.end(), std::ref(f32rng)); |
| 574 | std::fill(w226.begin(), w226.end(), 0.0f); |
| 575 | std::generate(w226.begin(), w226.end() - size_t(sparsity * w226.size()), std::ref(f32rng)); |
| 576 | std::shuffle(w226.begin(), w226.end(), rng); |
| 577 | std::generate(w227.begin(), w227.end(), std::ref(f32rng)); |
| 578 | std::generate(w228.begin(), w228.end(), std::ref(f32rng)); |
| 579 | std::generate(w229.begin(), w229.end(), std::ref(f32rng)); |
| 580 | std::fill(w230.begin(), w230.end(), 0.0f); |
| 581 | std::generate(w230.begin(), w230.end() - size_t(sparsity * w230.size()), std::ref(f32rng)); |
| 582 | std::shuffle(w230.begin(), w230.end(), rng); |
| 583 | std::generate(w231.begin(), w231.end(), std::ref(f32rng)); |
| 584 | std::fill(w232.begin(), w232.end(), 0.0f); |
| 585 | std::generate(w232.begin(), w232.end() - size_t(sparsity * w232.size()), std::ref(f32rng)); |
| 586 | std::shuffle(w232.begin(), w232.end(), rng); |
| 587 | std::generate(w233.begin(), w233.end(), std::ref(f32rng)); |
| 588 | std::fill(w234.begin(), w234.end(), 0.0f); |
| 589 | std::generate(w234.begin(), w234.end() - size_t(sparsity * w234.size()), std::ref(f32rng)); |
| 590 | std::shuffle(w234.begin(), w234.end(), rng); |
| 591 | std::generate(w235.begin(), w235.end(), std::ref(f32rng)); |
| 592 | std::fill(w236.begin(), w236.end(), 0.0f); |
| 593 | std::generate(w236.begin(), w236.end() - size_t(sparsity * w236.size()), std::ref(f32rng)); |
| 594 | std::shuffle(w236.begin(), w236.end(), rng); |
| 595 | std::generate(w237.begin(), w237.end(), std::ref(f32rng)); |
| 596 | std::generate(w238.begin(), w238.end(), std::ref(f32rng)); |
| 597 | std::generate(w239.begin(), w239.end(), std::ref(f32rng)); |
| 598 | std::generate(w240.begin(), w240.end(), std::ref(f32rng)); |
| 599 | std::generate(w241.begin(), w241.end(), std::ref(f32rng)); |
| 600 | |
| 601 | ExecutionPlan operators; |
| 602 | xnn_status status; |
| 603 | |
| 604 | xnn_operator_t op0 = nullptr; |
| 605 | status = xnn_create_convolution2d_nchw_f32( |
| 606 | 1 /* top padding */, 1 /* right padding */, |
| 607 | 1 /* bottom padding */, 1 /* left padding */, |
| 608 | 3 /* kernel height */, 3 /* kernel width */, |
| 609 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 610 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 611 | 1 /* groups */, |
| 612 | 3 /* input channels per group */, |
| 613 | 16 /* output_channels_per_group */, |
| 614 | 3 /* input pixel stride */, |
| 615 | 16 /* output pixel stride */, |
| 616 | w114.data(), w115.data(), |
| 617 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 618 | XNN_FLAG_INPUT_NHWC /* flags */, |
| 619 | &op0); |
| 620 | if (status != xnn_status_success) { |
| 621 | std::cerr << "failed to create operation #0" << std::endl; |
| 622 | return ExecutionPlan(); |
| 623 | } |
| 624 | operators.emplace_back(op0, xnn_delete_operator); |
| 625 | |
| 626 | xnn_operator_t op1 = nullptr; |
| 627 | status = xnn_create_hardswish_nc_f32( |
| 628 | 16 /* channels */, |
| 629 | 16 /* input stride */, |
| 630 | 16 /* output stride */, |
| 631 | 0 /* flags */, |
| 632 | &op1); |
| 633 | if (status != xnn_status_success) { |
| 634 | std::cerr << "failed to create operation #1" << std::endl; |
| 635 | return ExecutionPlan(); |
| 636 | } |
| 637 | operators.emplace_back(op1, xnn_delete_operator); |
| 638 | |
| 639 | xnn_operator_t op2 = nullptr; |
| 640 | status = xnn_create_convolution2d_nchw_f32( |
| 641 | 1 /* top padding */, 1 /* right padding */, |
| 642 | 1 /* bottom padding */, 1 /* left padding */, |
| 643 | 3 /* kernel height */, 3 /* kernel width */, |
| 644 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 645 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 646 | 16 /* groups */, |
| 647 | 1 /* input channels per group */, |
| 648 | 1 /* output_channels_per_group */, |
| 649 | 16 /* input pixel stride */, |
| 650 | 16 /* output pixel stride */, |
| 651 | w116.data(), w117.data(), |
| 652 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 653 | 0 /* flags */, |
| 654 | &op2); |
| 655 | if (status != xnn_status_success) { |
| 656 | std::cerr << "failed to create operation #2" << std::endl; |
| 657 | return ExecutionPlan(); |
| 658 | } |
| 659 | operators.emplace_back(op2, xnn_delete_operator); |
| 660 | |
| 661 | xnn_operator_t op3 = nullptr; |
| 662 | status = xnn_create_convolution2d_nchw_f32( |
| 663 | 0 /* top padding */, 0 /* right padding */, |
| 664 | 0 /* bottom padding */, 0 /* left padding */, |
| 665 | 1 /* kernel height */, 1 /* kernel width */, |
| 666 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 667 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 668 | 1 /* groups */, |
| 669 | 16 /* input channels per group */, |
| 670 | 16 /* output_channels_per_group */, |
| 671 | 16 /* input pixel stride */, |
| 672 | 16 /* output pixel stride */, |
| 673 | w118.data(), w119.data(), |
| 674 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 675 | 0 /* flags */, |
| 676 | &op3); |
| 677 | if (status != xnn_status_success) { |
| 678 | std::cerr << "failed to create operation #3" << std::endl; |
| 679 | return ExecutionPlan(); |
| 680 | } |
| 681 | operators.emplace_back(op3, xnn_delete_operator); |
| 682 | |
| 683 | xnn_operator_t op4 = nullptr; |
| 684 | status = xnn_create_add_nd_f32( |
| 685 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 686 | 0 /* flags */, |
| 687 | &op4); |
| 688 | if (status != xnn_status_success) { |
| 689 | std::cerr << "failed to create operation #4" << std::endl; |
| 690 | return ExecutionPlan(); |
| 691 | } |
| 692 | operators.emplace_back(op4, xnn_delete_operator); |
| 693 | |
| 694 | xnn_operator_t op5 = nullptr; |
| 695 | status = xnn_create_convolution2d_nchw_f32( |
| 696 | 0 /* top padding */, 0 /* right padding */, |
| 697 | 0 /* bottom padding */, 0 /* left padding */, |
| 698 | 1 /* kernel height */, 1 /* kernel width */, |
| 699 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 700 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 701 | 1 /* groups */, |
| 702 | 16 /* input channels per group */, |
| 703 | 64 /* output_channels_per_group */, |
| 704 | 16 /* input pixel stride */, |
| 705 | 64 /* output pixel stride */, |
| 706 | w120.data(), w121.data(), |
| 707 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 708 | 0 /* flags */, |
| 709 | &op5); |
| 710 | if (status != xnn_status_success) { |
| 711 | std::cerr << "failed to create operation #5" << std::endl; |
| 712 | return ExecutionPlan(); |
| 713 | } |
| 714 | operators.emplace_back(op5, xnn_delete_operator); |
| 715 | |
| 716 | xnn_operator_t op6 = nullptr; |
| 717 | status = xnn_create_convolution2d_nchw_f32( |
| 718 | 1 /* top padding */, 1 /* right padding */, |
| 719 | 1 /* bottom padding */, 1 /* left padding */, |
| 720 | 3 /* kernel height */, 3 /* kernel width */, |
| 721 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 722 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 723 | 64 /* groups */, |
| 724 | 1 /* input channels per group */, |
| 725 | 1 /* output_channels_per_group */, |
| 726 | 64 /* input pixel stride */, |
| 727 | 64 /* output pixel stride */, |
| 728 | w122.data(), w123.data(), |
| 729 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 730 | 0 /* flags */, |
| 731 | &op6); |
| 732 | if (status != xnn_status_success) { |
| 733 | std::cerr << "failed to create operation #6" << std::endl; |
| 734 | return ExecutionPlan(); |
| 735 | } |
| 736 | operators.emplace_back(op6, xnn_delete_operator); |
| 737 | |
| 738 | xnn_operator_t op7 = nullptr; |
| 739 | status = xnn_create_convolution2d_nchw_f32( |
| 740 | 0 /* top padding */, 0 /* right padding */, |
| 741 | 0 /* bottom padding */, 0 /* left padding */, |
| 742 | 1 /* kernel height */, 1 /* kernel width */, |
| 743 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 744 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 745 | 1 /* groups */, |
| 746 | 64 /* input channels per group */, |
| 747 | 24 /* output_channels_per_group */, |
| 748 | 64 /* input pixel stride */, |
| 749 | 24 /* output pixel stride */, |
| 750 | w124.data(), w125.data(), |
| 751 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 752 | 0 /* flags */, |
| 753 | &op7); |
| 754 | if (status != xnn_status_success) { |
| 755 | std::cerr << "failed to create operation #7" << std::endl; |
| 756 | return ExecutionPlan(); |
| 757 | } |
| 758 | operators.emplace_back(op7, xnn_delete_operator); |
| 759 | |
| 760 | xnn_operator_t op8 = nullptr; |
| 761 | status = xnn_create_convolution2d_nchw_f32( |
| 762 | 0 /* top padding */, 0 /* right padding */, |
| 763 | 0 /* bottom padding */, 0 /* left padding */, |
| 764 | 1 /* kernel height */, 1 /* kernel width */, |
| 765 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 766 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 767 | 1 /* groups */, |
| 768 | 24 /* input channels per group */, |
| 769 | 72 /* output_channels_per_group */, |
| 770 | 24 /* input pixel stride */, |
| 771 | 72 /* output pixel stride */, |
| 772 | w126.data(), w127.data(), |
| 773 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 774 | 0 /* flags */, |
| 775 | &op8); |
| 776 | if (status != xnn_status_success) { |
| 777 | std::cerr << "failed to create operation #8" << std::endl; |
| 778 | return ExecutionPlan(); |
| 779 | } |
| 780 | operators.emplace_back(op8, xnn_delete_operator); |
| 781 | |
| 782 | xnn_operator_t op9 = nullptr; |
| 783 | status = xnn_create_convolution2d_nchw_f32( |
| 784 | 1 /* top padding */, 1 /* right padding */, |
| 785 | 1 /* bottom padding */, 1 /* left padding */, |
| 786 | 3 /* kernel height */, 3 /* kernel width */, |
| 787 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 788 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 789 | 72 /* groups */, |
| 790 | 1 /* input channels per group */, |
| 791 | 1 /* output_channels_per_group */, |
| 792 | 72 /* input pixel stride */, |
| 793 | 72 /* output pixel stride */, |
| 794 | w128.data(), w129.data(), |
| 795 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 796 | 0 /* flags */, |
| 797 | &op9); |
| 798 | if (status != xnn_status_success) { |
| 799 | std::cerr << "failed to create operation #9" << std::endl; |
| 800 | return ExecutionPlan(); |
| 801 | } |
| 802 | operators.emplace_back(op9, xnn_delete_operator); |
| 803 | |
| 804 | xnn_operator_t op10 = nullptr; |
| 805 | status = xnn_create_convolution2d_nchw_f32( |
| 806 | 0 /* top padding */, 0 /* right padding */, |
| 807 | 0 /* bottom padding */, 0 /* left padding */, |
| 808 | 1 /* kernel height */, 1 /* kernel width */, |
| 809 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 810 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 811 | 1 /* groups */, |
| 812 | 72 /* input channels per group */, |
| 813 | 24 /* output_channels_per_group */, |
| 814 | 72 /* input pixel stride */, |
| 815 | 24 /* output pixel stride */, |
| 816 | w130.data(), w131.data(), |
| 817 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 818 | 0 /* flags */, |
| 819 | &op10); |
| 820 | if (status != xnn_status_success) { |
| 821 | std::cerr << "failed to create operation #10" << std::endl; |
| 822 | return ExecutionPlan(); |
| 823 | } |
| 824 | operators.emplace_back(op10, xnn_delete_operator); |
| 825 | |
| 826 | xnn_operator_t op11 = nullptr; |
| 827 | status = xnn_create_add_nd_f32( |
| 828 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 829 | 0 /* flags */, |
| 830 | &op11); |
| 831 | if (status != xnn_status_success) { |
| 832 | std::cerr << "failed to create operation #11" << std::endl; |
| 833 | return ExecutionPlan(); |
| 834 | } |
| 835 | operators.emplace_back(op11, xnn_delete_operator); |
| 836 | |
| 837 | xnn_operator_t op12 = nullptr; |
| 838 | status = xnn_create_convolution2d_nchw_f32( |
| 839 | 0 /* top padding */, 0 /* right padding */, |
| 840 | 0 /* bottom padding */, 0 /* left padding */, |
| 841 | 1 /* kernel height */, 1 /* kernel width */, |
| 842 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 843 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 844 | 1 /* groups */, |
| 845 | 24 /* input channels per group */, |
| 846 | 72 /* output_channels_per_group */, |
| 847 | 24 /* input pixel stride */, |
| 848 | 72 /* output pixel stride */, |
| 849 | w132.data(), w133.data(), |
| 850 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 851 | 0 /* flags */, |
| 852 | &op12); |
| 853 | if (status != xnn_status_success) { |
| 854 | std::cerr << "failed to create operation #12" << std::endl; |
| 855 | return ExecutionPlan(); |
| 856 | } |
| 857 | operators.emplace_back(op12, xnn_delete_operator); |
| 858 | |
| 859 | xnn_operator_t op13 = nullptr; |
| 860 | status = xnn_create_convolution2d_nchw_f32( |
| 861 | 2 /* top padding */, 2 /* right padding */, |
| 862 | 2 /* bottom padding */, 2 /* left padding */, |
| 863 | 5 /* kernel height */, 5 /* kernel width */, |
| 864 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 865 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 866 | 72 /* groups */, |
| 867 | 1 /* input channels per group */, |
| 868 | 1 /* output_channels_per_group */, |
| 869 | 72 /* input pixel stride */, |
| 870 | 72 /* output pixel stride */, |
| 871 | w134.data(), w135.data(), |
| 872 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 873 | 0 /* flags */, |
| 874 | &op13); |
| 875 | if (status != xnn_status_success) { |
| 876 | std::cerr << "failed to create operation #13" << std::endl; |
| 877 | return ExecutionPlan(); |
| 878 | } |
| 879 | operators.emplace_back(op13, xnn_delete_operator); |
| 880 | |
| 881 | xnn_operator_t op14 = nullptr; |
| 882 | status = xnn_create_global_average_pooling_ncw_f32( |
| 883 | 72 /* channels */, |
| 884 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 885 | 0 /* flags */, |
| 886 | &op14); |
| 887 | if (status != xnn_status_success) { |
| 888 | std::cerr << "failed to create operation #14" << std::endl; |
| 889 | return ExecutionPlan(); |
| 890 | } |
| 891 | operators.emplace_back(op14, xnn_delete_operator); |
| 892 | |
| 893 | xnn_operator_t op15 = nullptr; |
| 894 | status = xnn_create_convolution2d_nchw_f32( |
| 895 | 0 /* top padding */, 0 /* right padding */, |
| 896 | 0 /* bottom padding */, 0 /* left padding */, |
| 897 | 1 /* kernel height */, 1 /* kernel width */, |
| 898 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 899 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 900 | 1 /* groups */, |
| 901 | 72 /* input channels per group */, |
| 902 | 24 /* output_channels_per_group */, |
| 903 | 72 /* input pixel stride */, |
| 904 | 24 /* output pixel stride */, |
| 905 | w136.data(), w137.data(), |
| 906 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 907 | 0 /* flags */, |
| 908 | &op15); |
| 909 | if (status != xnn_status_success) { |
| 910 | std::cerr << "failed to create operation #15" << std::endl; |
| 911 | return ExecutionPlan(); |
| 912 | } |
| 913 | operators.emplace_back(op15, xnn_delete_operator); |
| 914 | |
| 915 | xnn_operator_t op16 = nullptr; |
| 916 | status = xnn_create_convolution2d_nchw_f32( |
| 917 | 0 /* top padding */, 0 /* right padding */, |
| 918 | 0 /* bottom padding */, 0 /* left padding */, |
| 919 | 1 /* kernel height */, 1 /* kernel width */, |
| 920 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 921 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 922 | 1 /* groups */, |
| 923 | 24 /* input channels per group */, |
| 924 | 72 /* output_channels_per_group */, |
| 925 | 24 /* input pixel stride */, |
| 926 | 72 /* output pixel stride */, |
| 927 | w138.data(), w139.data(), |
| 928 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 929 | 0 /* flags */, |
| 930 | &op16); |
| 931 | if (status != xnn_status_success) { |
| 932 | std::cerr << "failed to create operation #16" << std::endl; |
| 933 | return ExecutionPlan(); |
| 934 | } |
| 935 | operators.emplace_back(op16, xnn_delete_operator); |
| 936 | |
| 937 | xnn_operator_t op17 = nullptr; |
| 938 | status = xnn_create_multiply_nd_f32( |
| 939 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 940 | 0 /* flags */, |
| 941 | &op17); |
| 942 | if (status != xnn_status_success) { |
| 943 | std::cerr << "failed to create operation #17" << std::endl; |
| 944 | return ExecutionPlan(); |
| 945 | } |
| 946 | operators.emplace_back(op17, xnn_delete_operator); |
| 947 | |
| 948 | xnn_operator_t op18 = nullptr; |
| 949 | status = xnn_create_convolution2d_nchw_f32( |
| 950 | 0 /* top padding */, 0 /* right padding */, |
| 951 | 0 /* bottom padding */, 0 /* left padding */, |
| 952 | 1 /* kernel height */, 1 /* kernel width */, |
| 953 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 954 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 955 | 1 /* groups */, |
| 956 | 72 /* input channels per group */, |
| 957 | 40 /* output_channels_per_group */, |
| 958 | 72 /* input pixel stride */, |
| 959 | 40 /* output pixel stride */, |
| 960 | w140.data(), w141.data(), |
| 961 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 962 | 0 /* flags */, |
| 963 | &op18); |
| 964 | if (status != xnn_status_success) { |
| 965 | std::cerr << "failed to create operation #18" << std::endl; |
| 966 | return ExecutionPlan(); |
| 967 | } |
| 968 | operators.emplace_back(op18, xnn_delete_operator); |
| 969 | |
| 970 | xnn_operator_t op19 = nullptr; |
| 971 | status = xnn_create_convolution2d_nchw_f32( |
| 972 | 0 /* top padding */, 0 /* right padding */, |
| 973 | 0 /* bottom padding */, 0 /* left padding */, |
| 974 | 1 /* kernel height */, 1 /* kernel width */, |
| 975 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 976 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 977 | 1 /* groups */, |
| 978 | 40 /* input channels per group */, |
| 979 | 120 /* output_channels_per_group */, |
| 980 | 40 /* input pixel stride */, |
| 981 | 120 /* output pixel stride */, |
| 982 | w142.data(), w143.data(), |
| 983 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 984 | 0 /* flags */, |
| 985 | &op19); |
| 986 | if (status != xnn_status_success) { |
| 987 | std::cerr << "failed to create operation #19" << std::endl; |
| 988 | return ExecutionPlan(); |
| 989 | } |
| 990 | operators.emplace_back(op19, xnn_delete_operator); |
| 991 | |
| 992 | xnn_operator_t op20 = nullptr; |
| 993 | status = xnn_create_convolution2d_nchw_f32( |
| 994 | 2 /* top padding */, 2 /* right padding */, |
| 995 | 2 /* bottom padding */, 2 /* left padding */, |
| 996 | 5 /* kernel height */, 5 /* kernel width */, |
| 997 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 998 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 999 | 120 /* groups */, |
| 1000 | 1 /* input channels per group */, |
| 1001 | 1 /* output_channels_per_group */, |
| 1002 | 120 /* input pixel stride */, |
| 1003 | 120 /* output pixel stride */, |
| 1004 | w144.data(), w145.data(), |
| 1005 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1006 | 0 /* flags */, |
| 1007 | &op20); |
| 1008 | if (status != xnn_status_success) { |
| 1009 | std::cerr << "failed to create operation #20" << std::endl; |
| 1010 | return ExecutionPlan(); |
| 1011 | } |
| 1012 | operators.emplace_back(op20, xnn_delete_operator); |
| 1013 | |
| 1014 | xnn_operator_t op21 = nullptr; |
| 1015 | status = xnn_create_global_average_pooling_ncw_f32( |
| 1016 | 120 /* channels */, |
| 1017 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1018 | 0 /* flags */, |
| 1019 | &op21); |
| 1020 | if (status != xnn_status_success) { |
| 1021 | std::cerr << "failed to create operation #21" << std::endl; |
| 1022 | return ExecutionPlan(); |
| 1023 | } |
| 1024 | operators.emplace_back(op21, xnn_delete_operator); |
| 1025 | |
| 1026 | xnn_operator_t op22 = nullptr; |
| 1027 | status = xnn_create_convolution2d_nchw_f32( |
| 1028 | 0 /* top padding */, 0 /* right padding */, |
| 1029 | 0 /* bottom padding */, 0 /* left padding */, |
| 1030 | 1 /* kernel height */, 1 /* kernel width */, |
| 1031 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1032 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1033 | 1 /* groups */, |
| 1034 | 120 /* input channels per group */, |
| 1035 | 32 /* output_channels_per_group */, |
| 1036 | 120 /* input pixel stride */, |
| 1037 | 32 /* output pixel stride */, |
| 1038 | w146.data(), w147.data(), |
| 1039 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1040 | 0 /* flags */, |
| 1041 | &op22); |
| 1042 | if (status != xnn_status_success) { |
| 1043 | std::cerr << "failed to create operation #22" << std::endl; |
| 1044 | return ExecutionPlan(); |
| 1045 | } |
| 1046 | operators.emplace_back(op22, xnn_delete_operator); |
| 1047 | |
| 1048 | xnn_operator_t op23 = nullptr; |
| 1049 | status = xnn_create_convolution2d_nchw_f32( |
| 1050 | 0 /* top padding */, 0 /* right padding */, |
| 1051 | 0 /* bottom padding */, 0 /* left padding */, |
| 1052 | 1 /* kernel height */, 1 /* kernel width */, |
| 1053 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1054 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1055 | 1 /* groups */, |
| 1056 | 32 /* input channels per group */, |
| 1057 | 120 /* output_channels_per_group */, |
| 1058 | 32 /* input pixel stride */, |
| 1059 | 120 /* output pixel stride */, |
| 1060 | w148.data(), w149.data(), |
| 1061 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1062 | 0 /* flags */, |
| 1063 | &op23); |
| 1064 | if (status != xnn_status_success) { |
| 1065 | std::cerr << "failed to create operation #23" << std::endl; |
| 1066 | return ExecutionPlan(); |
| 1067 | } |
| 1068 | operators.emplace_back(op23, xnn_delete_operator); |
| 1069 | |
| 1070 | xnn_operator_t op24 = nullptr; |
| 1071 | status = xnn_create_multiply_nd_f32( |
| 1072 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1073 | 0 /* flags */, |
| 1074 | &op24); |
| 1075 | if (status != xnn_status_success) { |
| 1076 | std::cerr << "failed to create operation #24" << std::endl; |
| 1077 | return ExecutionPlan(); |
| 1078 | } |
| 1079 | operators.emplace_back(op24, xnn_delete_operator); |
| 1080 | |
| 1081 | xnn_operator_t op25 = nullptr; |
| 1082 | status = xnn_create_convolution2d_nchw_f32( |
| 1083 | 0 /* top padding */, 0 /* right padding */, |
| 1084 | 0 /* bottom padding */, 0 /* left padding */, |
| 1085 | 1 /* kernel height */, 1 /* kernel width */, |
| 1086 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1087 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1088 | 1 /* groups */, |
| 1089 | 120 /* input channels per group */, |
| 1090 | 40 /* output_channels_per_group */, |
| 1091 | 120 /* input pixel stride */, |
| 1092 | 40 /* output pixel stride */, |
| 1093 | w150.data(), w151.data(), |
| 1094 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1095 | 0 /* flags */, |
| 1096 | &op25); |
| 1097 | if (status != xnn_status_success) { |
| 1098 | std::cerr << "failed to create operation #25" << std::endl; |
| 1099 | return ExecutionPlan(); |
| 1100 | } |
| 1101 | operators.emplace_back(op25, xnn_delete_operator); |
| 1102 | |
| 1103 | xnn_operator_t op26 = nullptr; |
| 1104 | status = xnn_create_add_nd_f32( |
| 1105 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1106 | 0 /* flags */, |
| 1107 | &op26); |
| 1108 | if (status != xnn_status_success) { |
| 1109 | std::cerr << "failed to create operation #26" << std::endl; |
| 1110 | return ExecutionPlan(); |
| 1111 | } |
| 1112 | operators.emplace_back(op26, xnn_delete_operator); |
| 1113 | |
| 1114 | xnn_operator_t op27 = nullptr; |
| 1115 | status = xnn_create_convolution2d_nchw_f32( |
| 1116 | 0 /* top padding */, 0 /* right padding */, |
| 1117 | 0 /* bottom padding */, 0 /* left padding */, |
| 1118 | 1 /* kernel height */, 1 /* kernel width */, |
| 1119 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1120 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1121 | 1 /* groups */, |
| 1122 | 40 /* input channels per group */, |
| 1123 | 120 /* output_channels_per_group */, |
| 1124 | 40 /* input pixel stride */, |
| 1125 | 120 /* output pixel stride */, |
| 1126 | w152.data(), w153.data(), |
| 1127 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1128 | 0 /* flags */, |
| 1129 | &op27); |
| 1130 | if (status != xnn_status_success) { |
| 1131 | std::cerr << "failed to create operation #27" << std::endl; |
| 1132 | return ExecutionPlan(); |
| 1133 | } |
| 1134 | operators.emplace_back(op27, xnn_delete_operator); |
| 1135 | |
| 1136 | xnn_operator_t op28 = nullptr; |
| 1137 | status = xnn_create_convolution2d_nchw_f32( |
| 1138 | 2 /* top padding */, 2 /* right padding */, |
| 1139 | 2 /* bottom padding */, 2 /* left padding */, |
| 1140 | 5 /* kernel height */, 5 /* kernel width */, |
| 1141 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1142 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1143 | 120 /* groups */, |
| 1144 | 1 /* input channels per group */, |
| 1145 | 1 /* output_channels_per_group */, |
| 1146 | 120 /* input pixel stride */, |
| 1147 | 120 /* output pixel stride */, |
| 1148 | w154.data(), w155.data(), |
| 1149 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1150 | 0 /* flags */, |
| 1151 | &op28); |
| 1152 | if (status != xnn_status_success) { |
| 1153 | std::cerr << "failed to create operation #28" << std::endl; |
| 1154 | return ExecutionPlan(); |
| 1155 | } |
| 1156 | operators.emplace_back(op28, xnn_delete_operator); |
| 1157 | |
| 1158 | xnn_operator_t op29 = nullptr; |
| 1159 | status = xnn_create_global_average_pooling_ncw_f32( |
| 1160 | 120 /* channels */, |
| 1161 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1162 | 0 /* flags */, |
| 1163 | &op29); |
| 1164 | if (status != xnn_status_success) { |
| 1165 | std::cerr << "failed to create operation #29" << std::endl; |
| 1166 | return ExecutionPlan(); |
| 1167 | } |
| 1168 | operators.emplace_back(op29, xnn_delete_operator); |
| 1169 | |
| 1170 | xnn_operator_t op30 = nullptr; |
| 1171 | status = xnn_create_convolution2d_nchw_f32( |
| 1172 | 0 /* top padding */, 0 /* right padding */, |
| 1173 | 0 /* bottom padding */, 0 /* left padding */, |
| 1174 | 1 /* kernel height */, 1 /* kernel width */, |
| 1175 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1176 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1177 | 1 /* groups */, |
| 1178 | 120 /* input channels per group */, |
| 1179 | 32 /* output_channels_per_group */, |
| 1180 | 120 /* input pixel stride */, |
| 1181 | 32 /* output pixel stride */, |
| 1182 | w156.data(), w157.data(), |
| 1183 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1184 | 0 /* flags */, |
| 1185 | &op30); |
| 1186 | if (status != xnn_status_success) { |
| 1187 | std::cerr << "failed to create operation #30" << std::endl; |
| 1188 | return ExecutionPlan(); |
| 1189 | } |
| 1190 | operators.emplace_back(op30, xnn_delete_operator); |
| 1191 | |
| 1192 | xnn_operator_t op31 = nullptr; |
| 1193 | status = xnn_create_convolution2d_nchw_f32( |
| 1194 | 0 /* top padding */, 0 /* right padding */, |
| 1195 | 0 /* bottom padding */, 0 /* left padding */, |
| 1196 | 1 /* kernel height */, 1 /* kernel width */, |
| 1197 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1198 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1199 | 1 /* groups */, |
| 1200 | 32 /* input channels per group */, |
| 1201 | 120 /* output_channels_per_group */, |
| 1202 | 32 /* input pixel stride */, |
| 1203 | 120 /* output pixel stride */, |
| 1204 | w158.data(), w159.data(), |
| 1205 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1206 | 0 /* flags */, |
| 1207 | &op31); |
| 1208 | if (status != xnn_status_success) { |
| 1209 | std::cerr << "failed to create operation #31" << std::endl; |
| 1210 | return ExecutionPlan(); |
| 1211 | } |
| 1212 | operators.emplace_back(op31, xnn_delete_operator); |
| 1213 | |
| 1214 | xnn_operator_t op32 = nullptr; |
| 1215 | status = xnn_create_multiply_nd_f32( |
| 1216 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1217 | 0 /* flags */, |
| 1218 | &op32); |
| 1219 | if (status != xnn_status_success) { |
| 1220 | std::cerr << "failed to create operation #32" << std::endl; |
| 1221 | return ExecutionPlan(); |
| 1222 | } |
| 1223 | operators.emplace_back(op32, xnn_delete_operator); |
| 1224 | |
| 1225 | xnn_operator_t op33 = nullptr; |
| 1226 | status = xnn_create_convolution2d_nchw_f32( |
| 1227 | 0 /* top padding */, 0 /* right padding */, |
| 1228 | 0 /* bottom padding */, 0 /* left padding */, |
| 1229 | 1 /* kernel height */, 1 /* kernel width */, |
| 1230 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1231 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1232 | 1 /* groups */, |
| 1233 | 120 /* input channels per group */, |
| 1234 | 40 /* output_channels_per_group */, |
| 1235 | 120 /* input pixel stride */, |
| 1236 | 40 /* output pixel stride */, |
| 1237 | w160.data(), w161.data(), |
| 1238 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1239 | 0 /* flags */, |
| 1240 | &op33); |
| 1241 | if (status != xnn_status_success) { |
| 1242 | std::cerr << "failed to create operation #33" << std::endl; |
| 1243 | return ExecutionPlan(); |
| 1244 | } |
| 1245 | operators.emplace_back(op33, xnn_delete_operator); |
| 1246 | |
| 1247 | xnn_operator_t op34 = nullptr; |
| 1248 | status = xnn_create_add_nd_f32( |
| 1249 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1250 | 0 /* flags */, |
| 1251 | &op34); |
| 1252 | if (status != xnn_status_success) { |
| 1253 | std::cerr << "failed to create operation #34" << std::endl; |
| 1254 | return ExecutionPlan(); |
| 1255 | } |
| 1256 | operators.emplace_back(op34, xnn_delete_operator); |
| 1257 | |
| 1258 | xnn_operator_t op35 = nullptr; |
| 1259 | status = xnn_create_convolution2d_nchw_f32( |
| 1260 | 0 /* top padding */, 0 /* right padding */, |
| 1261 | 0 /* bottom padding */, 0 /* left padding */, |
| 1262 | 1 /* kernel height */, 1 /* kernel width */, |
| 1263 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1264 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1265 | 1 /* groups */, |
| 1266 | 40 /* input channels per group */, |
| 1267 | 240 /* output_channels_per_group */, |
| 1268 | 40 /* input pixel stride */, |
| 1269 | 240 /* output pixel stride */, |
| 1270 | w162.data(), w163.data(), |
| 1271 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1272 | 0 /* flags */, |
| 1273 | &op35); |
| 1274 | if (status != xnn_status_success) { |
| 1275 | std::cerr << "failed to create operation #35" << std::endl; |
| 1276 | return ExecutionPlan(); |
| 1277 | } |
| 1278 | operators.emplace_back(op35, xnn_delete_operator); |
| 1279 | |
| 1280 | xnn_operator_t op36 = nullptr; |
| 1281 | status = xnn_create_hardswish_nc_f32( |
| 1282 | 240 /* channels */, |
| 1283 | 240 /* input stride */, |
| 1284 | 240 /* output stride */, |
| 1285 | 0 /* flags */, |
| 1286 | &op36); |
| 1287 | if (status != xnn_status_success) { |
| 1288 | std::cerr << "failed to create operation #36" << std::endl; |
| 1289 | return ExecutionPlan(); |
| 1290 | } |
| 1291 | operators.emplace_back(op36, xnn_delete_operator); |
| 1292 | |
| 1293 | xnn_operator_t op37 = nullptr; |
| 1294 | status = xnn_create_convolution2d_nchw_f32( |
| 1295 | 1 /* top padding */, 1 /* right padding */, |
| 1296 | 1 /* bottom padding */, 1 /* left padding */, |
| 1297 | 3 /* kernel height */, 3 /* kernel width */, |
| 1298 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 1299 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1300 | 240 /* groups */, |
| 1301 | 1 /* input channels per group */, |
| 1302 | 1 /* output_channels_per_group */, |
| 1303 | 240 /* input pixel stride */, |
| 1304 | 240 /* output pixel stride */, |
| 1305 | w164.data(), w165.data(), |
| 1306 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1307 | 0 /* flags */, |
| 1308 | &op37); |
| 1309 | if (status != xnn_status_success) { |
| 1310 | std::cerr << "failed to create operation #37" << std::endl; |
| 1311 | return ExecutionPlan(); |
| 1312 | } |
| 1313 | operators.emplace_back(op37, xnn_delete_operator); |
| 1314 | |
| 1315 | xnn_operator_t op38 = nullptr; |
| 1316 | status = xnn_create_hardswish_nc_f32( |
| 1317 | 240 /* channels */, |
| 1318 | 240 /* input stride */, |
| 1319 | 240 /* output stride */, |
| 1320 | 0 /* flags */, |
| 1321 | &op38); |
| 1322 | if (status != xnn_status_success) { |
| 1323 | std::cerr << "failed to create operation #38" << std::endl; |
| 1324 | return ExecutionPlan(); |
| 1325 | } |
| 1326 | operators.emplace_back(op38, xnn_delete_operator); |
| 1327 | |
| 1328 | xnn_operator_t op39 = nullptr; |
| 1329 | status = xnn_create_convolution2d_nchw_f32( |
| 1330 | 0 /* top padding */, 0 /* right padding */, |
| 1331 | 0 /* bottom padding */, 0 /* left padding */, |
| 1332 | 1 /* kernel height */, 1 /* kernel width */, |
| 1333 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1334 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1335 | 1 /* groups */, |
| 1336 | 240 /* input channels per group */, |
| 1337 | 80 /* output_channels_per_group */, |
| 1338 | 240 /* input pixel stride */, |
| 1339 | 80 /* output pixel stride */, |
| 1340 | w166.data(), w167.data(), |
| 1341 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1342 | 0 /* flags */, |
| 1343 | &op39); |
| 1344 | if (status != xnn_status_success) { |
| 1345 | std::cerr << "failed to create operation #39" << std::endl; |
| 1346 | return ExecutionPlan(); |
| 1347 | } |
| 1348 | operators.emplace_back(op39, xnn_delete_operator); |
| 1349 | |
| 1350 | xnn_operator_t op40 = nullptr; |
| 1351 | status = xnn_create_convolution2d_nchw_f32( |
| 1352 | 0 /* top padding */, 0 /* right padding */, |
| 1353 | 0 /* bottom padding */, 0 /* left padding */, |
| 1354 | 1 /* kernel height */, 1 /* kernel width */, |
| 1355 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1356 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1357 | 1 /* groups */, |
| 1358 | 80 /* input channels per group */, |
| 1359 | 200 /* output_channels_per_group */, |
| 1360 | 80 /* input pixel stride */, |
| 1361 | 200 /* output pixel stride */, |
| 1362 | w168.data(), w169.data(), |
| 1363 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1364 | 0 /* flags */, |
| 1365 | &op40); |
| 1366 | if (status != xnn_status_success) { |
| 1367 | std::cerr << "failed to create operation #40" << std::endl; |
| 1368 | return ExecutionPlan(); |
| 1369 | } |
| 1370 | operators.emplace_back(op40, xnn_delete_operator); |
| 1371 | |
| 1372 | xnn_operator_t op41 = nullptr; |
| 1373 | status = xnn_create_hardswish_nc_f32( |
| 1374 | 200 /* channels */, |
| 1375 | 200 /* input stride */, |
| 1376 | 200 /* output stride */, |
| 1377 | 0 /* flags */, |
| 1378 | &op41); |
| 1379 | if (status != xnn_status_success) { |
| 1380 | std::cerr << "failed to create operation #41" << std::endl; |
| 1381 | return ExecutionPlan(); |
| 1382 | } |
| 1383 | operators.emplace_back(op41, xnn_delete_operator); |
| 1384 | |
| 1385 | xnn_operator_t op42 = nullptr; |
| 1386 | status = xnn_create_convolution2d_nchw_f32( |
| 1387 | 1 /* top padding */, 1 /* right padding */, |
| 1388 | 1 /* bottom padding */, 1 /* left padding */, |
| 1389 | 3 /* kernel height */, 3 /* kernel width */, |
| 1390 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1391 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1392 | 200 /* groups */, |
| 1393 | 1 /* input channels per group */, |
| 1394 | 1 /* output_channels_per_group */, |
| 1395 | 200 /* input pixel stride */, |
| 1396 | 200 /* output pixel stride */, |
| 1397 | w170.data(), w171.data(), |
| 1398 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1399 | 0 /* flags */, |
| 1400 | &op42); |
| 1401 | if (status != xnn_status_success) { |
| 1402 | std::cerr << "failed to create operation #42" << std::endl; |
| 1403 | return ExecutionPlan(); |
| 1404 | } |
| 1405 | operators.emplace_back(op42, xnn_delete_operator); |
| 1406 | |
| 1407 | xnn_operator_t op43 = nullptr; |
| 1408 | status = xnn_create_hardswish_nc_f32( |
| 1409 | 200 /* channels */, |
| 1410 | 200 /* input stride */, |
| 1411 | 200 /* output stride */, |
| 1412 | 0 /* flags */, |
| 1413 | &op43); |
| 1414 | if (status != xnn_status_success) { |
| 1415 | std::cerr << "failed to create operation #43" << std::endl; |
| 1416 | return ExecutionPlan(); |
| 1417 | } |
| 1418 | operators.emplace_back(op43, xnn_delete_operator); |
| 1419 | |
| 1420 | xnn_operator_t op44 = nullptr; |
| 1421 | status = xnn_create_convolution2d_nchw_f32( |
| 1422 | 0 /* top padding */, 0 /* right padding */, |
| 1423 | 0 /* bottom padding */, 0 /* left padding */, |
| 1424 | 1 /* kernel height */, 1 /* kernel width */, |
| 1425 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1426 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1427 | 1 /* groups */, |
| 1428 | 200 /* input channels per group */, |
| 1429 | 80 /* output_channels_per_group */, |
| 1430 | 200 /* input pixel stride */, |
| 1431 | 80 /* output pixel stride */, |
| 1432 | w172.data(), w173.data(), |
| 1433 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1434 | 0 /* flags */, |
| 1435 | &op44); |
| 1436 | if (status != xnn_status_success) { |
| 1437 | std::cerr << "failed to create operation #44" << std::endl; |
| 1438 | return ExecutionPlan(); |
| 1439 | } |
| 1440 | operators.emplace_back(op44, xnn_delete_operator); |
| 1441 | |
| 1442 | xnn_operator_t op45 = nullptr; |
| 1443 | status = xnn_create_add_nd_f32( |
| 1444 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1445 | 0 /* flags */, |
| 1446 | &op45); |
| 1447 | if (status != xnn_status_success) { |
| 1448 | std::cerr << "failed to create operation #45" << std::endl; |
| 1449 | return ExecutionPlan(); |
| 1450 | } |
| 1451 | operators.emplace_back(op45, xnn_delete_operator); |
| 1452 | |
| 1453 | xnn_operator_t op46 = nullptr; |
| 1454 | status = xnn_create_convolution2d_nchw_f32( |
| 1455 | 0 /* top padding */, 0 /* right padding */, |
| 1456 | 0 /* bottom padding */, 0 /* left padding */, |
| 1457 | 1 /* kernel height */, 1 /* kernel width */, |
| 1458 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1459 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1460 | 1 /* groups */, |
| 1461 | 80 /* input channels per group */, |
| 1462 | 184 /* output_channels_per_group */, |
| 1463 | 80 /* input pixel stride */, |
| 1464 | 184 /* output pixel stride */, |
| 1465 | w174.data(), w175.data(), |
| 1466 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1467 | 0 /* flags */, |
| 1468 | &op46); |
| 1469 | if (status != xnn_status_success) { |
| 1470 | std::cerr << "failed to create operation #46" << std::endl; |
| 1471 | return ExecutionPlan(); |
| 1472 | } |
| 1473 | operators.emplace_back(op46, xnn_delete_operator); |
| 1474 | |
| 1475 | xnn_operator_t op47 = nullptr; |
| 1476 | status = xnn_create_hardswish_nc_f32( |
| 1477 | 184 /* channels */, |
| 1478 | 184 /* input stride */, |
| 1479 | 184 /* output stride */, |
| 1480 | 0 /* flags */, |
| 1481 | &op47); |
| 1482 | if (status != xnn_status_success) { |
| 1483 | std::cerr << "failed to create operation #47" << std::endl; |
| 1484 | return ExecutionPlan(); |
| 1485 | } |
| 1486 | operators.emplace_back(op47, xnn_delete_operator); |
| 1487 | |
| 1488 | xnn_operator_t op48 = nullptr; |
| 1489 | status = xnn_create_convolution2d_nchw_f32( |
| 1490 | 1 /* top padding */, 1 /* right padding */, |
| 1491 | 1 /* bottom padding */, 1 /* left padding */, |
| 1492 | 3 /* kernel height */, 3 /* kernel width */, |
| 1493 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1494 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1495 | 184 /* groups */, |
| 1496 | 1 /* input channels per group */, |
| 1497 | 1 /* output_channels_per_group */, |
| 1498 | 184 /* input pixel stride */, |
| 1499 | 184 /* output pixel stride */, |
| 1500 | w176.data(), w177.data(), |
| 1501 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1502 | 0 /* flags */, |
| 1503 | &op48); |
| 1504 | if (status != xnn_status_success) { |
| 1505 | std::cerr << "failed to create operation #48" << std::endl; |
| 1506 | return ExecutionPlan(); |
| 1507 | } |
| 1508 | operators.emplace_back(op48, xnn_delete_operator); |
| 1509 | |
| 1510 | xnn_operator_t op49 = nullptr; |
| 1511 | status = xnn_create_hardswish_nc_f32( |
| 1512 | 184 /* channels */, |
| 1513 | 184 /* input stride */, |
| 1514 | 184 /* output stride */, |
| 1515 | 0 /* flags */, |
| 1516 | &op49); |
| 1517 | if (status != xnn_status_success) { |
| 1518 | std::cerr << "failed to create operation #49" << std::endl; |
| 1519 | return ExecutionPlan(); |
| 1520 | } |
| 1521 | operators.emplace_back(op49, xnn_delete_operator); |
| 1522 | |
| 1523 | xnn_operator_t op50 = nullptr; |
| 1524 | status = xnn_create_convolution2d_nchw_f32( |
| 1525 | 0 /* top padding */, 0 /* right padding */, |
| 1526 | 0 /* bottom padding */, 0 /* left padding */, |
| 1527 | 1 /* kernel height */, 1 /* kernel width */, |
| 1528 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1529 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1530 | 1 /* groups */, |
| 1531 | 184 /* input channels per group */, |
| 1532 | 80 /* output_channels_per_group */, |
| 1533 | 184 /* input pixel stride */, |
| 1534 | 80 /* output pixel stride */, |
| 1535 | w178.data(), w179.data(), |
| 1536 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1537 | 0 /* flags */, |
| 1538 | &op50); |
| 1539 | if (status != xnn_status_success) { |
| 1540 | std::cerr << "failed to create operation #50" << std::endl; |
| 1541 | return ExecutionPlan(); |
| 1542 | } |
| 1543 | operators.emplace_back(op50, xnn_delete_operator); |
| 1544 | |
| 1545 | xnn_operator_t op51 = nullptr; |
| 1546 | status = xnn_create_add_nd_f32( |
| 1547 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1548 | 0 /* flags */, |
| 1549 | &op51); |
| 1550 | if (status != xnn_status_success) { |
| 1551 | std::cerr << "failed to create operation #51" << std::endl; |
| 1552 | return ExecutionPlan(); |
| 1553 | } |
| 1554 | operators.emplace_back(op51, xnn_delete_operator); |
| 1555 | |
| 1556 | xnn_operator_t op52 = nullptr; |
| 1557 | status = xnn_create_convolution2d_nchw_f32( |
| 1558 | 0 /* top padding */, 0 /* right padding */, |
| 1559 | 0 /* bottom padding */, 0 /* left padding */, |
| 1560 | 1 /* kernel height */, 1 /* kernel width */, |
| 1561 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1562 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1563 | 1 /* groups */, |
| 1564 | 80 /* input channels per group */, |
| 1565 | 184 /* output_channels_per_group */, |
| 1566 | 80 /* input pixel stride */, |
| 1567 | 184 /* output pixel stride */, |
| 1568 | w180.data(), w181.data(), |
| 1569 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1570 | 0 /* flags */, |
| 1571 | &op52); |
| 1572 | if (status != xnn_status_success) { |
| 1573 | std::cerr << "failed to create operation #52" << std::endl; |
| 1574 | return ExecutionPlan(); |
| 1575 | } |
| 1576 | operators.emplace_back(op52, xnn_delete_operator); |
| 1577 | |
| 1578 | xnn_operator_t op53 = nullptr; |
| 1579 | status = xnn_create_hardswish_nc_f32( |
| 1580 | 184 /* channels */, |
| 1581 | 184 /* input stride */, |
| 1582 | 184 /* output stride */, |
| 1583 | 0 /* flags */, |
| 1584 | &op53); |
| 1585 | if (status != xnn_status_success) { |
| 1586 | std::cerr << "failed to create operation #53" << std::endl; |
| 1587 | return ExecutionPlan(); |
| 1588 | } |
| 1589 | operators.emplace_back(op53, xnn_delete_operator); |
| 1590 | |
| 1591 | xnn_operator_t op54 = nullptr; |
| 1592 | status = xnn_create_convolution2d_nchw_f32( |
| 1593 | 1 /* top padding */, 1 /* right padding */, |
| 1594 | 1 /* bottom padding */, 1 /* left padding */, |
| 1595 | 3 /* kernel height */, 3 /* kernel width */, |
| 1596 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1597 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1598 | 184 /* groups */, |
| 1599 | 1 /* input channels per group */, |
| 1600 | 1 /* output_channels_per_group */, |
| 1601 | 184 /* input pixel stride */, |
| 1602 | 184 /* output pixel stride */, |
| 1603 | w182.data(), w183.data(), |
| 1604 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1605 | 0 /* flags */, |
| 1606 | &op54); |
| 1607 | if (status != xnn_status_success) { |
| 1608 | std::cerr << "failed to create operation #54" << std::endl; |
| 1609 | return ExecutionPlan(); |
| 1610 | } |
| 1611 | operators.emplace_back(op54, xnn_delete_operator); |
| 1612 | |
| 1613 | xnn_operator_t op55 = nullptr; |
| 1614 | status = xnn_create_hardswish_nc_f32( |
| 1615 | 184 /* channels */, |
| 1616 | 184 /* input stride */, |
| 1617 | 184 /* output stride */, |
| 1618 | 0 /* flags */, |
| 1619 | &op55); |
| 1620 | if (status != xnn_status_success) { |
| 1621 | std::cerr << "failed to create operation #55" << std::endl; |
| 1622 | return ExecutionPlan(); |
| 1623 | } |
| 1624 | operators.emplace_back(op55, xnn_delete_operator); |
| 1625 | |
| 1626 | xnn_operator_t op56 = nullptr; |
| 1627 | status = xnn_create_convolution2d_nchw_f32( |
| 1628 | 0 /* top padding */, 0 /* right padding */, |
| 1629 | 0 /* bottom padding */, 0 /* left padding */, |
| 1630 | 1 /* kernel height */, 1 /* kernel width */, |
| 1631 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1632 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1633 | 1 /* groups */, |
| 1634 | 184 /* input channels per group */, |
| 1635 | 80 /* output_channels_per_group */, |
| 1636 | 184 /* input pixel stride */, |
| 1637 | 80 /* output pixel stride */, |
| 1638 | w184.data(), w185.data(), |
| 1639 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1640 | 0 /* flags */, |
| 1641 | &op56); |
| 1642 | if (status != xnn_status_success) { |
| 1643 | std::cerr << "failed to create operation #56" << std::endl; |
| 1644 | return ExecutionPlan(); |
| 1645 | } |
| 1646 | operators.emplace_back(op56, xnn_delete_operator); |
| 1647 | |
| 1648 | xnn_operator_t op57 = nullptr; |
| 1649 | status = xnn_create_add_nd_f32( |
| 1650 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1651 | 0 /* flags */, |
| 1652 | &op57); |
| 1653 | if (status != xnn_status_success) { |
| 1654 | std::cerr << "failed to create operation #57" << std::endl; |
| 1655 | return ExecutionPlan(); |
| 1656 | } |
| 1657 | operators.emplace_back(op57, xnn_delete_operator); |
| 1658 | |
| 1659 | xnn_operator_t op58 = nullptr; |
| 1660 | status = xnn_create_convolution2d_nchw_f32( |
| 1661 | 0 /* top padding */, 0 /* right padding */, |
| 1662 | 0 /* bottom padding */, 0 /* left padding */, |
| 1663 | 1 /* kernel height */, 1 /* kernel width */, |
| 1664 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1665 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1666 | 1 /* groups */, |
| 1667 | 80 /* input channels per group */, |
| 1668 | 480 /* output_channels_per_group */, |
| 1669 | 80 /* input pixel stride */, |
| 1670 | 480 /* output pixel stride */, |
| 1671 | w186.data(), w187.data(), |
| 1672 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1673 | 0 /* flags */, |
| 1674 | &op58); |
| 1675 | if (status != xnn_status_success) { |
| 1676 | std::cerr << "failed to create operation #58" << std::endl; |
| 1677 | return ExecutionPlan(); |
| 1678 | } |
| 1679 | operators.emplace_back(op58, xnn_delete_operator); |
| 1680 | |
| 1681 | xnn_operator_t op59 = nullptr; |
| 1682 | status = xnn_create_hardswish_nc_f32( |
| 1683 | 480 /* channels */, |
| 1684 | 480 /* input stride */, |
| 1685 | 480 /* output stride */, |
| 1686 | 0 /* flags */, |
| 1687 | &op59); |
| 1688 | if (status != xnn_status_success) { |
| 1689 | std::cerr << "failed to create operation #59" << std::endl; |
| 1690 | return ExecutionPlan(); |
| 1691 | } |
| 1692 | operators.emplace_back(op59, xnn_delete_operator); |
| 1693 | |
| 1694 | xnn_operator_t op60 = nullptr; |
| 1695 | status = xnn_create_convolution2d_nchw_f32( |
| 1696 | 1 /* top padding */, 1 /* right padding */, |
| 1697 | 1 /* bottom padding */, 1 /* left padding */, |
| 1698 | 3 /* kernel height */, 3 /* kernel width */, |
| 1699 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1700 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1701 | 480 /* groups */, |
| 1702 | 1 /* input channels per group */, |
| 1703 | 1 /* output_channels_per_group */, |
| 1704 | 480 /* input pixel stride */, |
| 1705 | 480 /* output pixel stride */, |
| 1706 | w188.data(), w189.data(), |
| 1707 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1708 | 0 /* flags */, |
| 1709 | &op60); |
| 1710 | if (status != xnn_status_success) { |
| 1711 | std::cerr << "failed to create operation #60" << std::endl; |
| 1712 | return ExecutionPlan(); |
| 1713 | } |
| 1714 | operators.emplace_back(op60, xnn_delete_operator); |
| 1715 | |
| 1716 | xnn_operator_t op61 = nullptr; |
| 1717 | status = xnn_create_hardswish_nc_f32( |
| 1718 | 480 /* channels */, |
| 1719 | 480 /* input stride */, |
| 1720 | 480 /* output stride */, |
| 1721 | 0 /* flags */, |
| 1722 | &op61); |
| 1723 | if (status != xnn_status_success) { |
| 1724 | std::cerr << "failed to create operation #61" << std::endl; |
| 1725 | return ExecutionPlan(); |
| 1726 | } |
| 1727 | operators.emplace_back(op61, xnn_delete_operator); |
| 1728 | |
| 1729 | xnn_operator_t op62 = nullptr; |
| 1730 | status = xnn_create_global_average_pooling_ncw_f32( |
| 1731 | 480 /* channels */, |
| 1732 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1733 | 0 /* flags */, |
| 1734 | &op62); |
| 1735 | if (status != xnn_status_success) { |
| 1736 | std::cerr << "failed to create operation #62" << std::endl; |
| 1737 | return ExecutionPlan(); |
| 1738 | } |
| 1739 | operators.emplace_back(op62, xnn_delete_operator); |
| 1740 | |
| 1741 | xnn_operator_t op63 = nullptr; |
| 1742 | status = xnn_create_convolution2d_nchw_f32( |
| 1743 | 0 /* top padding */, 0 /* right padding */, |
| 1744 | 0 /* bottom padding */, 0 /* left padding */, |
| 1745 | 1 /* kernel height */, 1 /* kernel width */, |
| 1746 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1747 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1748 | 1 /* groups */, |
| 1749 | 480 /* input channels per group */, |
| 1750 | 120 /* output_channels_per_group */, |
| 1751 | 480 /* input pixel stride */, |
| 1752 | 120 /* output pixel stride */, |
| 1753 | w190.data(), w191.data(), |
| 1754 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1755 | 0 /* flags */, |
| 1756 | &op63); |
| 1757 | if (status != xnn_status_success) { |
| 1758 | std::cerr << "failed to create operation #63" << std::endl; |
| 1759 | return ExecutionPlan(); |
| 1760 | } |
| 1761 | operators.emplace_back(op63, xnn_delete_operator); |
| 1762 | |
| 1763 | xnn_operator_t op64 = nullptr; |
| 1764 | status = xnn_create_convolution2d_nchw_f32( |
| 1765 | 0 /* top padding */, 0 /* right padding */, |
| 1766 | 0 /* bottom padding */, 0 /* left padding */, |
| 1767 | 1 /* kernel height */, 1 /* kernel width */, |
| 1768 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1769 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1770 | 1 /* groups */, |
| 1771 | 120 /* input channels per group */, |
| 1772 | 480 /* output_channels_per_group */, |
| 1773 | 120 /* input pixel stride */, |
| 1774 | 480 /* output pixel stride */, |
| 1775 | w192.data(), w193.data(), |
| 1776 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1777 | 0 /* flags */, |
| 1778 | &op64); |
| 1779 | if (status != xnn_status_success) { |
| 1780 | std::cerr << "failed to create operation #64" << std::endl; |
| 1781 | return ExecutionPlan(); |
| 1782 | } |
| 1783 | operators.emplace_back(op64, xnn_delete_operator); |
| 1784 | |
| 1785 | xnn_operator_t op65 = nullptr; |
| 1786 | status = xnn_create_multiply_nd_f32( |
| 1787 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1788 | 0 /* flags */, |
| 1789 | &op65); |
| 1790 | if (status != xnn_status_success) { |
| 1791 | std::cerr << "failed to create operation #65" << std::endl; |
| 1792 | return ExecutionPlan(); |
| 1793 | } |
| 1794 | operators.emplace_back(op65, xnn_delete_operator); |
| 1795 | |
| 1796 | xnn_operator_t op66 = nullptr; |
| 1797 | status = xnn_create_convolution2d_nchw_f32( |
| 1798 | 0 /* top padding */, 0 /* right padding */, |
| 1799 | 0 /* bottom padding */, 0 /* left padding */, |
| 1800 | 1 /* kernel height */, 1 /* kernel width */, |
| 1801 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1802 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1803 | 1 /* groups */, |
| 1804 | 480 /* input channels per group */, |
| 1805 | 112 /* output_channels_per_group */, |
| 1806 | 480 /* input pixel stride */, |
| 1807 | 112 /* output pixel stride */, |
| 1808 | w194.data(), w195.data(), |
| 1809 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1810 | 0 /* flags */, |
| 1811 | &op66); |
| 1812 | if (status != xnn_status_success) { |
| 1813 | std::cerr << "failed to create operation #66" << std::endl; |
| 1814 | return ExecutionPlan(); |
| 1815 | } |
| 1816 | operators.emplace_back(op66, xnn_delete_operator); |
| 1817 | |
| 1818 | xnn_operator_t op67 = nullptr; |
| 1819 | status = xnn_create_convolution2d_nchw_f32( |
| 1820 | 0 /* top padding */, 0 /* right padding */, |
| 1821 | 0 /* bottom padding */, 0 /* left padding */, |
| 1822 | 1 /* kernel height */, 1 /* kernel width */, |
| 1823 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1824 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1825 | 1 /* groups */, |
| 1826 | 112 /* input channels per group */, |
| 1827 | 672 /* output_channels_per_group */, |
| 1828 | 112 /* input pixel stride */, |
| 1829 | 672 /* output pixel stride */, |
| 1830 | w196.data(), w197.data(), |
| 1831 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1832 | 0 /* flags */, |
| 1833 | &op67); |
| 1834 | if (status != xnn_status_success) { |
| 1835 | std::cerr << "failed to create operation #67" << std::endl; |
| 1836 | return ExecutionPlan(); |
| 1837 | } |
| 1838 | operators.emplace_back(op67, xnn_delete_operator); |
| 1839 | |
| 1840 | xnn_operator_t op68 = nullptr; |
| 1841 | status = xnn_create_hardswish_nc_f32( |
| 1842 | 672 /* channels */, |
| 1843 | 672 /* input stride */, |
| 1844 | 672 /* output stride */, |
| 1845 | 0 /* flags */, |
| 1846 | &op68); |
| 1847 | if (status != xnn_status_success) { |
| 1848 | std::cerr << "failed to create operation #68" << std::endl; |
| 1849 | return ExecutionPlan(); |
| 1850 | } |
| 1851 | operators.emplace_back(op68, xnn_delete_operator); |
| 1852 | |
| 1853 | xnn_operator_t op69 = nullptr; |
| 1854 | status = xnn_create_convolution2d_nchw_f32( |
| 1855 | 1 /* top padding */, 1 /* right padding */, |
| 1856 | 1 /* bottom padding */, 1 /* left padding */, |
| 1857 | 3 /* kernel height */, 3 /* kernel width */, |
| 1858 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1859 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1860 | 672 /* groups */, |
| 1861 | 1 /* input channels per group */, |
| 1862 | 1 /* output_channels_per_group */, |
| 1863 | 672 /* input pixel stride */, |
| 1864 | 672 /* output pixel stride */, |
| 1865 | w198.data(), w199.data(), |
| 1866 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1867 | 0 /* flags */, |
| 1868 | &op69); |
| 1869 | if (status != xnn_status_success) { |
| 1870 | std::cerr << "failed to create operation #69" << std::endl; |
| 1871 | return ExecutionPlan(); |
| 1872 | } |
| 1873 | operators.emplace_back(op69, xnn_delete_operator); |
| 1874 | |
| 1875 | xnn_operator_t op70 = nullptr; |
| 1876 | status = xnn_create_hardswish_nc_f32( |
| 1877 | 672 /* channels */, |
| 1878 | 672 /* input stride */, |
| 1879 | 672 /* output stride */, |
| 1880 | 0 /* flags */, |
| 1881 | &op70); |
| 1882 | if (status != xnn_status_success) { |
| 1883 | std::cerr << "failed to create operation #70" << std::endl; |
| 1884 | return ExecutionPlan(); |
| 1885 | } |
| 1886 | operators.emplace_back(op70, xnn_delete_operator); |
| 1887 | |
| 1888 | xnn_operator_t op71 = nullptr; |
| 1889 | status = xnn_create_global_average_pooling_ncw_f32( |
| 1890 | 672 /* channels */, |
| 1891 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1892 | 0 /* flags */, |
| 1893 | &op71); |
| 1894 | if (status != xnn_status_success) { |
| 1895 | std::cerr << "failed to create operation #71" << std::endl; |
| 1896 | return ExecutionPlan(); |
| 1897 | } |
| 1898 | operators.emplace_back(op71, xnn_delete_operator); |
| 1899 | |
| 1900 | xnn_operator_t op72 = nullptr; |
| 1901 | status = xnn_create_convolution2d_nchw_f32( |
| 1902 | 0 /* top padding */, 0 /* right padding */, |
| 1903 | 0 /* bottom padding */, 0 /* left padding */, |
| 1904 | 1 /* kernel height */, 1 /* kernel width */, |
| 1905 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1906 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1907 | 1 /* groups */, |
| 1908 | 672 /* input channels per group */, |
| 1909 | 168 /* output_channels_per_group */, |
| 1910 | 672 /* input pixel stride */, |
| 1911 | 168 /* output pixel stride */, |
| 1912 | w200.data(), w201.data(), |
| 1913 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1914 | 0 /* flags */, |
| 1915 | &op72); |
| 1916 | if (status != xnn_status_success) { |
| 1917 | std::cerr << "failed to create operation #72" << std::endl; |
| 1918 | return ExecutionPlan(); |
| 1919 | } |
| 1920 | operators.emplace_back(op72, xnn_delete_operator); |
| 1921 | |
| 1922 | xnn_operator_t op73 = nullptr; |
| 1923 | status = xnn_create_convolution2d_nchw_f32( |
| 1924 | 0 /* top padding */, 0 /* right padding */, |
| 1925 | 0 /* bottom padding */, 0 /* left padding */, |
| 1926 | 1 /* kernel height */, 1 /* kernel width */, |
| 1927 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1928 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1929 | 1 /* groups */, |
| 1930 | 168 /* input channels per group */, |
| 1931 | 672 /* output_channels_per_group */, |
| 1932 | 168 /* input pixel stride */, |
| 1933 | 672 /* output pixel stride */, |
| 1934 | w202.data(), w203.data(), |
| 1935 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1936 | 0 /* flags */, |
| 1937 | &op73); |
| 1938 | if (status != xnn_status_success) { |
| 1939 | std::cerr << "failed to create operation #73" << std::endl; |
| 1940 | return ExecutionPlan(); |
| 1941 | } |
| 1942 | operators.emplace_back(op73, xnn_delete_operator); |
| 1943 | |
| 1944 | xnn_operator_t op74 = nullptr; |
| 1945 | status = xnn_create_multiply_nd_f32( |
| 1946 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1947 | 0 /* flags */, |
| 1948 | &op74); |
| 1949 | if (status != xnn_status_success) { |
| 1950 | std::cerr << "failed to create operation #74" << std::endl; |
| 1951 | return ExecutionPlan(); |
| 1952 | } |
| 1953 | operators.emplace_back(op74, xnn_delete_operator); |
| 1954 | |
| 1955 | xnn_operator_t op75 = nullptr; |
| 1956 | status = xnn_create_convolution2d_nchw_f32( |
| 1957 | 0 /* top padding */, 0 /* right padding */, |
| 1958 | 0 /* bottom padding */, 0 /* left padding */, |
| 1959 | 1 /* kernel height */, 1 /* kernel width */, |
| 1960 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1961 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1962 | 1 /* groups */, |
| 1963 | 672 /* input channels per group */, |
| 1964 | 112 /* output_channels_per_group */, |
| 1965 | 672 /* input pixel stride */, |
| 1966 | 112 /* output pixel stride */, |
| 1967 | w204.data(), w205.data(), |
| 1968 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1969 | 0 /* flags */, |
| 1970 | &op75); |
| 1971 | if (status != xnn_status_success) { |
| 1972 | std::cerr << "failed to create operation #75" << std::endl; |
| 1973 | return ExecutionPlan(); |
| 1974 | } |
| 1975 | operators.emplace_back(op75, xnn_delete_operator); |
| 1976 | |
| 1977 | xnn_operator_t op76 = nullptr; |
| 1978 | status = xnn_create_add_nd_f32( |
| 1979 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1980 | 0 /* flags */, |
| 1981 | &op76); |
| 1982 | if (status != xnn_status_success) { |
| 1983 | std::cerr << "failed to create operation #76" << std::endl; |
| 1984 | return ExecutionPlan(); |
| 1985 | } |
| 1986 | operators.emplace_back(op76, xnn_delete_operator); |
| 1987 | |
| 1988 | xnn_operator_t op77 = nullptr; |
| 1989 | status = xnn_create_convolution2d_nchw_f32( |
| 1990 | 0 /* top padding */, 0 /* right padding */, |
| 1991 | 0 /* bottom padding */, 0 /* left padding */, |
| 1992 | 1 /* kernel height */, 1 /* kernel width */, |
| 1993 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1994 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1995 | 1 /* groups */, |
| 1996 | 112 /* input channels per group */, |
| 1997 | 672 /* output_channels_per_group */, |
| 1998 | 112 /* input pixel stride */, |
| 1999 | 672 /* output pixel stride */, |
| 2000 | w206.data(), w207.data(), |
| 2001 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2002 | 0 /* flags */, |
| 2003 | &op77); |
| 2004 | if (status != xnn_status_success) { |
| 2005 | std::cerr << "failed to create operation #77" << std::endl; |
| 2006 | return ExecutionPlan(); |
| 2007 | } |
| 2008 | operators.emplace_back(op77, xnn_delete_operator); |
| 2009 | |
| 2010 | xnn_operator_t op78 = nullptr; |
| 2011 | status = xnn_create_hardswish_nc_f32( |
| 2012 | 672 /* channels */, |
| 2013 | 672 /* input stride */, |
| 2014 | 672 /* output stride */, |
| 2015 | 0 /* flags */, |
| 2016 | &op78); |
| 2017 | if (status != xnn_status_success) { |
| 2018 | std::cerr << "failed to create operation #78" << std::endl; |
| 2019 | return ExecutionPlan(); |
| 2020 | } |
| 2021 | operators.emplace_back(op78, xnn_delete_operator); |
| 2022 | |
| 2023 | xnn_operator_t op79 = nullptr; |
| 2024 | status = xnn_create_convolution2d_nchw_f32( |
| 2025 | 2 /* top padding */, 2 /* right padding */, |
| 2026 | 2 /* bottom padding */, 2 /* left padding */, |
| 2027 | 5 /* kernel height */, 5 /* kernel width */, |
| 2028 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 2029 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2030 | 672 /* groups */, |
| 2031 | 1 /* input channels per group */, |
| 2032 | 1 /* output_channels_per_group */, |
| 2033 | 672 /* input pixel stride */, |
| 2034 | 672 /* output pixel stride */, |
| 2035 | w208.data(), w209.data(), |
| 2036 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2037 | 0 /* flags */, |
| 2038 | &op79); |
| 2039 | if (status != xnn_status_success) { |
| 2040 | std::cerr << "failed to create operation #79" << std::endl; |
| 2041 | return ExecutionPlan(); |
| 2042 | } |
| 2043 | operators.emplace_back(op79, xnn_delete_operator); |
| 2044 | |
| 2045 | xnn_operator_t op80 = nullptr; |
| 2046 | status = xnn_create_hardswish_nc_f32( |
| 2047 | 672 /* channels */, |
| 2048 | 672 /* input stride */, |
| 2049 | 672 /* output stride */, |
| 2050 | 0 /* flags */, |
| 2051 | &op80); |
| 2052 | if (status != xnn_status_success) { |
| 2053 | std::cerr << "failed to create operation #80" << std::endl; |
| 2054 | return ExecutionPlan(); |
| 2055 | } |
| 2056 | operators.emplace_back(op80, xnn_delete_operator); |
| 2057 | |
| 2058 | xnn_operator_t op81 = nullptr; |
| 2059 | status = xnn_create_global_average_pooling_ncw_f32( |
| 2060 | 672 /* channels */, |
| 2061 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 2062 | 0 /* flags */, |
| 2063 | &op81); |
| 2064 | if (status != xnn_status_success) { |
| 2065 | std::cerr << "failed to create operation #81" << std::endl; |
| 2066 | return ExecutionPlan(); |
| 2067 | } |
| 2068 | operators.emplace_back(op81, xnn_delete_operator); |
| 2069 | |
| 2070 | xnn_operator_t op82 = nullptr; |
| 2071 | status = xnn_create_convolution2d_nchw_f32( |
| 2072 | 0 /* top padding */, 0 /* right padding */, |
| 2073 | 0 /* bottom padding */, 0 /* left padding */, |
| 2074 | 1 /* kernel height */, 1 /* kernel width */, |
| 2075 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2076 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2077 | 1 /* groups */, |
| 2078 | 672 /* input channels per group */, |
| 2079 | 168 /* output_channels_per_group */, |
| 2080 | 672 /* input pixel stride */, |
| 2081 | 168 /* output pixel stride */, |
| 2082 | w210.data(), w211.data(), |
| 2083 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2084 | 0 /* flags */, |
| 2085 | &op82); |
| 2086 | if (status != xnn_status_success) { |
| 2087 | std::cerr << "failed to create operation #82" << std::endl; |
| 2088 | return ExecutionPlan(); |
| 2089 | } |
| 2090 | operators.emplace_back(op82, xnn_delete_operator); |
| 2091 | |
| 2092 | xnn_operator_t op83 = nullptr; |
| 2093 | status = xnn_create_convolution2d_nchw_f32( |
| 2094 | 0 /* top padding */, 0 /* right padding */, |
| 2095 | 0 /* bottom padding */, 0 /* left padding */, |
| 2096 | 1 /* kernel height */, 1 /* kernel width */, |
| 2097 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2098 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2099 | 1 /* groups */, |
| 2100 | 168 /* input channels per group */, |
| 2101 | 672 /* output_channels_per_group */, |
| 2102 | 168 /* input pixel stride */, |
| 2103 | 672 /* output pixel stride */, |
| 2104 | w212.data(), w213.data(), |
| 2105 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 2106 | 0 /* flags */, |
| 2107 | &op83); |
| 2108 | if (status != xnn_status_success) { |
| 2109 | std::cerr << "failed to create operation #83" << std::endl; |
| 2110 | return ExecutionPlan(); |
| 2111 | } |
| 2112 | operators.emplace_back(op83, xnn_delete_operator); |
| 2113 | |
| 2114 | xnn_operator_t op84 = nullptr; |
| 2115 | status = xnn_create_multiply_nd_f32( |
| 2116 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2117 | 0 /* flags */, |
| 2118 | &op84); |
| 2119 | if (status != xnn_status_success) { |
| 2120 | std::cerr << "failed to create operation #84" << std::endl; |
| 2121 | return ExecutionPlan(); |
| 2122 | } |
| 2123 | operators.emplace_back(op84, xnn_delete_operator); |
| 2124 | |
| 2125 | xnn_operator_t op85 = nullptr; |
| 2126 | status = xnn_create_convolution2d_nchw_f32( |
| 2127 | 0 /* top padding */, 0 /* right padding */, |
| 2128 | 0 /* bottom padding */, 0 /* left padding */, |
| 2129 | 1 /* kernel height */, 1 /* kernel width */, |
| 2130 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2131 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2132 | 1 /* groups */, |
| 2133 | 672 /* input channels per group */, |
| 2134 | 160 /* output_channels_per_group */, |
| 2135 | 672 /* input pixel stride */, |
| 2136 | 160 /* output pixel stride */, |
| 2137 | w214.data(), w215.data(), |
| 2138 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2139 | 0 /* flags */, |
| 2140 | &op85); |
| 2141 | if (status != xnn_status_success) { |
| 2142 | std::cerr << "failed to create operation #85" << std::endl; |
| 2143 | return ExecutionPlan(); |
| 2144 | } |
| 2145 | operators.emplace_back(op85, xnn_delete_operator); |
| 2146 | |
| 2147 | xnn_operator_t op86 = nullptr; |
| 2148 | status = xnn_create_convolution2d_nchw_f32( |
| 2149 | 0 /* top padding */, 0 /* right padding */, |
| 2150 | 0 /* bottom padding */, 0 /* left padding */, |
| 2151 | 1 /* kernel height */, 1 /* kernel width */, |
| 2152 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2153 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2154 | 1 /* groups */, |
| 2155 | 160 /* input channels per group */, |
| 2156 | 960 /* output_channels_per_group */, |
| 2157 | 160 /* input pixel stride */, |
| 2158 | 960 /* output pixel stride */, |
| 2159 | w216.data(), w217.data(), |
| 2160 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2161 | 0 /* flags */, |
| 2162 | &op86); |
| 2163 | if (status != xnn_status_success) { |
| 2164 | std::cerr << "failed to create operation #86" << std::endl; |
| 2165 | return ExecutionPlan(); |
| 2166 | } |
| 2167 | operators.emplace_back(op86, xnn_delete_operator); |
| 2168 | |
| 2169 | xnn_operator_t op87 = nullptr; |
| 2170 | status = xnn_create_hardswish_nc_f32( |
| 2171 | 960 /* channels */, |
| 2172 | 960 /* input stride */, |
| 2173 | 960 /* output stride */, |
| 2174 | 0 /* flags */, |
| 2175 | &op87); |
| 2176 | if (status != xnn_status_success) { |
| 2177 | std::cerr << "failed to create operation #87" << std::endl; |
| 2178 | return ExecutionPlan(); |
| 2179 | } |
| 2180 | operators.emplace_back(op87, xnn_delete_operator); |
| 2181 | |
| 2182 | xnn_operator_t op88 = nullptr; |
| 2183 | status = xnn_create_convolution2d_nchw_f32( |
| 2184 | 2 /* top padding */, 2 /* right padding */, |
| 2185 | 2 /* bottom padding */, 2 /* left padding */, |
| 2186 | 5 /* kernel height */, 5 /* kernel width */, |
| 2187 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2188 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2189 | 960 /* groups */, |
| 2190 | 1 /* input channels per group */, |
| 2191 | 1 /* output_channels_per_group */, |
| 2192 | 960 /* input pixel stride */, |
| 2193 | 960 /* output pixel stride */, |
| 2194 | w218.data(), w219.data(), |
| 2195 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2196 | 0 /* flags */, |
| 2197 | &op88); |
| 2198 | if (status != xnn_status_success) { |
| 2199 | std::cerr << "failed to create operation #88" << std::endl; |
| 2200 | return ExecutionPlan(); |
| 2201 | } |
| 2202 | operators.emplace_back(op88, xnn_delete_operator); |
| 2203 | |
| 2204 | xnn_operator_t op89 = nullptr; |
| 2205 | status = xnn_create_hardswish_nc_f32( |
| 2206 | 960 /* channels */, |
| 2207 | 960 /* input stride */, |
| 2208 | 960 /* output stride */, |
| 2209 | 0 /* flags */, |
| 2210 | &op89); |
| 2211 | if (status != xnn_status_success) { |
| 2212 | std::cerr << "failed to create operation #89" << std::endl; |
| 2213 | return ExecutionPlan(); |
| 2214 | } |
| 2215 | operators.emplace_back(op89, xnn_delete_operator); |
| 2216 | |
| 2217 | xnn_operator_t op90 = nullptr; |
| 2218 | status = xnn_create_global_average_pooling_ncw_f32( |
| 2219 | 960 /* channels */, |
| 2220 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 2221 | 0 /* flags */, |
| 2222 | &op90); |
| 2223 | if (status != xnn_status_success) { |
| 2224 | std::cerr << "failed to create operation #90" << std::endl; |
| 2225 | return ExecutionPlan(); |
| 2226 | } |
| 2227 | operators.emplace_back(op90, xnn_delete_operator); |
| 2228 | |
| 2229 | xnn_operator_t op91 = nullptr; |
| 2230 | status = xnn_create_convolution2d_nchw_f32( |
| 2231 | 0 /* top padding */, 0 /* right padding */, |
| 2232 | 0 /* bottom padding */, 0 /* left padding */, |
| 2233 | 1 /* kernel height */, 1 /* kernel width */, |
| 2234 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2235 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2236 | 1 /* groups */, |
| 2237 | 960 /* input channels per group */, |
| 2238 | 240 /* output_channels_per_group */, |
| 2239 | 960 /* input pixel stride */, |
| 2240 | 240 /* output pixel stride */, |
| 2241 | w220.data(), w221.data(), |
| 2242 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2243 | 0 /* flags */, |
| 2244 | &op91); |
| 2245 | if (status != xnn_status_success) { |
| 2246 | std::cerr << "failed to create operation #91" << std::endl; |
| 2247 | return ExecutionPlan(); |
| 2248 | } |
| 2249 | operators.emplace_back(op91, xnn_delete_operator); |
| 2250 | |
| 2251 | xnn_operator_t op92 = nullptr; |
| 2252 | status = xnn_create_convolution2d_nchw_f32( |
| 2253 | 0 /* top padding */, 0 /* right padding */, |
| 2254 | 0 /* bottom padding */, 0 /* left padding */, |
| 2255 | 1 /* kernel height */, 1 /* kernel width */, |
| 2256 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2257 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2258 | 1 /* groups */, |
| 2259 | 240 /* input channels per group */, |
| 2260 | 960 /* output_channels_per_group */, |
| 2261 | 240 /* input pixel stride */, |
| 2262 | 960 /* output pixel stride */, |
| 2263 | w222.data(), w223.data(), |
| 2264 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 2265 | 0 /* flags */, |
| 2266 | &op92); |
| 2267 | if (status != xnn_status_success) { |
| 2268 | std::cerr << "failed to create operation #92" << std::endl; |
| 2269 | return ExecutionPlan(); |
| 2270 | } |
| 2271 | operators.emplace_back(op92, xnn_delete_operator); |
| 2272 | |
| 2273 | xnn_operator_t op93 = nullptr; |
| 2274 | status = xnn_create_multiply_nd_f32( |
| 2275 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2276 | 0 /* flags */, |
| 2277 | &op93); |
| 2278 | if (status != xnn_status_success) { |
| 2279 | std::cerr << "failed to create operation #93" << std::endl; |
| 2280 | return ExecutionPlan(); |
| 2281 | } |
| 2282 | operators.emplace_back(op93, xnn_delete_operator); |
| 2283 | |
| 2284 | xnn_operator_t op94 = nullptr; |
| 2285 | status = xnn_create_convolution2d_nchw_f32( |
| 2286 | 0 /* top padding */, 0 /* right padding */, |
| 2287 | 0 /* bottom padding */, 0 /* left padding */, |
| 2288 | 1 /* kernel height */, 1 /* kernel width */, |
| 2289 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2290 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2291 | 1 /* groups */, |
| 2292 | 960 /* input channels per group */, |
| 2293 | 160 /* output_channels_per_group */, |
| 2294 | 960 /* input pixel stride */, |
| 2295 | 160 /* output pixel stride */, |
| 2296 | w224.data(), w225.data(), |
| 2297 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2298 | 0 /* flags */, |
| 2299 | &op94); |
| 2300 | if (status != xnn_status_success) { |
| 2301 | std::cerr << "failed to create operation #94" << std::endl; |
| 2302 | return ExecutionPlan(); |
| 2303 | } |
| 2304 | operators.emplace_back(op94, xnn_delete_operator); |
| 2305 | |
| 2306 | xnn_operator_t op95 = nullptr; |
| 2307 | status = xnn_create_add_nd_f32( |
| 2308 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2309 | 0 /* flags */, |
| 2310 | &op95); |
| 2311 | if (status != xnn_status_success) { |
| 2312 | std::cerr << "failed to create operation #95" << std::endl; |
| 2313 | return ExecutionPlan(); |
| 2314 | } |
| 2315 | operators.emplace_back(op95, xnn_delete_operator); |
| 2316 | |
| 2317 | xnn_operator_t op96 = nullptr; |
| 2318 | status = xnn_create_convolution2d_nchw_f32( |
| 2319 | 0 /* top padding */, 0 /* right padding */, |
| 2320 | 0 /* bottom padding */, 0 /* left padding */, |
| 2321 | 1 /* kernel height */, 1 /* kernel width */, |
| 2322 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2323 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2324 | 1 /* groups */, |
| 2325 | 160 /* input channels per group */, |
| 2326 | 960 /* output_channels_per_group */, |
| 2327 | 160 /* input pixel stride */, |
| 2328 | 960 /* output pixel stride */, |
| 2329 | w226.data(), w227.data(), |
| 2330 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2331 | 0 /* flags */, |
| 2332 | &op96); |
| 2333 | if (status != xnn_status_success) { |
| 2334 | std::cerr << "failed to create operation #96" << std::endl; |
| 2335 | return ExecutionPlan(); |
| 2336 | } |
| 2337 | operators.emplace_back(op96, xnn_delete_operator); |
| 2338 | |
| 2339 | xnn_operator_t op97 = nullptr; |
| 2340 | status = xnn_create_hardswish_nc_f32( |
| 2341 | 960 /* channels */, |
| 2342 | 960 /* input stride */, |
| 2343 | 960 /* output stride */, |
| 2344 | 0 /* flags */, |
| 2345 | &op97); |
| 2346 | if (status != xnn_status_success) { |
| 2347 | std::cerr << "failed to create operation #97" << std::endl; |
| 2348 | return ExecutionPlan(); |
| 2349 | } |
| 2350 | operators.emplace_back(op97, xnn_delete_operator); |
| 2351 | |
| 2352 | xnn_operator_t op98 = nullptr; |
| 2353 | status = xnn_create_convolution2d_nchw_f32( |
| 2354 | 2 /* top padding */, 2 /* right padding */, |
| 2355 | 2 /* bottom padding */, 2 /* left padding */, |
| 2356 | 5 /* kernel height */, 5 /* kernel width */, |
| 2357 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2358 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2359 | 960 /* groups */, |
| 2360 | 1 /* input channels per group */, |
| 2361 | 1 /* output_channels_per_group */, |
| 2362 | 960 /* input pixel stride */, |
| 2363 | 960 /* output pixel stride */, |
| 2364 | w228.data(), w229.data(), |
| 2365 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2366 | 0 /* flags */, |
| 2367 | &op98); |
| 2368 | if (status != xnn_status_success) { |
| 2369 | std::cerr << "failed to create operation #98" << std::endl; |
| 2370 | return ExecutionPlan(); |
| 2371 | } |
| 2372 | operators.emplace_back(op98, xnn_delete_operator); |
| 2373 | |
| 2374 | xnn_operator_t op99 = nullptr; |
| 2375 | status = xnn_create_hardswish_nc_f32( |
| 2376 | 960 /* channels */, |
| 2377 | 960 /* input stride */, |
| 2378 | 960 /* output stride */, |
| 2379 | 0 /* flags */, |
| 2380 | &op99); |
| 2381 | if (status != xnn_status_success) { |
| 2382 | std::cerr << "failed to create operation #99" << std::endl; |
| 2383 | return ExecutionPlan(); |
| 2384 | } |
| 2385 | operators.emplace_back(op99, xnn_delete_operator); |
| 2386 | |
| 2387 | xnn_operator_t op100 = nullptr; |
| 2388 | status = xnn_create_global_average_pooling_ncw_f32( |
| 2389 | 960 /* channels */, |
| 2390 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 2391 | 0 /* flags */, |
| 2392 | &op100); |
| 2393 | if (status != xnn_status_success) { |
| 2394 | std::cerr << "failed to create operation #100" << std::endl; |
| 2395 | return ExecutionPlan(); |
| 2396 | } |
| 2397 | operators.emplace_back(op100, xnn_delete_operator); |
| 2398 | |
| 2399 | xnn_operator_t op101 = nullptr; |
| 2400 | status = xnn_create_convolution2d_nchw_f32( |
| 2401 | 0 /* top padding */, 0 /* right padding */, |
| 2402 | 0 /* bottom padding */, 0 /* left padding */, |
| 2403 | 1 /* kernel height */, 1 /* kernel width */, |
| 2404 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2405 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2406 | 1 /* groups */, |
| 2407 | 960 /* input channels per group */, |
| 2408 | 240 /* output_channels_per_group */, |
| 2409 | 960 /* input pixel stride */, |
| 2410 | 240 /* output pixel stride */, |
| 2411 | w230.data(), w231.data(), |
| 2412 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2413 | 0 /* flags */, |
| 2414 | &op101); |
| 2415 | if (status != xnn_status_success) { |
| 2416 | std::cerr << "failed to create operation #101" << std::endl; |
| 2417 | return ExecutionPlan(); |
| 2418 | } |
| 2419 | operators.emplace_back(op101, xnn_delete_operator); |
| 2420 | |
| 2421 | xnn_operator_t op102 = nullptr; |
| 2422 | status = xnn_create_convolution2d_nchw_f32( |
| 2423 | 0 /* top padding */, 0 /* right padding */, |
| 2424 | 0 /* bottom padding */, 0 /* left padding */, |
| 2425 | 1 /* kernel height */, 1 /* kernel width */, |
| 2426 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2427 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2428 | 1 /* groups */, |
| 2429 | 240 /* input channels per group */, |
| 2430 | 960 /* output_channels_per_group */, |
| 2431 | 240 /* input pixel stride */, |
| 2432 | 960 /* output pixel stride */, |
| 2433 | w232.data(), w233.data(), |
| 2434 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 2435 | 0 /* flags */, |
| 2436 | &op102); |
| 2437 | if (status != xnn_status_success) { |
| 2438 | std::cerr << "failed to create operation #102" << std::endl; |
| 2439 | return ExecutionPlan(); |
| 2440 | } |
| 2441 | operators.emplace_back(op102, xnn_delete_operator); |
| 2442 | |
| 2443 | xnn_operator_t op103 = nullptr; |
| 2444 | status = xnn_create_multiply_nd_f32( |
| 2445 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2446 | 0 /* flags */, |
| 2447 | &op103); |
| 2448 | if (status != xnn_status_success) { |
| 2449 | std::cerr << "failed to create operation #103" << std::endl; |
| 2450 | return ExecutionPlan(); |
| 2451 | } |
| 2452 | operators.emplace_back(op103, xnn_delete_operator); |
| 2453 | |
| 2454 | xnn_operator_t op104 = nullptr; |
| 2455 | status = xnn_create_convolution2d_nchw_f32( |
| 2456 | 0 /* top padding */, 0 /* right padding */, |
| 2457 | 0 /* bottom padding */, 0 /* left padding */, |
| 2458 | 1 /* kernel height */, 1 /* kernel width */, |
| 2459 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2460 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2461 | 1 /* groups */, |
| 2462 | 960 /* input channels per group */, |
| 2463 | 160 /* output_channels_per_group */, |
| 2464 | 960 /* input pixel stride */, |
| 2465 | 160 /* output pixel stride */, |
| 2466 | w234.data(), w235.data(), |
| 2467 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2468 | 0 /* flags */, |
| 2469 | &op104); |
| 2470 | if (status != xnn_status_success) { |
| 2471 | std::cerr << "failed to create operation #104" << std::endl; |
| 2472 | return ExecutionPlan(); |
| 2473 | } |
| 2474 | operators.emplace_back(op104, xnn_delete_operator); |
| 2475 | |
| 2476 | xnn_operator_t op105 = nullptr; |
| 2477 | status = xnn_create_add_nd_f32( |
| 2478 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2479 | 0 /* flags */, |
| 2480 | &op105); |
| 2481 | if (status != xnn_status_success) { |
| 2482 | std::cerr << "failed to create operation #105" << std::endl; |
| 2483 | return ExecutionPlan(); |
| 2484 | } |
| 2485 | operators.emplace_back(op105, xnn_delete_operator); |
| 2486 | |
| 2487 | xnn_operator_t op106 = nullptr; |
| 2488 | status = xnn_create_convolution2d_nchw_f32( |
| 2489 | 0 /* top padding */, 0 /* right padding */, |
| 2490 | 0 /* bottom padding */, 0 /* left padding */, |
| 2491 | 1 /* kernel height */, 1 /* kernel width */, |
| 2492 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2493 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2494 | 1 /* groups */, |
| 2495 | 160 /* input channels per group */, |
| 2496 | 960 /* output_channels_per_group */, |
| 2497 | 160 /* input pixel stride */, |
| 2498 | 960 /* output pixel stride */, |
| 2499 | w236.data(), w237.data(), |
| 2500 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2501 | 0 /* flags */, |
| 2502 | &op106); |
| 2503 | if (status != xnn_status_success) { |
| 2504 | std::cerr << "failed to create operation #106" << std::endl; |
| 2505 | return ExecutionPlan(); |
| 2506 | } |
| 2507 | operators.emplace_back(op106, xnn_delete_operator); |
| 2508 | |
| 2509 | xnn_operator_t op107 = nullptr; |
| 2510 | status = xnn_create_hardswish_nc_f32( |
| 2511 | 960 /* channels */, |
| 2512 | 960 /* input stride */, |
| 2513 | 960 /* output stride */, |
| 2514 | 0 /* flags */, |
| 2515 | &op107); |
| 2516 | if (status != xnn_status_success) { |
| 2517 | std::cerr << "failed to create operation #107" << std::endl; |
| 2518 | return ExecutionPlan(); |
| 2519 | } |
| 2520 | operators.emplace_back(op107, xnn_delete_operator); |
| 2521 | |
| 2522 | xnn_operator_t op108 = nullptr; |
| 2523 | status = xnn_create_global_average_pooling_ncw_f32( |
| 2524 | 960 /* channels */, |
| 2525 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 2526 | 0 /* flags */, |
| 2527 | &op108); |
| 2528 | if (status != xnn_status_success) { |
| 2529 | std::cerr << "failed to create operation #108" << std::endl; |
| 2530 | return ExecutionPlan(); |
| 2531 | } |
| 2532 | operators.emplace_back(op108, xnn_delete_operator); |
| 2533 | |
| 2534 | xnn_operator_t op109 = nullptr; |
| 2535 | status = xnn_create_convolution2d_nhwc_f32( |
| 2536 | 0 /* top padding */, 0 /* right padding */, |
| 2537 | 0 /* bottom padding */, 0 /* left padding */, |
| 2538 | 1 /* kernel height */, 1 /* kernel width */, |
| 2539 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2540 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2541 | 1 /* groups */, |
| 2542 | 960 /* input channels per group */, |
| 2543 | 1280 /* output_channels_per_group */, |
| 2544 | 960 /* input pixel stride */, |
| 2545 | 1280 /* output pixel stride */, |
| 2546 | w238.data(), w239.data(), |
| 2547 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2548 | 0 /* flags */, |
| 2549 | &op109); |
| 2550 | if (status != xnn_status_success) { |
| 2551 | std::cerr << "failed to create operation #109" << std::endl; |
| 2552 | return ExecutionPlan(); |
| 2553 | } |
| 2554 | operators.emplace_back(op109, xnn_delete_operator); |
| 2555 | |
| 2556 | xnn_operator_t op110 = nullptr; |
| 2557 | status = xnn_create_hardswish_nc_f32( |
| 2558 | 1280 /* channels */, |
| 2559 | 1280 /* input stride */, |
| 2560 | 1280 /* output stride */, |
| 2561 | 0 /* flags */, |
| 2562 | &op110); |
| 2563 | if (status != xnn_status_success) { |
| 2564 | std::cerr << "failed to create operation #110" << std::endl; |
| 2565 | return ExecutionPlan(); |
| 2566 | } |
| 2567 | operators.emplace_back(op110, xnn_delete_operator); |
| 2568 | |
| 2569 | xnn_operator_t op111 = nullptr; |
| 2570 | status = xnn_create_global_average_pooling_nwc_f32( |
| 2571 | 1280 /* channels */, 1280 /* input stride */, 1280 /* output stride */, |
| 2572 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 2573 | 0 /* flags */, |
| 2574 | &op111); |
| 2575 | if (status != xnn_status_success) { |
| 2576 | std::cerr << "failed to create operation #111" << std::endl; |
| 2577 | return ExecutionPlan(); |
| 2578 | } |
| 2579 | operators.emplace_back(op111, xnn_delete_operator); |
| 2580 | |
| 2581 | xnn_operator_t op112 = nullptr; |
| 2582 | status = xnn_create_convolution2d_nhwc_f32( |
| 2583 | 0 /* top padding */, 0 /* right padding */, |
| 2584 | 0 /* bottom padding */, 0 /* left padding */, |
| 2585 | 1 /* kernel height */, 1 /* kernel width */, |
| 2586 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2587 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2588 | 1 /* groups */, |
| 2589 | 1280 /* input channels per group */, |
| 2590 | 1001 /* output_channels_per_group */, |
| 2591 | 1280 /* input pixel stride */, |
| 2592 | 1001 /* output pixel stride */, |
| 2593 | w240.data(), w241.data(), |
| 2594 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2595 | 0 /* flags */, |
| 2596 | &op112); |
| 2597 | if (status != xnn_status_success) { |
| 2598 | std::cerr << "failed to create operation #112" << std::endl; |
| 2599 | return ExecutionPlan(); |
| 2600 | } |
| 2601 | operators.emplace_back(op112, xnn_delete_operator); |
| 2602 | |
| 2603 | |
| 2604 | |
| 2605 | status = xnn_setup_convolution2d_nchw_f32( |
| 2606 | op0, |
| 2607 | 1 /* batch size */, 224 /* input height */, 224 /* input width */, |
| 2608 | v0.data() /* input */, v1.data() /* output */, |
| 2609 | threadpool /* threadpool */); |
| 2610 | if (status != xnn_status_success) { |
| 2611 | std::cerr << "failed to setup operation #0" << std::endl; |
| 2612 | return ExecutionPlan(); |
| 2613 | } |
| 2614 | |
| 2615 | status = xnn_setup_hardswish_nc_f32( |
| 2616 | op1, |
| 2617 | 12544 /* batch size */, |
| 2618 | v1.data() /* input */, v2.data() /* output */, |
| 2619 | threadpool /* threadpool */); |
| 2620 | if (status != xnn_status_success) { |
| 2621 | std::cerr << "failed to setup operation #1" << std::endl; |
| 2622 | return ExecutionPlan(); |
| 2623 | } |
| 2624 | |
| 2625 | status = xnn_setup_convolution2d_nchw_f32( |
| 2626 | op2, |
| 2627 | 1 /* batch size */, 112 /* input height */, 112 /* input width */, |
| 2628 | v2.data() /* input */, v3.data() /* output */, |
| 2629 | threadpool /* threadpool */); |
| 2630 | if (status != xnn_status_success) { |
| 2631 | std::cerr << "failed to setup operation #2" << std::endl; |
| 2632 | return ExecutionPlan(); |
| 2633 | } |
| 2634 | |
| 2635 | status = xnn_setup_convolution2d_nchw_f32( |
| 2636 | op3, |
| 2637 | 1 /* batch size */, 112 /* input height */, 112 /* input width */, |
| 2638 | v3.data() /* input */, v4.data() /* output */, |
| 2639 | threadpool /* threadpool */); |
| 2640 | if (status != xnn_status_success) { |
| 2641 | std::cerr << "failed to setup operation #3" << std::endl; |
| 2642 | return ExecutionPlan(); |
| 2643 | } |
| 2644 | |
| 2645 | { |
| 2646 | const size_t a_shape[] = { 1, 16, 112, 112 }; |
| 2647 | const size_t b_shape[] = { 1, 16, 112, 112 }; |
| 2648 | status = xnn_setup_add_nd_f32( |
| 2649 | op4, |
| 2650 | 4, a_shape, 4, b_shape, |
| 2651 | v4.data() /* a */, v2.data() /* b */, v5.data() /* output */, |
| 2652 | threadpool /* threadpool */); |
| 2653 | } |
| 2654 | if (status != xnn_status_success) { |
| 2655 | std::cerr << "failed to setup operation #4" << std::endl; |
| 2656 | return ExecutionPlan(); |
| 2657 | } |
| 2658 | |
| 2659 | status = xnn_setup_convolution2d_nchw_f32( |
| 2660 | op5, |
| 2661 | 1 /* batch size */, 112 /* input height */, 112 /* input width */, |
| 2662 | v5.data() /* input */, v6.data() /* output */, |
| 2663 | threadpool /* threadpool */); |
| 2664 | if (status != xnn_status_success) { |
| 2665 | std::cerr << "failed to setup operation #5" << std::endl; |
| 2666 | return ExecutionPlan(); |
| 2667 | } |
| 2668 | |
| 2669 | status = xnn_setup_convolution2d_nchw_f32( |
| 2670 | op6, |
| 2671 | 1 /* batch size */, 112 /* input height */, 112 /* input width */, |
| 2672 | v6.data() /* input */, v7.data() /* output */, |
| 2673 | threadpool /* threadpool */); |
| 2674 | if (status != xnn_status_success) { |
| 2675 | std::cerr << "failed to setup operation #6" << std::endl; |
| 2676 | return ExecutionPlan(); |
| 2677 | } |
| 2678 | |
| 2679 | status = xnn_setup_convolution2d_nchw_f32( |
| 2680 | op7, |
| 2681 | 1 /* batch size */, 56 /* input height */, 56 /* input width */, |
| 2682 | v7.data() /* input */, v8.data() /* output */, |
| 2683 | threadpool /* threadpool */); |
| 2684 | if (status != xnn_status_success) { |
| 2685 | std::cerr << "failed to setup operation #7" << std::endl; |
| 2686 | return ExecutionPlan(); |
| 2687 | } |
| 2688 | |
| 2689 | status = xnn_setup_convolution2d_nchw_f32( |
| 2690 | op8, |
| 2691 | 1 /* batch size */, 56 /* input height */, 56 /* input width */, |
| 2692 | v8.data() /* input */, v9.data() /* output */, |
| 2693 | threadpool /* threadpool */); |
| 2694 | if (status != xnn_status_success) { |
| 2695 | std::cerr << "failed to setup operation #8" << std::endl; |
| 2696 | return ExecutionPlan(); |
| 2697 | } |
| 2698 | |
| 2699 | status = xnn_setup_convolution2d_nchw_f32( |
| 2700 | op9, |
| 2701 | 1 /* batch size */, 56 /* input height */, 56 /* input width */, |
| 2702 | v9.data() /* input */, v10.data() /* output */, |
| 2703 | threadpool /* threadpool */); |
| 2704 | if (status != xnn_status_success) { |
| 2705 | std::cerr << "failed to setup operation #9" << std::endl; |
| 2706 | return ExecutionPlan(); |
| 2707 | } |
| 2708 | |
| 2709 | status = xnn_setup_convolution2d_nchw_f32( |
| 2710 | op10, |
| 2711 | 1 /* batch size */, 56 /* input height */, 56 /* input width */, |
| 2712 | v10.data() /* input */, v11.data() /* output */, |
| 2713 | threadpool /* threadpool */); |
| 2714 | if (status != xnn_status_success) { |
| 2715 | std::cerr << "failed to setup operation #10" << std::endl; |
| 2716 | return ExecutionPlan(); |
| 2717 | } |
| 2718 | |
| 2719 | { |
| 2720 | const size_t a_shape[] = { 1, 24, 56, 56 }; |
| 2721 | const size_t b_shape[] = { 1, 24, 56, 56 }; |
| 2722 | status = xnn_setup_add_nd_f32( |
| 2723 | op11, |
| 2724 | 4, a_shape, 4, b_shape, |
| 2725 | v11.data() /* a */, v8.data() /* b */, v12.data() /* output */, |
| 2726 | threadpool /* threadpool */); |
| 2727 | } |
| 2728 | if (status != xnn_status_success) { |
| 2729 | std::cerr << "failed to setup operation #11" << std::endl; |
| 2730 | return ExecutionPlan(); |
| 2731 | } |
| 2732 | |
| 2733 | status = xnn_setup_convolution2d_nchw_f32( |
| 2734 | op12, |
| 2735 | 1 /* batch size */, 56 /* input height */, 56 /* input width */, |
| 2736 | v12.data() /* input */, v13.data() /* output */, |
| 2737 | threadpool /* threadpool */); |
| 2738 | if (status != xnn_status_success) { |
| 2739 | std::cerr << "failed to setup operation #12" << std::endl; |
| 2740 | return ExecutionPlan(); |
| 2741 | } |
| 2742 | |
| 2743 | status = xnn_setup_convolution2d_nchw_f32( |
| 2744 | op13, |
| 2745 | 1 /* batch size */, 56 /* input height */, 56 /* input width */, |
| 2746 | v13.data() /* input */, v14.data() /* output */, |
| 2747 | threadpool /* threadpool */); |
| 2748 | if (status != xnn_status_success) { |
| 2749 | std::cerr << "failed to setup operation #13" << std::endl; |
| 2750 | return ExecutionPlan(); |
| 2751 | } |
| 2752 | |
| 2753 | status = xnn_setup_global_average_pooling_ncw_f32( |
| 2754 | op14, |
| 2755 | 1 /* batch size */, 784 /* width */, |
| 2756 | v14.data() /* input */, v15.data() /* output */, |
| 2757 | threadpool /* threadpool */); |
| 2758 | if (status != xnn_status_success) { |
| 2759 | std::cerr << "failed to setup operation #14" << std::endl; |
| 2760 | return ExecutionPlan(); |
| 2761 | } |
| 2762 | |
| 2763 | status = xnn_setup_convolution2d_nchw_f32( |
| 2764 | op15, |
| 2765 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2766 | v15.data() /* input */, v16.data() /* output */, |
| 2767 | threadpool /* threadpool */); |
| 2768 | if (status != xnn_status_success) { |
| 2769 | std::cerr << "failed to setup operation #15" << std::endl; |
| 2770 | return ExecutionPlan(); |
| 2771 | } |
| 2772 | |
| 2773 | status = xnn_setup_convolution2d_nchw_f32( |
| 2774 | op16, |
| 2775 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2776 | v16.data() /* input */, v17.data() /* output */, |
| 2777 | threadpool /* threadpool */); |
| 2778 | if (status != xnn_status_success) { |
| 2779 | std::cerr << "failed to setup operation #16" << std::endl; |
| 2780 | return ExecutionPlan(); |
| 2781 | } |
| 2782 | |
| 2783 | { |
| 2784 | const size_t a_shape[] = { 1, 72, 28, 28 }; |
| 2785 | const size_t b_shape[] = { 1, 72, 1, 1 }; |
| 2786 | status = xnn_setup_multiply_nd_f32( |
| 2787 | op17, |
| 2788 | 4, a_shape, 4, b_shape, |
| 2789 | v14.data() /* a */, v17.data() /* b */, v18.data() /* output */, |
| 2790 | threadpool /* threadpool */); |
| 2791 | } |
| 2792 | if (status != xnn_status_success) { |
| 2793 | std::cerr << "failed to setup operation #17" << std::endl; |
| 2794 | return ExecutionPlan(); |
| 2795 | } |
| 2796 | |
| 2797 | status = xnn_setup_convolution2d_nchw_f32( |
| 2798 | op18, |
| 2799 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2800 | v18.data() /* input */, v19.data() /* output */, |
| 2801 | threadpool /* threadpool */); |
| 2802 | if (status != xnn_status_success) { |
| 2803 | std::cerr << "failed to setup operation #18" << std::endl; |
| 2804 | return ExecutionPlan(); |
| 2805 | } |
| 2806 | |
| 2807 | status = xnn_setup_convolution2d_nchw_f32( |
| 2808 | op19, |
| 2809 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2810 | v19.data() /* input */, v20.data() /* output */, |
| 2811 | threadpool /* threadpool */); |
| 2812 | if (status != xnn_status_success) { |
| 2813 | std::cerr << "failed to setup operation #19" << std::endl; |
| 2814 | return ExecutionPlan(); |
| 2815 | } |
| 2816 | |
| 2817 | status = xnn_setup_convolution2d_nchw_f32( |
| 2818 | op20, |
| 2819 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2820 | v20.data() /* input */, v21.data() /* output */, |
| 2821 | threadpool /* threadpool */); |
| 2822 | if (status != xnn_status_success) { |
| 2823 | std::cerr << "failed to setup operation #20" << std::endl; |
| 2824 | return ExecutionPlan(); |
| 2825 | } |
| 2826 | |
| 2827 | status = xnn_setup_global_average_pooling_ncw_f32( |
| 2828 | op21, |
| 2829 | 1 /* batch size */, 784 /* width */, |
| 2830 | v21.data() /* input */, v22.data() /* output */, |
| 2831 | threadpool /* threadpool */); |
| 2832 | if (status != xnn_status_success) { |
| 2833 | std::cerr << "failed to setup operation #21" << std::endl; |
| 2834 | return ExecutionPlan(); |
| 2835 | } |
| 2836 | |
| 2837 | status = xnn_setup_convolution2d_nchw_f32( |
| 2838 | op22, |
| 2839 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2840 | v22.data() /* input */, v23.data() /* output */, |
| 2841 | threadpool /* threadpool */); |
| 2842 | if (status != xnn_status_success) { |
| 2843 | std::cerr << "failed to setup operation #22" << std::endl; |
| 2844 | return ExecutionPlan(); |
| 2845 | } |
| 2846 | |
| 2847 | status = xnn_setup_convolution2d_nchw_f32( |
| 2848 | op23, |
| 2849 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2850 | v23.data() /* input */, v24.data() /* output */, |
| 2851 | threadpool /* threadpool */); |
| 2852 | if (status != xnn_status_success) { |
| 2853 | std::cerr << "failed to setup operation #23" << std::endl; |
| 2854 | return ExecutionPlan(); |
| 2855 | } |
| 2856 | |
| 2857 | { |
| 2858 | const size_t a_shape[] = { 1, 120, 28, 28 }; |
| 2859 | const size_t b_shape[] = { 1, 120, 1, 1 }; |
| 2860 | status = xnn_setup_multiply_nd_f32( |
| 2861 | op24, |
| 2862 | 4, a_shape, 4, b_shape, |
| 2863 | v21.data() /* a */, v24.data() /* b */, v25.data() /* output */, |
| 2864 | threadpool /* threadpool */); |
| 2865 | } |
| 2866 | if (status != xnn_status_success) { |
| 2867 | std::cerr << "failed to setup operation #24" << std::endl; |
| 2868 | return ExecutionPlan(); |
| 2869 | } |
| 2870 | |
| 2871 | status = xnn_setup_convolution2d_nchw_f32( |
| 2872 | op25, |
| 2873 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2874 | v25.data() /* input */, v26.data() /* output */, |
| 2875 | threadpool /* threadpool */); |
| 2876 | if (status != xnn_status_success) { |
| 2877 | std::cerr << "failed to setup operation #25" << std::endl; |
| 2878 | return ExecutionPlan(); |
| 2879 | } |
| 2880 | |
| 2881 | { |
| 2882 | const size_t a_shape[] = { 1, 40, 28, 28 }; |
| 2883 | const size_t b_shape[] = { 1, 40, 28, 28 }; |
| 2884 | status = xnn_setup_add_nd_f32( |
| 2885 | op26, |
| 2886 | 4, a_shape, 4, b_shape, |
| 2887 | v26.data() /* a */, v19.data() /* b */, v27.data() /* output */, |
| 2888 | threadpool /* threadpool */); |
| 2889 | } |
| 2890 | if (status != xnn_status_success) { |
| 2891 | std::cerr << "failed to setup operation #26" << std::endl; |
| 2892 | return ExecutionPlan(); |
| 2893 | } |
| 2894 | |
| 2895 | status = xnn_setup_convolution2d_nchw_f32( |
| 2896 | op27, |
| 2897 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2898 | v27.data() /* input */, v28.data() /* output */, |
| 2899 | threadpool /* threadpool */); |
| 2900 | if (status != xnn_status_success) { |
| 2901 | std::cerr << "failed to setup operation #27" << std::endl; |
| 2902 | return ExecutionPlan(); |
| 2903 | } |
| 2904 | |
| 2905 | status = xnn_setup_convolution2d_nchw_f32( |
| 2906 | op28, |
| 2907 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2908 | v28.data() /* input */, v29.data() /* output */, |
| 2909 | threadpool /* threadpool */); |
| 2910 | if (status != xnn_status_success) { |
| 2911 | std::cerr << "failed to setup operation #28" << std::endl; |
| 2912 | return ExecutionPlan(); |
| 2913 | } |
| 2914 | |
| 2915 | status = xnn_setup_global_average_pooling_ncw_f32( |
| 2916 | op29, |
| 2917 | 1 /* batch size */, 784 /* width */, |
| 2918 | v29.data() /* input */, v30.data() /* output */, |
| 2919 | threadpool /* threadpool */); |
| 2920 | if (status != xnn_status_success) { |
| 2921 | std::cerr << "failed to setup operation #29" << std::endl; |
| 2922 | return ExecutionPlan(); |
| 2923 | } |
| 2924 | |
| 2925 | status = xnn_setup_convolution2d_nchw_f32( |
| 2926 | op30, |
| 2927 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2928 | v30.data() /* input */, v31.data() /* output */, |
| 2929 | threadpool /* threadpool */); |
| 2930 | if (status != xnn_status_success) { |
| 2931 | std::cerr << "failed to setup operation #30" << std::endl; |
| 2932 | return ExecutionPlan(); |
| 2933 | } |
| 2934 | |
| 2935 | status = xnn_setup_convolution2d_nchw_f32( |
| 2936 | op31, |
| 2937 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2938 | v31.data() /* input */, v32.data() /* output */, |
| 2939 | threadpool /* threadpool */); |
| 2940 | if (status != xnn_status_success) { |
| 2941 | std::cerr << "failed to setup operation #31" << std::endl; |
| 2942 | return ExecutionPlan(); |
| 2943 | } |
| 2944 | |
| 2945 | { |
| 2946 | const size_t a_shape[] = { 1, 120, 28, 28 }; |
| 2947 | const size_t b_shape[] = { 1, 120, 1, 1 }; |
| 2948 | status = xnn_setup_multiply_nd_f32( |
| 2949 | op32, |
| 2950 | 4, a_shape, 4, b_shape, |
| 2951 | v29.data() /* a */, v32.data() /* b */, v33.data() /* output */, |
| 2952 | threadpool /* threadpool */); |
| 2953 | } |
| 2954 | if (status != xnn_status_success) { |
| 2955 | std::cerr << "failed to setup operation #32" << std::endl; |
| 2956 | return ExecutionPlan(); |
| 2957 | } |
| 2958 | |
| 2959 | status = xnn_setup_convolution2d_nchw_f32( |
| 2960 | op33, |
| 2961 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2962 | v33.data() /* input */, v34.data() /* output */, |
| 2963 | threadpool /* threadpool */); |
| 2964 | if (status != xnn_status_success) { |
| 2965 | std::cerr << "failed to setup operation #33" << std::endl; |
| 2966 | return ExecutionPlan(); |
| 2967 | } |
| 2968 | |
| 2969 | { |
| 2970 | const size_t a_shape[] = { 1, 40, 28, 28 }; |
| 2971 | const size_t b_shape[] = { 1, 40, 28, 28 }; |
| 2972 | status = xnn_setup_add_nd_f32( |
| 2973 | op34, |
| 2974 | 4, a_shape, 4, b_shape, |
| 2975 | v34.data() /* a */, v27.data() /* b */, v35.data() /* output */, |
| 2976 | threadpool /* threadpool */); |
| 2977 | } |
| 2978 | if (status != xnn_status_success) { |
| 2979 | std::cerr << "failed to setup operation #34" << std::endl; |
| 2980 | return ExecutionPlan(); |
| 2981 | } |
| 2982 | |
| 2983 | status = xnn_setup_convolution2d_nchw_f32( |
| 2984 | op35, |
| 2985 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2986 | v35.data() /* input */, v36.data() /* output */, |
| 2987 | threadpool /* threadpool */); |
| 2988 | if (status != xnn_status_success) { |
| 2989 | std::cerr << "failed to setup operation #35" << std::endl; |
| 2990 | return ExecutionPlan(); |
| 2991 | } |
| 2992 | |
| 2993 | status = xnn_setup_hardswish_nc_f32( |
| 2994 | op36, |
| 2995 | 784 /* batch size */, |
| 2996 | v36.data() /* input */, v37.data() /* output */, |
| 2997 | threadpool /* threadpool */); |
| 2998 | if (status != xnn_status_success) { |
| 2999 | std::cerr << "failed to setup operation #36" << std::endl; |
| 3000 | return ExecutionPlan(); |
| 3001 | } |
| 3002 | |
| 3003 | status = xnn_setup_convolution2d_nchw_f32( |
| 3004 | op37, |
| 3005 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 3006 | v37.data() /* input */, v38.data() /* output */, |
| 3007 | threadpool /* threadpool */); |
| 3008 | if (status != xnn_status_success) { |
| 3009 | std::cerr << "failed to setup operation #37" << std::endl; |
| 3010 | return ExecutionPlan(); |
| 3011 | } |
| 3012 | |
| 3013 | status = xnn_setup_hardswish_nc_f32( |
| 3014 | op38, |
| 3015 | 196 /* batch size */, |
| 3016 | v38.data() /* input */, v39.data() /* output */, |
| 3017 | threadpool /* threadpool */); |
| 3018 | if (status != xnn_status_success) { |
| 3019 | std::cerr << "failed to setup operation #38" << std::endl; |
| 3020 | return ExecutionPlan(); |
| 3021 | } |
| 3022 | |
| 3023 | status = xnn_setup_convolution2d_nchw_f32( |
| 3024 | op39, |
| 3025 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3026 | v39.data() /* input */, v40.data() /* output */, |
| 3027 | threadpool /* threadpool */); |
| 3028 | if (status != xnn_status_success) { |
| 3029 | std::cerr << "failed to setup operation #39" << std::endl; |
| 3030 | return ExecutionPlan(); |
| 3031 | } |
| 3032 | |
| 3033 | status = xnn_setup_convolution2d_nchw_f32( |
| 3034 | op40, |
| 3035 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3036 | v40.data() /* input */, v41.data() /* output */, |
| 3037 | threadpool /* threadpool */); |
| 3038 | if (status != xnn_status_success) { |
| 3039 | std::cerr << "failed to setup operation #40" << std::endl; |
| 3040 | return ExecutionPlan(); |
| 3041 | } |
| 3042 | |
| 3043 | status = xnn_setup_hardswish_nc_f32( |
| 3044 | op41, |
| 3045 | 196 /* batch size */, |
| 3046 | v41.data() /* input */, v42.data() /* output */, |
| 3047 | threadpool /* threadpool */); |
| 3048 | if (status != xnn_status_success) { |
| 3049 | std::cerr << "failed to setup operation #41" << std::endl; |
| 3050 | return ExecutionPlan(); |
| 3051 | } |
| 3052 | |
| 3053 | status = xnn_setup_convolution2d_nchw_f32( |
| 3054 | op42, |
| 3055 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3056 | v42.data() /* input */, v43.data() /* output */, |
| 3057 | threadpool /* threadpool */); |
| 3058 | if (status != xnn_status_success) { |
| 3059 | std::cerr << "failed to setup operation #42" << std::endl; |
| 3060 | return ExecutionPlan(); |
| 3061 | } |
| 3062 | |
| 3063 | status = xnn_setup_hardswish_nc_f32( |
| 3064 | op43, |
| 3065 | 196 /* batch size */, |
| 3066 | v43.data() /* input */, v44.data() /* output */, |
| 3067 | threadpool /* threadpool */); |
| 3068 | if (status != xnn_status_success) { |
| 3069 | std::cerr << "failed to setup operation #43" << std::endl; |
| 3070 | return ExecutionPlan(); |
| 3071 | } |
| 3072 | |
| 3073 | status = xnn_setup_convolution2d_nchw_f32( |
| 3074 | op44, |
| 3075 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3076 | v44.data() /* input */, v45.data() /* output */, |
| 3077 | threadpool /* threadpool */); |
| 3078 | if (status != xnn_status_success) { |
| 3079 | std::cerr << "failed to setup operation #44" << std::endl; |
| 3080 | return ExecutionPlan(); |
| 3081 | } |
| 3082 | |
| 3083 | { |
| 3084 | const size_t a_shape[] = { 1, 80, 14, 14 }; |
| 3085 | const size_t b_shape[] = { 1, 80, 14, 14 }; |
| 3086 | status = xnn_setup_add_nd_f32( |
| 3087 | op45, |
| 3088 | 4, a_shape, 4, b_shape, |
| 3089 | v45.data() /* a */, v40.data() /* b */, v46.data() /* output */, |
| 3090 | threadpool /* threadpool */); |
| 3091 | } |
| 3092 | if (status != xnn_status_success) { |
| 3093 | std::cerr << "failed to setup operation #45" << std::endl; |
| 3094 | return ExecutionPlan(); |
| 3095 | } |
| 3096 | |
| 3097 | status = xnn_setup_convolution2d_nchw_f32( |
| 3098 | op46, |
| 3099 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3100 | v46.data() /* input */, v47.data() /* output */, |
| 3101 | threadpool /* threadpool */); |
| 3102 | if (status != xnn_status_success) { |
| 3103 | std::cerr << "failed to setup operation #46" << std::endl; |
| 3104 | return ExecutionPlan(); |
| 3105 | } |
| 3106 | |
| 3107 | status = xnn_setup_hardswish_nc_f32( |
| 3108 | op47, |
| 3109 | 196 /* batch size */, |
| 3110 | v47.data() /* input */, v48.data() /* output */, |
| 3111 | threadpool /* threadpool */); |
| 3112 | if (status != xnn_status_success) { |
| 3113 | std::cerr << "failed to setup operation #47" << std::endl; |
| 3114 | return ExecutionPlan(); |
| 3115 | } |
| 3116 | |
| 3117 | status = xnn_setup_convolution2d_nchw_f32( |
| 3118 | op48, |
| 3119 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3120 | v48.data() /* input */, v49.data() /* output */, |
| 3121 | threadpool /* threadpool */); |
| 3122 | if (status != xnn_status_success) { |
| 3123 | std::cerr << "failed to setup operation #48" << std::endl; |
| 3124 | return ExecutionPlan(); |
| 3125 | } |
| 3126 | |
| 3127 | status = xnn_setup_hardswish_nc_f32( |
| 3128 | op49, |
| 3129 | 196 /* batch size */, |
| 3130 | v49.data() /* input */, v50.data() /* output */, |
| 3131 | threadpool /* threadpool */); |
| 3132 | if (status != xnn_status_success) { |
| 3133 | std::cerr << "failed to setup operation #49" << std::endl; |
| 3134 | return ExecutionPlan(); |
| 3135 | } |
| 3136 | |
| 3137 | status = xnn_setup_convolution2d_nchw_f32( |
| 3138 | op50, |
| 3139 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3140 | v50.data() /* input */, v51.data() /* output */, |
| 3141 | threadpool /* threadpool */); |
| 3142 | if (status != xnn_status_success) { |
| 3143 | std::cerr << "failed to setup operation #50" << std::endl; |
| 3144 | return ExecutionPlan(); |
| 3145 | } |
| 3146 | |
| 3147 | { |
| 3148 | const size_t a_shape[] = { 1, 80, 14, 14 }; |
| 3149 | const size_t b_shape[] = { 1, 80, 14, 14 }; |
| 3150 | status = xnn_setup_add_nd_f32( |
| 3151 | op51, |
| 3152 | 4, a_shape, 4, b_shape, |
| 3153 | v51.data() /* a */, v46.data() /* b */, v52.data() /* output */, |
| 3154 | threadpool /* threadpool */); |
| 3155 | } |
| 3156 | if (status != xnn_status_success) { |
| 3157 | std::cerr << "failed to setup operation #51" << std::endl; |
| 3158 | return ExecutionPlan(); |
| 3159 | } |
| 3160 | |
| 3161 | status = xnn_setup_convolution2d_nchw_f32( |
| 3162 | op52, |
| 3163 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3164 | v52.data() /* input */, v53.data() /* output */, |
| 3165 | threadpool /* threadpool */); |
| 3166 | if (status != xnn_status_success) { |
| 3167 | std::cerr << "failed to setup operation #52" << std::endl; |
| 3168 | return ExecutionPlan(); |
| 3169 | } |
| 3170 | |
| 3171 | status = xnn_setup_hardswish_nc_f32( |
| 3172 | op53, |
| 3173 | 196 /* batch size */, |
| 3174 | v53.data() /* input */, v54.data() /* output */, |
| 3175 | threadpool /* threadpool */); |
| 3176 | if (status != xnn_status_success) { |
| 3177 | std::cerr << "failed to setup operation #53" << std::endl; |
| 3178 | return ExecutionPlan(); |
| 3179 | } |
| 3180 | |
| 3181 | status = xnn_setup_convolution2d_nchw_f32( |
| 3182 | op54, |
| 3183 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3184 | v54.data() /* input */, v55.data() /* output */, |
| 3185 | threadpool /* threadpool */); |
| 3186 | if (status != xnn_status_success) { |
| 3187 | std::cerr << "failed to setup operation #54" << std::endl; |
| 3188 | return ExecutionPlan(); |
| 3189 | } |
| 3190 | |
| 3191 | status = xnn_setup_hardswish_nc_f32( |
| 3192 | op55, |
| 3193 | 196 /* batch size */, |
| 3194 | v55.data() /* input */, v56.data() /* output */, |
| 3195 | threadpool /* threadpool */); |
| 3196 | if (status != xnn_status_success) { |
| 3197 | std::cerr << "failed to setup operation #55" << std::endl; |
| 3198 | return ExecutionPlan(); |
| 3199 | } |
| 3200 | |
| 3201 | status = xnn_setup_convolution2d_nchw_f32( |
| 3202 | op56, |
| 3203 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3204 | v56.data() /* input */, v57.data() /* output */, |
| 3205 | threadpool /* threadpool */); |
| 3206 | if (status != xnn_status_success) { |
| 3207 | std::cerr << "failed to setup operation #56" << std::endl; |
| 3208 | return ExecutionPlan(); |
| 3209 | } |
| 3210 | |
| 3211 | { |
| 3212 | const size_t a_shape[] = { 1, 80, 14, 14 }; |
| 3213 | const size_t b_shape[] = { 1, 80, 14, 14 }; |
| 3214 | status = xnn_setup_add_nd_f32( |
| 3215 | op57, |
| 3216 | 4, a_shape, 4, b_shape, |
| 3217 | v57.data() /* a */, v52.data() /* b */, v58.data() /* output */, |
| 3218 | threadpool /* threadpool */); |
| 3219 | } |
| 3220 | if (status != xnn_status_success) { |
| 3221 | std::cerr << "failed to setup operation #57" << std::endl; |
| 3222 | return ExecutionPlan(); |
| 3223 | } |
| 3224 | |
| 3225 | status = xnn_setup_convolution2d_nchw_f32( |
| 3226 | op58, |
| 3227 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3228 | v58.data() /* input */, v59.data() /* output */, |
| 3229 | threadpool /* threadpool */); |
| 3230 | if (status != xnn_status_success) { |
| 3231 | std::cerr << "failed to setup operation #58" << std::endl; |
| 3232 | return ExecutionPlan(); |
| 3233 | } |
| 3234 | |
| 3235 | status = xnn_setup_hardswish_nc_f32( |
| 3236 | op59, |
| 3237 | 196 /* batch size */, |
| 3238 | v59.data() /* input */, v60.data() /* output */, |
| 3239 | threadpool /* threadpool */); |
| 3240 | if (status != xnn_status_success) { |
| 3241 | std::cerr << "failed to setup operation #59" << std::endl; |
| 3242 | return ExecutionPlan(); |
| 3243 | } |
| 3244 | |
| 3245 | status = xnn_setup_convolution2d_nchw_f32( |
| 3246 | op60, |
| 3247 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3248 | v60.data() /* input */, v61.data() /* output */, |
| 3249 | threadpool /* threadpool */); |
| 3250 | if (status != xnn_status_success) { |
| 3251 | std::cerr << "failed to setup operation #60" << std::endl; |
| 3252 | return ExecutionPlan(); |
| 3253 | } |
| 3254 | |
| 3255 | status = xnn_setup_hardswish_nc_f32( |
| 3256 | op61, |
| 3257 | 196 /* batch size */, |
| 3258 | v61.data() /* input */, v62.data() /* output */, |
| 3259 | threadpool /* threadpool */); |
| 3260 | if (status != xnn_status_success) { |
| 3261 | std::cerr << "failed to setup operation #61" << std::endl; |
| 3262 | return ExecutionPlan(); |
| 3263 | } |
| 3264 | |
| 3265 | status = xnn_setup_global_average_pooling_ncw_f32( |
| 3266 | op62, |
| 3267 | 1 /* batch size */, 196 /* width */, |
| 3268 | v62.data() /* input */, v63.data() /* output */, |
| 3269 | threadpool /* threadpool */); |
| 3270 | if (status != xnn_status_success) { |
| 3271 | std::cerr << "failed to setup operation #62" << std::endl; |
| 3272 | return ExecutionPlan(); |
| 3273 | } |
| 3274 | |
| 3275 | status = xnn_setup_convolution2d_nchw_f32( |
| 3276 | op63, |
| 3277 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3278 | v63.data() /* input */, v64.data() /* output */, |
| 3279 | threadpool /* threadpool */); |
| 3280 | if (status != xnn_status_success) { |
| 3281 | std::cerr << "failed to setup operation #63" << std::endl; |
| 3282 | return ExecutionPlan(); |
| 3283 | } |
| 3284 | |
| 3285 | status = xnn_setup_convolution2d_nchw_f32( |
| 3286 | op64, |
| 3287 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3288 | v64.data() /* input */, v65.data() /* output */, |
| 3289 | threadpool /* threadpool */); |
| 3290 | if (status != xnn_status_success) { |
| 3291 | std::cerr << "failed to setup operation #64" << std::endl; |
| 3292 | return ExecutionPlan(); |
| 3293 | } |
| 3294 | |
| 3295 | { |
| 3296 | const size_t a_shape[] = { 1, 480, 14, 14 }; |
| 3297 | const size_t b_shape[] = { 1, 480, 1, 1 }; |
| 3298 | status = xnn_setup_multiply_nd_f32( |
| 3299 | op65, |
| 3300 | 4, a_shape, 4, b_shape, |
| 3301 | v62.data() /* a */, v65.data() /* b */, v66.data() /* output */, |
| 3302 | threadpool /* threadpool */); |
| 3303 | } |
| 3304 | if (status != xnn_status_success) { |
| 3305 | std::cerr << "failed to setup operation #65" << std::endl; |
| 3306 | return ExecutionPlan(); |
| 3307 | } |
| 3308 | |
| 3309 | status = xnn_setup_convolution2d_nchw_f32( |
| 3310 | op66, |
| 3311 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3312 | v66.data() /* input */, v67.data() /* output */, |
| 3313 | threadpool /* threadpool */); |
| 3314 | if (status != xnn_status_success) { |
| 3315 | std::cerr << "failed to setup operation #66" << std::endl; |
| 3316 | return ExecutionPlan(); |
| 3317 | } |
| 3318 | |
| 3319 | status = xnn_setup_convolution2d_nchw_f32( |
| 3320 | op67, |
| 3321 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3322 | v67.data() /* input */, v68.data() /* output */, |
| 3323 | threadpool /* threadpool */); |
| 3324 | if (status != xnn_status_success) { |
| 3325 | std::cerr << "failed to setup operation #67" << std::endl; |
| 3326 | return ExecutionPlan(); |
| 3327 | } |
| 3328 | |
| 3329 | status = xnn_setup_hardswish_nc_f32( |
| 3330 | op68, |
| 3331 | 196 /* batch size */, |
| 3332 | v68.data() /* input */, v69.data() /* output */, |
| 3333 | threadpool /* threadpool */); |
| 3334 | if (status != xnn_status_success) { |
| 3335 | std::cerr << "failed to setup operation #68" << std::endl; |
| 3336 | return ExecutionPlan(); |
| 3337 | } |
| 3338 | |
| 3339 | status = xnn_setup_convolution2d_nchw_f32( |
| 3340 | op69, |
| 3341 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3342 | v69.data() /* input */, v70.data() /* output */, |
| 3343 | threadpool /* threadpool */); |
| 3344 | if (status != xnn_status_success) { |
| 3345 | std::cerr << "failed to setup operation #69" << std::endl; |
| 3346 | return ExecutionPlan(); |
| 3347 | } |
| 3348 | |
| 3349 | status = xnn_setup_hardswish_nc_f32( |
| 3350 | op70, |
| 3351 | 196 /* batch size */, |
| 3352 | v70.data() /* input */, v71.data() /* output */, |
| 3353 | threadpool /* threadpool */); |
| 3354 | if (status != xnn_status_success) { |
| 3355 | std::cerr << "failed to setup operation #70" << std::endl; |
| 3356 | return ExecutionPlan(); |
| 3357 | } |
| 3358 | |
| 3359 | status = xnn_setup_global_average_pooling_ncw_f32( |
| 3360 | op71, |
| 3361 | 1 /* batch size */, 196 /* width */, |
| 3362 | v71.data() /* input */, v72.data() /* output */, |
| 3363 | threadpool /* threadpool */); |
| 3364 | if (status != xnn_status_success) { |
| 3365 | std::cerr << "failed to setup operation #71" << std::endl; |
| 3366 | return ExecutionPlan(); |
| 3367 | } |
| 3368 | |
| 3369 | status = xnn_setup_convolution2d_nchw_f32( |
| 3370 | op72, |
| 3371 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3372 | v72.data() /* input */, v73.data() /* output */, |
| 3373 | threadpool /* threadpool */); |
| 3374 | if (status != xnn_status_success) { |
| 3375 | std::cerr << "failed to setup operation #72" << std::endl; |
| 3376 | return ExecutionPlan(); |
| 3377 | } |
| 3378 | |
| 3379 | status = xnn_setup_convolution2d_nchw_f32( |
| 3380 | op73, |
| 3381 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3382 | v73.data() /* input */, v74.data() /* output */, |
| 3383 | threadpool /* threadpool */); |
| 3384 | if (status != xnn_status_success) { |
| 3385 | std::cerr << "failed to setup operation #73" << std::endl; |
| 3386 | return ExecutionPlan(); |
| 3387 | } |
| 3388 | |
| 3389 | { |
| 3390 | const size_t a_shape[] = { 1, 672, 14, 14 }; |
| 3391 | const size_t b_shape[] = { 1, 672, 1, 1 }; |
| 3392 | status = xnn_setup_multiply_nd_f32( |
| 3393 | op74, |
| 3394 | 4, a_shape, 4, b_shape, |
| 3395 | v71.data() /* a */, v74.data() /* b */, v75.data() /* output */, |
| 3396 | threadpool /* threadpool */); |
| 3397 | } |
| 3398 | if (status != xnn_status_success) { |
| 3399 | std::cerr << "failed to setup operation #74" << std::endl; |
| 3400 | return ExecutionPlan(); |
| 3401 | } |
| 3402 | |
| 3403 | status = xnn_setup_convolution2d_nchw_f32( |
| 3404 | op75, |
| 3405 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3406 | v75.data() /* input */, v76.data() /* output */, |
| 3407 | threadpool /* threadpool */); |
| 3408 | if (status != xnn_status_success) { |
| 3409 | std::cerr << "failed to setup operation #75" << std::endl; |
| 3410 | return ExecutionPlan(); |
| 3411 | } |
| 3412 | |
| 3413 | { |
| 3414 | const size_t a_shape[] = { 1, 112, 14, 14 }; |
| 3415 | const size_t b_shape[] = { 1, 112, 14, 14 }; |
| 3416 | status = xnn_setup_add_nd_f32( |
| 3417 | op76, |
| 3418 | 4, a_shape, 4, b_shape, |
| 3419 | v76.data() /* a */, v67.data() /* b */, v77.data() /* output */, |
| 3420 | threadpool /* threadpool */); |
| 3421 | } |
| 3422 | if (status != xnn_status_success) { |
| 3423 | std::cerr << "failed to setup operation #76" << std::endl; |
| 3424 | return ExecutionPlan(); |
| 3425 | } |
| 3426 | |
| 3427 | status = xnn_setup_convolution2d_nchw_f32( |
| 3428 | op77, |
| 3429 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3430 | v77.data() /* input */, v78.data() /* output */, |
| 3431 | threadpool /* threadpool */); |
| 3432 | if (status != xnn_status_success) { |
| 3433 | std::cerr << "failed to setup operation #77" << std::endl; |
| 3434 | return ExecutionPlan(); |
| 3435 | } |
| 3436 | |
| 3437 | status = xnn_setup_hardswish_nc_f32( |
| 3438 | op78, |
| 3439 | 196 /* batch size */, |
| 3440 | v78.data() /* input */, v79.data() /* output */, |
| 3441 | threadpool /* threadpool */); |
| 3442 | if (status != xnn_status_success) { |
| 3443 | std::cerr << "failed to setup operation #78" << std::endl; |
| 3444 | return ExecutionPlan(); |
| 3445 | } |
| 3446 | |
| 3447 | status = xnn_setup_convolution2d_nchw_f32( |
| 3448 | op79, |
| 3449 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 3450 | v79.data() /* input */, v80.data() /* output */, |
| 3451 | threadpool /* threadpool */); |
| 3452 | if (status != xnn_status_success) { |
| 3453 | std::cerr << "failed to setup operation #79" << std::endl; |
| 3454 | return ExecutionPlan(); |
| 3455 | } |
| 3456 | |
| 3457 | status = xnn_setup_hardswish_nc_f32( |
| 3458 | op80, |
| 3459 | 49 /* batch size */, |
| 3460 | v80.data() /* input */, v81.data() /* output */, |
| 3461 | threadpool /* threadpool */); |
| 3462 | if (status != xnn_status_success) { |
| 3463 | std::cerr << "failed to setup operation #80" << std::endl; |
| 3464 | return ExecutionPlan(); |
| 3465 | } |
| 3466 | |
| 3467 | status = xnn_setup_global_average_pooling_ncw_f32( |
| 3468 | op81, |
| 3469 | 1 /* batch size */, 49 /* width */, |
| 3470 | v81.data() /* input */, v82.data() /* output */, |
| 3471 | threadpool /* threadpool */); |
| 3472 | if (status != xnn_status_success) { |
| 3473 | std::cerr << "failed to setup operation #81" << std::endl; |
| 3474 | return ExecutionPlan(); |
| 3475 | } |
| 3476 | |
| 3477 | status = xnn_setup_convolution2d_nchw_f32( |
| 3478 | op82, |
| 3479 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3480 | v82.data() /* input */, v83.data() /* output */, |
| 3481 | threadpool /* threadpool */); |
| 3482 | if (status != xnn_status_success) { |
| 3483 | std::cerr << "failed to setup operation #82" << std::endl; |
| 3484 | return ExecutionPlan(); |
| 3485 | } |
| 3486 | |
| 3487 | status = xnn_setup_convolution2d_nchw_f32( |
| 3488 | op83, |
| 3489 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3490 | v83.data() /* input */, v84.data() /* output */, |
| 3491 | threadpool /* threadpool */); |
| 3492 | if (status != xnn_status_success) { |
| 3493 | std::cerr << "failed to setup operation #83" << std::endl; |
| 3494 | return ExecutionPlan(); |
| 3495 | } |
| 3496 | |
| 3497 | { |
| 3498 | const size_t a_shape[] = { 1, 672, 7, 7 }; |
| 3499 | const size_t b_shape[] = { 1, 672, 1, 1 }; |
| 3500 | status = xnn_setup_multiply_nd_f32( |
| 3501 | op84, |
| 3502 | 4, a_shape, 4, b_shape, |
| 3503 | v81.data() /* a */, v84.data() /* b */, v85.data() /* output */, |
| 3504 | threadpool /* threadpool */); |
| 3505 | } |
| 3506 | if (status != xnn_status_success) { |
| 3507 | std::cerr << "failed to setup operation #84" << std::endl; |
| 3508 | return ExecutionPlan(); |
| 3509 | } |
| 3510 | |
| 3511 | status = xnn_setup_convolution2d_nchw_f32( |
| 3512 | op85, |
| 3513 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3514 | v85.data() /* input */, v86.data() /* output */, |
| 3515 | threadpool /* threadpool */); |
| 3516 | if (status != xnn_status_success) { |
| 3517 | std::cerr << "failed to setup operation #85" << std::endl; |
| 3518 | return ExecutionPlan(); |
| 3519 | } |
| 3520 | |
| 3521 | status = xnn_setup_convolution2d_nchw_f32( |
| 3522 | op86, |
| 3523 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3524 | v86.data() /* input */, v87.data() /* output */, |
| 3525 | threadpool /* threadpool */); |
| 3526 | if (status != xnn_status_success) { |
| 3527 | std::cerr << "failed to setup operation #86" << std::endl; |
| 3528 | return ExecutionPlan(); |
| 3529 | } |
| 3530 | |
| 3531 | status = xnn_setup_hardswish_nc_f32( |
| 3532 | op87, |
| 3533 | 49 /* batch size */, |
| 3534 | v87.data() /* input */, v88.data() /* output */, |
| 3535 | threadpool /* threadpool */); |
| 3536 | if (status != xnn_status_success) { |
| 3537 | std::cerr << "failed to setup operation #87" << std::endl; |
| 3538 | return ExecutionPlan(); |
| 3539 | } |
| 3540 | |
| 3541 | status = xnn_setup_convolution2d_nchw_f32( |
| 3542 | op88, |
| 3543 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3544 | v88.data() /* input */, v89.data() /* output */, |
| 3545 | threadpool /* threadpool */); |
| 3546 | if (status != xnn_status_success) { |
| 3547 | std::cerr << "failed to setup operation #88" << std::endl; |
| 3548 | return ExecutionPlan(); |
| 3549 | } |
| 3550 | |
| 3551 | status = xnn_setup_hardswish_nc_f32( |
| 3552 | op89, |
| 3553 | 49 /* batch size */, |
| 3554 | v89.data() /* input */, v90.data() /* output */, |
| 3555 | threadpool /* threadpool */); |
| 3556 | if (status != xnn_status_success) { |
| 3557 | std::cerr << "failed to setup operation #89" << std::endl; |
| 3558 | return ExecutionPlan(); |
| 3559 | } |
| 3560 | |
| 3561 | status = xnn_setup_global_average_pooling_ncw_f32( |
| 3562 | op90, |
| 3563 | 1 /* batch size */, 49 /* width */, |
| 3564 | v90.data() /* input */, v91.data() /* output */, |
| 3565 | threadpool /* threadpool */); |
| 3566 | if (status != xnn_status_success) { |
| 3567 | std::cerr << "failed to setup operation #90" << std::endl; |
| 3568 | return ExecutionPlan(); |
| 3569 | } |
| 3570 | |
| 3571 | status = xnn_setup_convolution2d_nchw_f32( |
| 3572 | op91, |
| 3573 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3574 | v91.data() /* input */, v92.data() /* output */, |
| 3575 | threadpool /* threadpool */); |
| 3576 | if (status != xnn_status_success) { |
| 3577 | std::cerr << "failed to setup operation #91" << std::endl; |
| 3578 | return ExecutionPlan(); |
| 3579 | } |
| 3580 | |
| 3581 | status = xnn_setup_convolution2d_nchw_f32( |
| 3582 | op92, |
| 3583 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3584 | v92.data() /* input */, v93.data() /* output */, |
| 3585 | threadpool /* threadpool */); |
| 3586 | if (status != xnn_status_success) { |
| 3587 | std::cerr << "failed to setup operation #92" << std::endl; |
| 3588 | return ExecutionPlan(); |
| 3589 | } |
| 3590 | |
| 3591 | { |
| 3592 | const size_t a_shape[] = { 1, 960, 7, 7 }; |
| 3593 | const size_t b_shape[] = { 1, 960, 1, 1 }; |
| 3594 | status = xnn_setup_multiply_nd_f32( |
| 3595 | op93, |
| 3596 | 4, a_shape, 4, b_shape, |
| 3597 | v90.data() /* a */, v93.data() /* b */, v94.data() /* output */, |
| 3598 | threadpool /* threadpool */); |
| 3599 | } |
| 3600 | if (status != xnn_status_success) { |
| 3601 | std::cerr << "failed to setup operation #93" << std::endl; |
| 3602 | return ExecutionPlan(); |
| 3603 | } |
| 3604 | |
| 3605 | status = xnn_setup_convolution2d_nchw_f32( |
| 3606 | op94, |
| 3607 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3608 | v94.data() /* input */, v95.data() /* output */, |
| 3609 | threadpool /* threadpool */); |
| 3610 | if (status != xnn_status_success) { |
| 3611 | std::cerr << "failed to setup operation #94" << std::endl; |
| 3612 | return ExecutionPlan(); |
| 3613 | } |
| 3614 | |
| 3615 | { |
| 3616 | const size_t a_shape[] = { 1, 160, 7, 7 }; |
| 3617 | const size_t b_shape[] = { 1, 160, 7, 7 }; |
| 3618 | status = xnn_setup_add_nd_f32( |
| 3619 | op95, |
| 3620 | 4, a_shape, 4, b_shape, |
| 3621 | v95.data() /* a */, v86.data() /* b */, v96.data() /* output */, |
| 3622 | threadpool /* threadpool */); |
| 3623 | } |
| 3624 | if (status != xnn_status_success) { |
| 3625 | std::cerr << "failed to setup operation #95" << std::endl; |
| 3626 | return ExecutionPlan(); |
| 3627 | } |
| 3628 | |
| 3629 | status = xnn_setup_convolution2d_nchw_f32( |
| 3630 | op96, |
| 3631 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3632 | v96.data() /* input */, v97.data() /* output */, |
| 3633 | threadpool /* threadpool */); |
| 3634 | if (status != xnn_status_success) { |
| 3635 | std::cerr << "failed to setup operation #96" << std::endl; |
| 3636 | return ExecutionPlan(); |
| 3637 | } |
| 3638 | |
| 3639 | status = xnn_setup_hardswish_nc_f32( |
| 3640 | op97, |
| 3641 | 49 /* batch size */, |
| 3642 | v97.data() /* input */, v98.data() /* output */, |
| 3643 | threadpool /* threadpool */); |
| 3644 | if (status != xnn_status_success) { |
| 3645 | std::cerr << "failed to setup operation #97" << std::endl; |
| 3646 | return ExecutionPlan(); |
| 3647 | } |
| 3648 | |
| 3649 | status = xnn_setup_convolution2d_nchw_f32( |
| 3650 | op98, |
| 3651 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3652 | v98.data() /* input */, v99.data() /* output */, |
| 3653 | threadpool /* threadpool */); |
| 3654 | if (status != xnn_status_success) { |
| 3655 | std::cerr << "failed to setup operation #98" << std::endl; |
| 3656 | return ExecutionPlan(); |
| 3657 | } |
| 3658 | |
| 3659 | status = xnn_setup_hardswish_nc_f32( |
| 3660 | op99, |
| 3661 | 49 /* batch size */, |
| 3662 | v99.data() /* input */, v100.data() /* output */, |
| 3663 | threadpool /* threadpool */); |
| 3664 | if (status != xnn_status_success) { |
| 3665 | std::cerr << "failed to setup operation #99" << std::endl; |
| 3666 | return ExecutionPlan(); |
| 3667 | } |
| 3668 | |
| 3669 | status = xnn_setup_global_average_pooling_ncw_f32( |
| 3670 | op100, |
| 3671 | 1 /* batch size */, 49 /* width */, |
| 3672 | v100.data() /* input */, v101.data() /* output */, |
| 3673 | threadpool /* threadpool */); |
| 3674 | if (status != xnn_status_success) { |
| 3675 | std::cerr << "failed to setup operation #100" << std::endl; |
| 3676 | return ExecutionPlan(); |
| 3677 | } |
| 3678 | |
| 3679 | status = xnn_setup_convolution2d_nchw_f32( |
| 3680 | op101, |
| 3681 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3682 | v101.data() /* input */, v102.data() /* output */, |
| 3683 | threadpool /* threadpool */); |
| 3684 | if (status != xnn_status_success) { |
| 3685 | std::cerr << "failed to setup operation #101" << std::endl; |
| 3686 | return ExecutionPlan(); |
| 3687 | } |
| 3688 | |
| 3689 | status = xnn_setup_convolution2d_nchw_f32( |
| 3690 | op102, |
| 3691 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3692 | v102.data() /* input */, v103.data() /* output */, |
| 3693 | threadpool /* threadpool */); |
| 3694 | if (status != xnn_status_success) { |
| 3695 | std::cerr << "failed to setup operation #102" << std::endl; |
| 3696 | return ExecutionPlan(); |
| 3697 | } |
| 3698 | |
| 3699 | { |
| 3700 | const size_t a_shape[] = { 1, 960, 7, 7 }; |
| 3701 | const size_t b_shape[] = { 1, 960, 1, 1 }; |
| 3702 | status = xnn_setup_multiply_nd_f32( |
| 3703 | op103, |
| 3704 | 4, a_shape, 4, b_shape, |
| 3705 | v100.data() /* a */, v103.data() /* b */, v104.data() /* output */, |
| 3706 | threadpool /* threadpool */); |
| 3707 | } |
| 3708 | if (status != xnn_status_success) { |
| 3709 | std::cerr << "failed to setup operation #103" << std::endl; |
| 3710 | return ExecutionPlan(); |
| 3711 | } |
| 3712 | |
| 3713 | status = xnn_setup_convolution2d_nchw_f32( |
| 3714 | op104, |
| 3715 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3716 | v104.data() /* input */, v105.data() /* output */, |
| 3717 | threadpool /* threadpool */); |
| 3718 | if (status != xnn_status_success) { |
| 3719 | std::cerr << "failed to setup operation #104" << std::endl; |
| 3720 | return ExecutionPlan(); |
| 3721 | } |
| 3722 | |
| 3723 | { |
| 3724 | const size_t a_shape[] = { 1, 160, 7, 7 }; |
| 3725 | const size_t b_shape[] = { 1, 160, 7, 7 }; |
| 3726 | status = xnn_setup_add_nd_f32( |
| 3727 | op105, |
| 3728 | 4, a_shape, 4, b_shape, |
| 3729 | v105.data() /* a */, v96.data() /* b */, v106.data() /* output */, |
| 3730 | threadpool /* threadpool */); |
| 3731 | } |
| 3732 | if (status != xnn_status_success) { |
| 3733 | std::cerr << "failed to setup operation #105" << std::endl; |
| 3734 | return ExecutionPlan(); |
| 3735 | } |
| 3736 | |
| 3737 | status = xnn_setup_convolution2d_nchw_f32( |
| 3738 | op106, |
| 3739 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3740 | v106.data() /* input */, v107.data() /* output */, |
| 3741 | threadpool /* threadpool */); |
| 3742 | if (status != xnn_status_success) { |
| 3743 | std::cerr << "failed to setup operation #106" << std::endl; |
| 3744 | return ExecutionPlan(); |
| 3745 | } |
| 3746 | |
| 3747 | status = xnn_setup_hardswish_nc_f32( |
| 3748 | op107, |
| 3749 | 49 /* batch size */, |
| 3750 | v107.data() /* input */, v108.data() /* output */, |
| 3751 | threadpool /* threadpool */); |
| 3752 | if (status != xnn_status_success) { |
| 3753 | std::cerr << "failed to setup operation #107" << std::endl; |
| 3754 | return ExecutionPlan(); |
| 3755 | } |
| 3756 | |
| 3757 | status = xnn_setup_global_average_pooling_ncw_f32( |
| 3758 | op108, |
| 3759 | 1 /* batch size */, 49 /* width */, |
| 3760 | v108.data() /* input */, v109.data() /* output */, |
| 3761 | threadpool /* threadpool */); |
| 3762 | if (status != xnn_status_success) { |
| 3763 | std::cerr << "failed to setup operation #108" << std::endl; |
| 3764 | return ExecutionPlan(); |
| 3765 | } |
| 3766 | |
| 3767 | status = xnn_setup_convolution2d_nhwc_f32( |
| 3768 | op109, |
| 3769 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3770 | v109.data() /* input */, v110.data() /* output */, |
| 3771 | threadpool /* threadpool */); |
| 3772 | if (status != xnn_status_success) { |
| 3773 | std::cerr << "failed to setup operation #109" << std::endl; |
| 3774 | return ExecutionPlan(); |
| 3775 | } |
| 3776 | |
| 3777 | status = xnn_setup_hardswish_nc_f32( |
| 3778 | op110, |
| 3779 | 1 /* batch size */, |
| 3780 | v110.data() /* input */, v111.data() /* output */, |
| 3781 | threadpool /* threadpool */); |
| 3782 | if (status != xnn_status_success) { |
| 3783 | std::cerr << "failed to setup operation #110" << std::endl; |
| 3784 | return ExecutionPlan(); |
| 3785 | } |
| 3786 | |
| 3787 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 3788 | op111, |
| 3789 | 1 /* batch size */, 1 /* width */, |
| 3790 | v111.data() /* input */, v112.data() /* output */, |
| 3791 | threadpool /* threadpool */); |
| 3792 | if (status != xnn_status_success) { |
| 3793 | std::cerr << "failed to setup operation #111" << std::endl; |
| 3794 | return ExecutionPlan(); |
| 3795 | } |
| 3796 | |
| 3797 | status = xnn_setup_convolution2d_nhwc_f32( |
| 3798 | op112, |
| 3799 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3800 | v112.data() /* input */, v113.data() /* output */, |
| 3801 | threadpool /* threadpool */); |
| 3802 | if (status != xnn_status_success) { |
| 3803 | std::cerr << "failed to setup operation #112" << std::endl; |
| 3804 | return ExecutionPlan(); |
| 3805 | } |
| 3806 | |
| 3807 | #pragma clang diagnostic push |
| 3808 | #pragma clang diagnostic ignored "-Wpessimizing-move" |
| 3809 | return operators; |
| 3810 | #pragma clang diagnostic pop |
| 3811 | } |
| 3812 | |
| 3813 | } // namespace models |