Marat Dukhan | c08cdf5 | 2019-12-09 09:17:51 -0800 | [diff] [blame] | 1 | // Copyright 2019 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 <algorithm> |
| 9 | #include <functional> |
| 10 | #include <iostream> |
| 11 | #include <limits> |
| 12 | #include <random> |
| 13 | |
| 14 | #include "models/models.h" |
| 15 | |
| 16 | namespace models { |
| 17 | |
| 18 | ExecutionPlan MobileNetV3Small(pthreadpool_t threadpool) { |
| 19 | alignas(16) static float v0[150528]; |
| 20 | alignas(16) static float v1[200704]; |
| 21 | alignas(16) static float v2[200704]; |
| 22 | alignas(16) static float v3[50176]; |
| 23 | alignas(16) static float v4[16]; |
| 24 | alignas(16) static float v5[8]; |
| 25 | alignas(16) static float v6[16]; |
| 26 | alignas(16) static float v7[50176]; |
| 27 | alignas(16) static float v8[50176]; |
| 28 | alignas(16) static float v9[225792]; |
| 29 | alignas(16) static float v10[56448]; |
| 30 | alignas(16) static float v11[18816]; |
| 31 | alignas(16) static float v12[68992]; |
| 32 | alignas(16) static float v13[68992]; |
| 33 | alignas(16) static float v14[18816]; |
| 34 | alignas(16) static float v15[18816]; |
| 35 | alignas(16) static float v16[75264]; |
| 36 | alignas(16) static float v17[75264]; |
| 37 | alignas(16) static float v18[18816]; |
| 38 | alignas(16) static float v19[18816]; |
| 39 | alignas(16) static float v20[96]; |
| 40 | alignas(16) static float v21[24]; |
| 41 | alignas(16) static float v22[96]; |
| 42 | alignas(16) static float v23[18816]; |
| 43 | alignas(16) static float v24[7840]; |
| 44 | alignas(16) static float v25[47040]; |
| 45 | alignas(16) static float v26[47040]; |
| 46 | alignas(16) static float v27[47040]; |
| 47 | alignas(16) static float v28[47040]; |
| 48 | alignas(16) static float v29[240]; |
| 49 | alignas(16) static float v30[64]; |
| 50 | alignas(16) static float v31[240]; |
| 51 | alignas(16) static float v32[47040]; |
| 52 | alignas(16) static float v33[7840]; |
| 53 | alignas(16) static float v34[7840]; |
| 54 | alignas(16) static float v35[47040]; |
| 55 | alignas(16) static float v36[47040]; |
| 56 | alignas(16) static float v37[47040]; |
| 57 | alignas(16) static float v38[47040]; |
| 58 | alignas(16) static float v39[240]; |
| 59 | alignas(16) static float v40[64]; |
| 60 | alignas(16) static float v41[240]; |
| 61 | alignas(16) static float v42[47040]; |
| 62 | alignas(16) static float v43[7840]; |
| 63 | alignas(16) static float v44[7840]; |
| 64 | alignas(16) static float v45[23520]; |
| 65 | alignas(16) static float v46[23520]; |
| 66 | alignas(16) static float v47[23520]; |
| 67 | alignas(16) static float v48[23520]; |
| 68 | alignas(16) static float v49[120]; |
| 69 | alignas(16) static float v50[32]; |
| 70 | alignas(16) static float v51[120]; |
| 71 | alignas(16) static float v52[23520]; |
| 72 | alignas(16) static float v53[9408]; |
| 73 | alignas(16) static float v54[28224]; |
| 74 | alignas(16) static float v55[28224]; |
| 75 | alignas(16) static float v56[28224]; |
| 76 | alignas(16) static float v57[28224]; |
| 77 | alignas(16) static float v58[144]; |
| 78 | alignas(16) static float v59[40]; |
| 79 | alignas(16) static float v60[144]; |
| 80 | alignas(16) static float v61[28224]; |
| 81 | alignas(16) static float v62[9408]; |
| 82 | alignas(16) static float v63[9408]; |
| 83 | alignas(16) static float v64[56448]; |
| 84 | alignas(16) static float v65[56448]; |
| 85 | alignas(16) static float v66[14112]; |
| 86 | alignas(16) static float v67[14112]; |
| 87 | alignas(16) static float v68[288]; |
| 88 | alignas(16) static float v69[72]; |
| 89 | alignas(16) static float v70[288]; |
| 90 | alignas(16) static float v71[14112]; |
| 91 | alignas(16) static float v72[4704]; |
| 92 | alignas(16) static float v73[28224]; |
| 93 | alignas(16) static float v74[28224]; |
| 94 | alignas(16) static float v75[28224]; |
| 95 | alignas(16) static float v76[28224]; |
| 96 | alignas(16) static float v77[576]; |
| 97 | alignas(16) static float v78[144]; |
| 98 | alignas(16) static float v79[576]; |
| 99 | alignas(16) static float v80[28224]; |
| 100 | alignas(16) static float v81[4704]; |
| 101 | alignas(16) static float v82[4704]; |
| 102 | alignas(16) static float v83[28224]; |
| 103 | alignas(16) static float v84[28224]; |
| 104 | alignas(16) static float v85[28224]; |
| 105 | alignas(16) static float v86[28224]; |
| 106 | alignas(16) static float v87[576]; |
| 107 | alignas(16) static float v88[144]; |
| 108 | alignas(16) static float v89[576]; |
| 109 | alignas(16) static float v90[28224]; |
| 110 | alignas(16) static float v91[4704]; |
| 111 | alignas(16) static float v92[4704]; |
| 112 | alignas(16) static float v93[28224]; |
| 113 | alignas(16) static float v94[28224]; |
| 114 | alignas(16) static float v95[576]; |
| 115 | alignas(16) static float v96[1024]; |
| 116 | alignas(16) static float v97[1024]; |
| 117 | alignas(16) static float v98[1024]; |
| 118 | alignas(16) static float v99[1001]; |
| 119 | alignas(16) static float w100[432]; |
| 120 | alignas(16) static float w101[16]; |
| 121 | alignas(16) static float w102[144]; |
| 122 | alignas(16) static float w103[16]; |
| 123 | alignas(16) static float w104[128]; |
| 124 | alignas(16) static float w105[8]; |
| 125 | alignas(16) static float w106[128]; |
| 126 | alignas(16) static float w107[16]; |
| 127 | alignas(16) static float w108[256]; |
| 128 | alignas(16) static float w109[16]; |
| 129 | alignas(16) static float w110[1152]; |
| 130 | alignas(16) static float w111[72]; |
| 131 | alignas(16) static float w112[648]; |
| 132 | alignas(16) static float w113[72]; |
| 133 | alignas(16) static float w114[1728]; |
| 134 | alignas(16) static float w115[24]; |
| 135 | alignas(16) static float w116[2112]; |
| 136 | alignas(16) static float w117[88]; |
| 137 | alignas(16) static float w118[792]; |
| 138 | alignas(16) static float w119[88]; |
| 139 | alignas(16) static float w120[2112]; |
| 140 | alignas(16) static float w121[24]; |
| 141 | alignas(16) static float w122[2304]; |
| 142 | alignas(16) static float w123[96]; |
| 143 | alignas(16) static float w124[2400]; |
| 144 | alignas(16) static float w125[96]; |
| 145 | alignas(16) static float w126[2304]; |
| 146 | alignas(16) static float w127[24]; |
| 147 | alignas(16) static float w128[2304]; |
| 148 | alignas(16) static float w129[96]; |
| 149 | alignas(16) static float w130[3840]; |
| 150 | alignas(16) static float w131[40]; |
| 151 | alignas(16) static float w132[9600]; |
| 152 | alignas(16) static float w133[240]; |
| 153 | alignas(16) static float w134[6000]; |
| 154 | alignas(16) static float w135[240]; |
| 155 | alignas(16) static float w136[15360]; |
| 156 | alignas(16) static float w137[64]; |
| 157 | alignas(16) static float w138[15360]; |
| 158 | alignas(16) static float w139[240]; |
| 159 | alignas(16) static float w140[9600]; |
| 160 | alignas(16) static float w141[40]; |
| 161 | alignas(16) static float w142[9600]; |
| 162 | alignas(16) static float w143[240]; |
| 163 | alignas(16) static float w144[6000]; |
| 164 | alignas(16) static float w145[240]; |
| 165 | alignas(16) static float w146[15360]; |
| 166 | alignas(16) static float w147[64]; |
| 167 | alignas(16) static float w148[15360]; |
| 168 | alignas(16) static float w149[240]; |
| 169 | alignas(16) static float w150[9600]; |
| 170 | alignas(16) static float w151[40]; |
| 171 | alignas(16) static float w152[4800]; |
| 172 | alignas(16) static float w153[120]; |
| 173 | alignas(16) static float w154[3000]; |
| 174 | alignas(16) static float w155[120]; |
| 175 | alignas(16) static float w156[3840]; |
| 176 | alignas(16) static float w157[32]; |
| 177 | alignas(16) static float w158[3840]; |
| 178 | alignas(16) static float w159[120]; |
| 179 | alignas(16) static float w160[5760]; |
| 180 | alignas(16) static float w161[48]; |
| 181 | alignas(16) static float w162[6912]; |
| 182 | alignas(16) static float w163[144]; |
| 183 | alignas(16) static float w164[3600]; |
| 184 | alignas(16) static float w165[144]; |
| 185 | alignas(16) static float w166[5760]; |
| 186 | alignas(16) static float w167[40]; |
| 187 | alignas(16) static float w168[5760]; |
| 188 | alignas(16) static float w169[144]; |
| 189 | alignas(16) static float w170[6912]; |
| 190 | alignas(16) static float w171[48]; |
| 191 | alignas(16) static float w172[13824]; |
| 192 | alignas(16) static float w173[288]; |
| 193 | alignas(16) static float w174[7200]; |
| 194 | alignas(16) static float w175[288]; |
| 195 | alignas(16) static float w176[20736]; |
| 196 | alignas(16) static float w177[72]; |
| 197 | alignas(16) static float w178[20736]; |
| 198 | alignas(16) static float w179[288]; |
| 199 | alignas(16) static float w180[27648]; |
| 200 | alignas(16) static float w181[96]; |
| 201 | alignas(16) static float w182[55296]; |
| 202 | alignas(16) static float w183[576]; |
| 203 | alignas(16) static float w184[14400]; |
| 204 | alignas(16) static float w185[576]; |
| 205 | alignas(16) static float w186[82944]; |
| 206 | alignas(16) static float w187[144]; |
| 207 | alignas(16) static float w188[82944]; |
| 208 | alignas(16) static float w189[576]; |
| 209 | alignas(16) static float w190[55296]; |
| 210 | alignas(16) static float w191[96]; |
| 211 | alignas(16) static float w192[55296]; |
| 212 | alignas(16) static float w193[576]; |
| 213 | alignas(16) static float w194[14400]; |
| 214 | alignas(16) static float w195[576]; |
| 215 | alignas(16) static float w196[82944]; |
| 216 | alignas(16) static float w197[144]; |
| 217 | alignas(16) static float w198[82944]; |
| 218 | alignas(16) static float w199[576]; |
| 219 | alignas(16) static float w200[55296]; |
| 220 | alignas(16) static float w201[96]; |
| 221 | alignas(16) static float w202[55296]; |
| 222 | alignas(16) static float w203[576]; |
| 223 | alignas(16) static float w204[589824]; |
| 224 | alignas(16) static float w205[1024]; |
| 225 | alignas(16) static float w206[1025024]; |
| 226 | alignas(16) static float w207[1001]; |
| 227 | |
| 228 | std::random_device random_device; |
| 229 | auto rng = std::mt19937(random_device()); |
| 230 | auto f32rng = std::bind(std::uniform_real_distribution<float>(-1.0f, +1.0f), rng); |
| 231 | std::generate(v0, v0 + 150528, std::ref(f32rng)); |
| 232 | std::generate(w100, w100 + 432, std::ref(f32rng)); |
| 233 | std::generate(w101, w101 + 16, std::ref(f32rng)); |
| 234 | std::generate(w102, w102 + 144, std::ref(f32rng)); |
| 235 | std::generate(w103, w103 + 16, std::ref(f32rng)); |
| 236 | std::generate(w104, w104 + 128, std::ref(f32rng)); |
| 237 | std::generate(w105, w105 + 8, std::ref(f32rng)); |
| 238 | std::generate(w106, w106 + 128, std::ref(f32rng)); |
| 239 | std::generate(w107, w107 + 16, std::ref(f32rng)); |
| 240 | std::generate(w108, w108 + 256, std::ref(f32rng)); |
| 241 | std::generate(w109, w109 + 16, std::ref(f32rng)); |
| 242 | std::generate(w110, w110 + 1152, std::ref(f32rng)); |
| 243 | std::generate(w111, w111 + 72, std::ref(f32rng)); |
| 244 | std::generate(w112, w112 + 648, std::ref(f32rng)); |
| 245 | std::generate(w113, w113 + 72, std::ref(f32rng)); |
| 246 | std::generate(w114, w114 + 1728, std::ref(f32rng)); |
| 247 | std::generate(w115, w115 + 24, std::ref(f32rng)); |
| 248 | std::generate(w116, w116 + 2112, std::ref(f32rng)); |
| 249 | std::generate(w117, w117 + 88, std::ref(f32rng)); |
| 250 | std::generate(w118, w118 + 792, std::ref(f32rng)); |
| 251 | std::generate(w119, w119 + 88, std::ref(f32rng)); |
| 252 | std::generate(w120, w120 + 2112, std::ref(f32rng)); |
| 253 | std::generate(w121, w121 + 24, std::ref(f32rng)); |
| 254 | std::generate(w122, w122 + 2304, std::ref(f32rng)); |
| 255 | std::generate(w123, w123 + 96, std::ref(f32rng)); |
| 256 | std::generate(w124, w124 + 2400, std::ref(f32rng)); |
| 257 | std::generate(w125, w125 + 96, std::ref(f32rng)); |
| 258 | std::generate(w126, w126 + 2304, std::ref(f32rng)); |
| 259 | std::generate(w127, w127 + 24, std::ref(f32rng)); |
| 260 | std::generate(w128, w128 + 2304, std::ref(f32rng)); |
| 261 | std::generate(w129, w129 + 96, std::ref(f32rng)); |
| 262 | std::generate(w130, w130 + 3840, std::ref(f32rng)); |
| 263 | std::generate(w131, w131 + 40, std::ref(f32rng)); |
| 264 | std::generate(w132, w132 + 9600, std::ref(f32rng)); |
| 265 | std::generate(w133, w133 + 240, std::ref(f32rng)); |
| 266 | std::generate(w134, w134 + 6000, std::ref(f32rng)); |
| 267 | std::generate(w135, w135 + 240, std::ref(f32rng)); |
| 268 | std::generate(w136, w136 + 15360, std::ref(f32rng)); |
| 269 | std::generate(w137, w137 + 64, std::ref(f32rng)); |
| 270 | std::generate(w138, w138 + 15360, std::ref(f32rng)); |
| 271 | std::generate(w139, w139 + 240, std::ref(f32rng)); |
| 272 | std::generate(w140, w140 + 9600, std::ref(f32rng)); |
| 273 | std::generate(w141, w141 + 40, std::ref(f32rng)); |
| 274 | std::generate(w142, w142 + 9600, std::ref(f32rng)); |
| 275 | std::generate(w143, w143 + 240, std::ref(f32rng)); |
| 276 | std::generate(w144, w144 + 6000, std::ref(f32rng)); |
| 277 | std::generate(w145, w145 + 240, std::ref(f32rng)); |
| 278 | std::generate(w146, w146 + 15360, std::ref(f32rng)); |
| 279 | std::generate(w147, w147 + 64, std::ref(f32rng)); |
| 280 | std::generate(w148, w148 + 15360, std::ref(f32rng)); |
| 281 | std::generate(w149, w149 + 240, std::ref(f32rng)); |
| 282 | std::generate(w150, w150 + 9600, std::ref(f32rng)); |
| 283 | std::generate(w151, w151 + 40, std::ref(f32rng)); |
| 284 | std::generate(w152, w152 + 4800, std::ref(f32rng)); |
| 285 | std::generate(w153, w153 + 120, std::ref(f32rng)); |
| 286 | std::generate(w154, w154 + 3000, std::ref(f32rng)); |
| 287 | std::generate(w155, w155 + 120, std::ref(f32rng)); |
| 288 | std::generate(w156, w156 + 3840, std::ref(f32rng)); |
| 289 | std::generate(w157, w157 + 32, std::ref(f32rng)); |
| 290 | std::generate(w158, w158 + 3840, std::ref(f32rng)); |
| 291 | std::generate(w159, w159 + 120, std::ref(f32rng)); |
| 292 | std::generate(w160, w160 + 5760, std::ref(f32rng)); |
| 293 | std::generate(w161, w161 + 48, std::ref(f32rng)); |
| 294 | std::generate(w162, w162 + 6912, std::ref(f32rng)); |
| 295 | std::generate(w163, w163 + 144, std::ref(f32rng)); |
| 296 | std::generate(w164, w164 + 3600, std::ref(f32rng)); |
| 297 | std::generate(w165, w165 + 144, std::ref(f32rng)); |
| 298 | std::generate(w166, w166 + 5760, std::ref(f32rng)); |
| 299 | std::generate(w167, w167 + 40, std::ref(f32rng)); |
| 300 | std::generate(w168, w168 + 5760, std::ref(f32rng)); |
| 301 | std::generate(w169, w169 + 144, std::ref(f32rng)); |
| 302 | std::generate(w170, w170 + 6912, std::ref(f32rng)); |
| 303 | std::generate(w171, w171 + 48, std::ref(f32rng)); |
| 304 | std::generate(w172, w172 + 13824, std::ref(f32rng)); |
| 305 | std::generate(w173, w173 + 288, std::ref(f32rng)); |
| 306 | std::generate(w174, w174 + 7200, std::ref(f32rng)); |
| 307 | std::generate(w175, w175 + 288, std::ref(f32rng)); |
| 308 | std::generate(w176, w176 + 20736, std::ref(f32rng)); |
| 309 | std::generate(w177, w177 + 72, std::ref(f32rng)); |
| 310 | std::generate(w178, w178 + 20736, std::ref(f32rng)); |
| 311 | std::generate(w179, w179 + 288, std::ref(f32rng)); |
| 312 | std::generate(w180, w180 + 27648, std::ref(f32rng)); |
| 313 | std::generate(w181, w181 + 96, std::ref(f32rng)); |
| 314 | std::generate(w182, w182 + 55296, std::ref(f32rng)); |
| 315 | std::generate(w183, w183 + 576, std::ref(f32rng)); |
| 316 | std::generate(w184, w184 + 14400, std::ref(f32rng)); |
| 317 | std::generate(w185, w185 + 576, std::ref(f32rng)); |
| 318 | std::generate(w186, w186 + 82944, std::ref(f32rng)); |
| 319 | std::generate(w187, w187 + 144, std::ref(f32rng)); |
| 320 | std::generate(w188, w188 + 82944, std::ref(f32rng)); |
| 321 | std::generate(w189, w189 + 576, std::ref(f32rng)); |
| 322 | std::generate(w190, w190 + 55296, std::ref(f32rng)); |
| 323 | std::generate(w191, w191 + 96, std::ref(f32rng)); |
| 324 | std::generate(w192, w192 + 55296, std::ref(f32rng)); |
| 325 | std::generate(w193, w193 + 576, std::ref(f32rng)); |
| 326 | std::generate(w194, w194 + 14400, std::ref(f32rng)); |
| 327 | std::generate(w195, w195 + 576, std::ref(f32rng)); |
| 328 | std::generate(w196, w196 + 82944, std::ref(f32rng)); |
| 329 | std::generate(w197, w197 + 144, std::ref(f32rng)); |
| 330 | std::generate(w198, w198 + 82944, std::ref(f32rng)); |
| 331 | std::generate(w199, w199 + 576, std::ref(f32rng)); |
| 332 | std::generate(w200, w200 + 55296, std::ref(f32rng)); |
| 333 | std::generate(w201, w201 + 96, std::ref(f32rng)); |
| 334 | std::generate(w202, w202 + 55296, std::ref(f32rng)); |
| 335 | std::generate(w203, w203 + 576, std::ref(f32rng)); |
| 336 | std::generate(w204, w204 + 589824, std::ref(f32rng)); |
| 337 | std::generate(w205, w205 + 1024, std::ref(f32rng)); |
| 338 | std::generate(w206, w206 + 1025024, std::ref(f32rng)); |
| 339 | std::generate(w207, w207 + 1001, std::ref(f32rng)); |
| 340 | |
| 341 | ExecutionPlan operators; |
| 342 | xnn_status status; |
| 343 | |
| 344 | xnn_operator_t op0 = nullptr; |
| 345 | status = xnn_create_convolution2d_nhwc_f32( |
| 346 | 0 /* top padding */, 1 /* right padding */, |
| 347 | 1 /* bottom padding */, 0 /* left padding */, |
| 348 | 3 /* kernel height */, 3 /* kernel width */, |
| 349 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 350 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 351 | 1 /* groups */, |
| 352 | 3 /* input channels per group */, |
| 353 | 16 /* output_channels_per_group */, |
| 354 | 3 /* input pixel stride */, |
| 355 | 16 /* output pixel stride */, |
| 356 | w100, w101, |
| 357 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 358 | 0 /* flags */, |
| 359 | &op0); |
| 360 | if (status != xnn_status_success) { |
| 361 | std::cerr << "failed to create operation #0" << std::endl; |
| 362 | return ExecutionPlan(); |
| 363 | } |
| 364 | operators.emplace_back(op0, xnn_delete_operator); |
| 365 | |
| 366 | xnn_operator_t op1 = nullptr; |
| 367 | status = xnn_create_hardswish_nc_f32( |
| 368 | 16 /* channels */, |
| 369 | 16 /* input stride */, |
| 370 | 16 /* output stride */, |
| 371 | 0 /* flags */, |
| 372 | &op1); |
| 373 | if (status != xnn_status_success) { |
| 374 | std::cerr << "failed to create operation #1" << std::endl; |
| 375 | return ExecutionPlan(); |
| 376 | } |
| 377 | operators.emplace_back(op1, xnn_delete_operator); |
| 378 | |
| 379 | xnn_operator_t op2 = nullptr; |
| 380 | status = xnn_create_convolution2d_nhwc_f32( |
| 381 | 0 /* top padding */, 1 /* right padding */, |
| 382 | 1 /* bottom padding */, 0 /* left padding */, |
| 383 | 3 /* kernel height */, 3 /* kernel width */, |
| 384 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 385 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 386 | 16 /* groups */, |
| 387 | 1 /* input channels per group */, |
| 388 | 1 /* output_channels_per_group */, |
| 389 | 16 /* input pixel stride */, |
| 390 | 16 /* output pixel stride */, |
| 391 | w102, w103, |
| 392 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 393 | 0 /* flags */, |
| 394 | &op2); |
| 395 | if (status != xnn_status_success) { |
| 396 | std::cerr << "failed to create operation #2" << std::endl; |
| 397 | return ExecutionPlan(); |
| 398 | } |
| 399 | operators.emplace_back(op2, xnn_delete_operator); |
| 400 | |
| 401 | xnn_operator_t op3 = nullptr; |
| 402 | status = xnn_create_global_average_pooling_nwc_f32( |
| 403 | 16 /* channels */, 16 /* input stride */, 16 /* output stride */, |
| 404 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 405 | 0 /* flags */, |
| 406 | &op3); |
| 407 | if (status != xnn_status_success) { |
| 408 | std::cerr << "failed to create operation #3" << std::endl; |
| 409 | return ExecutionPlan(); |
| 410 | } |
| 411 | operators.emplace_back(op3, xnn_delete_operator); |
| 412 | |
| 413 | xnn_operator_t op4 = nullptr; |
| 414 | status = xnn_create_convolution2d_nhwc_f32( |
| 415 | 0 /* top padding */, 0 /* right padding */, |
| 416 | 0 /* bottom padding */, 0 /* left padding */, |
| 417 | 1 /* kernel height */, 1 /* kernel width */, |
| 418 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 419 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 420 | 1 /* groups */, |
| 421 | 16 /* input channels per group */, |
| 422 | 8 /* output_channels_per_group */, |
| 423 | 16 /* input pixel stride */, |
| 424 | 8 /* output pixel stride */, |
| 425 | w104, w105, |
| 426 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 427 | 0 /* flags */, |
| 428 | &op4); |
| 429 | if (status != xnn_status_success) { |
| 430 | std::cerr << "failed to create operation #4" << std::endl; |
| 431 | return ExecutionPlan(); |
| 432 | } |
| 433 | operators.emplace_back(op4, xnn_delete_operator); |
| 434 | |
| 435 | xnn_operator_t op5 = nullptr; |
| 436 | status = xnn_create_convolution2d_nhwc_f32( |
| 437 | 0 /* top padding */, 0 /* right padding */, |
| 438 | 0 /* bottom padding */, 0 /* left padding */, |
| 439 | 1 /* kernel height */, 1 /* kernel width */, |
| 440 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 441 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 442 | 1 /* groups */, |
| 443 | 8 /* input channels per group */, |
| 444 | 16 /* output_channels_per_group */, |
| 445 | 8 /* input pixel stride */, |
| 446 | 16 /* output pixel stride */, |
| 447 | w106, w107, |
| 448 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 449 | 0 /* flags */, |
| 450 | &op5); |
| 451 | if (status != xnn_status_success) { |
| 452 | std::cerr << "failed to create operation #5" << std::endl; |
| 453 | return ExecutionPlan(); |
| 454 | } |
| 455 | operators.emplace_back(op5, xnn_delete_operator); |
| 456 | |
| 457 | xnn_operator_t op6 = nullptr; |
| 458 | status = xnn_create_multiply_nd_f32( |
| 459 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 460 | 0 /* flags */, |
| 461 | &op6); |
| 462 | if (status != xnn_status_success) { |
| 463 | std::cerr << "failed to create operation #6" << std::endl; |
| 464 | return ExecutionPlan(); |
| 465 | } |
| 466 | operators.emplace_back(op6, xnn_delete_operator); |
| 467 | |
| 468 | xnn_operator_t op7 = nullptr; |
| 469 | status = xnn_create_convolution2d_nhwc_f32( |
| 470 | 0 /* top padding */, 0 /* right padding */, |
| 471 | 0 /* bottom padding */, 0 /* left padding */, |
| 472 | 1 /* kernel height */, 1 /* kernel width */, |
| 473 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 474 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 475 | 1 /* groups */, |
| 476 | 16 /* input channels per group */, |
| 477 | 16 /* output_channels_per_group */, |
| 478 | 16 /* input pixel stride */, |
| 479 | 16 /* output pixel stride */, |
| 480 | w108, w109, |
| 481 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 482 | 0 /* flags */, |
| 483 | &op7); |
| 484 | if (status != xnn_status_success) { |
| 485 | std::cerr << "failed to create operation #7" << std::endl; |
| 486 | return ExecutionPlan(); |
| 487 | } |
| 488 | operators.emplace_back(op7, xnn_delete_operator); |
| 489 | |
| 490 | xnn_operator_t op8 = nullptr; |
| 491 | status = xnn_create_convolution2d_nhwc_f32( |
| 492 | 0 /* top padding */, 0 /* right padding */, |
| 493 | 0 /* bottom padding */, 0 /* left padding */, |
| 494 | 1 /* kernel height */, 1 /* kernel width */, |
| 495 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 496 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 497 | 1 /* groups */, |
| 498 | 16 /* input channels per group */, |
| 499 | 72 /* output_channels_per_group */, |
| 500 | 16 /* input pixel stride */, |
| 501 | 72 /* output pixel stride */, |
| 502 | w110, w111, |
| 503 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 504 | 0 /* flags */, |
| 505 | &op8); |
| 506 | if (status != xnn_status_success) { |
| 507 | std::cerr << "failed to create operation #8" << std::endl; |
| 508 | return ExecutionPlan(); |
| 509 | } |
| 510 | operators.emplace_back(op8, xnn_delete_operator); |
| 511 | |
| 512 | xnn_operator_t op9 = nullptr; |
| 513 | status = xnn_create_convolution2d_nhwc_f32( |
| 514 | 0 /* top padding */, 1 /* right padding */, |
| 515 | 1 /* bottom padding */, 0 /* left padding */, |
| 516 | 3 /* kernel height */, 3 /* kernel width */, |
| 517 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 518 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 519 | 72 /* groups */, |
| 520 | 1 /* input channels per group */, |
| 521 | 1 /* output_channels_per_group */, |
| 522 | 72 /* input pixel stride */, |
| 523 | 72 /* output pixel stride */, |
| 524 | w112, w113, |
| 525 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 526 | 0 /* flags */, |
| 527 | &op9); |
| 528 | if (status != xnn_status_success) { |
| 529 | std::cerr << "failed to create operation #9" << std::endl; |
| 530 | return ExecutionPlan(); |
| 531 | } |
| 532 | operators.emplace_back(op9, xnn_delete_operator); |
| 533 | |
| 534 | xnn_operator_t op10 = nullptr; |
| 535 | status = xnn_create_convolution2d_nhwc_f32( |
| 536 | 0 /* top padding */, 0 /* right padding */, |
| 537 | 0 /* bottom padding */, 0 /* left padding */, |
| 538 | 1 /* kernel height */, 1 /* kernel width */, |
| 539 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 540 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 541 | 1 /* groups */, |
| 542 | 72 /* input channels per group */, |
| 543 | 24 /* output_channels_per_group */, |
| 544 | 72 /* input pixel stride */, |
| 545 | 24 /* output pixel stride */, |
| 546 | w114, w115, |
| 547 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 548 | 0 /* flags */, |
| 549 | &op10); |
| 550 | if (status != xnn_status_success) { |
| 551 | std::cerr << "failed to create operation #10" << std::endl; |
| 552 | return ExecutionPlan(); |
| 553 | } |
| 554 | operators.emplace_back(op10, xnn_delete_operator); |
| 555 | |
| 556 | xnn_operator_t op11 = nullptr; |
| 557 | status = xnn_create_convolution2d_nhwc_f32( |
| 558 | 0 /* top padding */, 0 /* right padding */, |
| 559 | 0 /* bottom padding */, 0 /* left padding */, |
| 560 | 1 /* kernel height */, 1 /* kernel width */, |
| 561 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 562 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 563 | 1 /* groups */, |
| 564 | 24 /* input channels per group */, |
| 565 | 88 /* output_channels_per_group */, |
| 566 | 24 /* input pixel stride */, |
| 567 | 88 /* output pixel stride */, |
| 568 | w116, w117, |
| 569 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 570 | 0 /* flags */, |
| 571 | &op11); |
| 572 | if (status != xnn_status_success) { |
| 573 | std::cerr << "failed to create operation #11" << std::endl; |
| 574 | return ExecutionPlan(); |
| 575 | } |
| 576 | operators.emplace_back(op11, xnn_delete_operator); |
| 577 | |
| 578 | xnn_operator_t op12 = nullptr; |
| 579 | status = xnn_create_convolution2d_nhwc_f32( |
| 580 | 1 /* top padding */, 1 /* right padding */, |
| 581 | 1 /* bottom padding */, 1 /* left padding */, |
| 582 | 3 /* kernel height */, 3 /* kernel width */, |
| 583 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 584 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 585 | 88 /* groups */, |
| 586 | 1 /* input channels per group */, |
| 587 | 1 /* output_channels_per_group */, |
| 588 | 88 /* input pixel stride */, |
| 589 | 88 /* output pixel stride */, |
| 590 | w118, w119, |
| 591 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 592 | 0 /* flags */, |
| 593 | &op12); |
| 594 | if (status != xnn_status_success) { |
| 595 | std::cerr << "failed to create operation #12" << std::endl; |
| 596 | return ExecutionPlan(); |
| 597 | } |
| 598 | operators.emplace_back(op12, xnn_delete_operator); |
| 599 | |
| 600 | xnn_operator_t op13 = nullptr; |
| 601 | status = xnn_create_convolution2d_nhwc_f32( |
| 602 | 0 /* top padding */, 0 /* right padding */, |
| 603 | 0 /* bottom padding */, 0 /* left padding */, |
| 604 | 1 /* kernel height */, 1 /* kernel width */, |
| 605 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 606 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 607 | 1 /* groups */, |
| 608 | 88 /* input channels per group */, |
| 609 | 24 /* output_channels_per_group */, |
| 610 | 88 /* input pixel stride */, |
| 611 | 24 /* output pixel stride */, |
| 612 | w120, w121, |
| 613 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 614 | 0 /* flags */, |
| 615 | &op13); |
| 616 | if (status != xnn_status_success) { |
| 617 | std::cerr << "failed to create operation #13" << std::endl; |
| 618 | return ExecutionPlan(); |
| 619 | } |
| 620 | operators.emplace_back(op13, xnn_delete_operator); |
| 621 | |
| 622 | xnn_operator_t op14 = nullptr; |
| 623 | status = xnn_create_add_nc_f32( |
| 624 | 24 /* channels */, |
| 625 | 24 /* a stride */, |
| 626 | 24 /* b stride */, |
| 627 | 24 /* c stride */, |
| 628 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 629 | 0 /* flags */, |
| 630 | &op14); |
| 631 | if (status != xnn_status_success) { |
| 632 | std::cerr << "failed to create operation #14" << std::endl; |
| 633 | return ExecutionPlan(); |
| 634 | } |
| 635 | operators.emplace_back(op14, xnn_delete_operator); |
| 636 | |
| 637 | xnn_operator_t op15 = nullptr; |
| 638 | status = xnn_create_convolution2d_nhwc_f32( |
| 639 | 0 /* top padding */, 0 /* right padding */, |
| 640 | 0 /* bottom padding */, 0 /* left padding */, |
| 641 | 1 /* kernel height */, 1 /* kernel width */, |
| 642 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 643 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 644 | 1 /* groups */, |
| 645 | 24 /* input channels per group */, |
| 646 | 96 /* output_channels_per_group */, |
| 647 | 24 /* input pixel stride */, |
| 648 | 96 /* output pixel stride */, |
| 649 | w122, w123, |
| 650 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 651 | 0 /* flags */, |
| 652 | &op15); |
| 653 | if (status != xnn_status_success) { |
| 654 | std::cerr << "failed to create operation #15" << std::endl; |
| 655 | return ExecutionPlan(); |
| 656 | } |
| 657 | operators.emplace_back(op15, xnn_delete_operator); |
| 658 | |
| 659 | xnn_operator_t op16 = nullptr; |
| 660 | status = xnn_create_hardswish_nc_f32( |
| 661 | 96 /* channels */, |
| 662 | 96 /* input stride */, |
| 663 | 96 /* output stride */, |
| 664 | 0 /* flags */, |
| 665 | &op16); |
| 666 | if (status != xnn_status_success) { |
| 667 | std::cerr << "failed to create operation #16" << std::endl; |
| 668 | return ExecutionPlan(); |
| 669 | } |
| 670 | operators.emplace_back(op16, xnn_delete_operator); |
| 671 | |
| 672 | xnn_operator_t op17 = nullptr; |
| 673 | status = xnn_create_convolution2d_nhwc_f32( |
| 674 | 1 /* top padding */, 2 /* right padding */, |
| 675 | 2 /* bottom padding */, 1 /* left padding */, |
| 676 | 5 /* kernel height */, 5 /* kernel width */, |
| 677 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 678 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 679 | 96 /* groups */, |
| 680 | 1 /* input channels per group */, |
| 681 | 1 /* output_channels_per_group */, |
| 682 | 96 /* input pixel stride */, |
| 683 | 96 /* output pixel stride */, |
| 684 | w124, w125, |
| 685 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 686 | 0 /* flags */, |
| 687 | &op17); |
| 688 | if (status != xnn_status_success) { |
| 689 | std::cerr << "failed to create operation #17" << std::endl; |
| 690 | return ExecutionPlan(); |
| 691 | } |
| 692 | operators.emplace_back(op17, xnn_delete_operator); |
| 693 | |
| 694 | xnn_operator_t op18 = nullptr; |
| 695 | status = xnn_create_hardswish_nc_f32( |
| 696 | 96 /* channels */, |
| 697 | 96 /* input stride */, |
| 698 | 96 /* output stride */, |
| 699 | 0 /* flags */, |
| 700 | &op18); |
| 701 | if (status != xnn_status_success) { |
| 702 | std::cerr << "failed to create operation #18" << std::endl; |
| 703 | return ExecutionPlan(); |
| 704 | } |
| 705 | operators.emplace_back(op18, xnn_delete_operator); |
| 706 | |
| 707 | xnn_operator_t op19 = nullptr; |
| 708 | status = xnn_create_global_average_pooling_nwc_f32( |
| 709 | 96 /* channels */, 96 /* input stride */, 96 /* output stride */, |
| 710 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 711 | 0 /* flags */, |
| 712 | &op19); |
| 713 | if (status != xnn_status_success) { |
| 714 | std::cerr << "failed to create operation #19" << std::endl; |
| 715 | return ExecutionPlan(); |
| 716 | } |
| 717 | operators.emplace_back(op19, xnn_delete_operator); |
| 718 | |
| 719 | xnn_operator_t op20 = nullptr; |
| 720 | status = xnn_create_convolution2d_nhwc_f32( |
| 721 | 0 /* top padding */, 0 /* right padding */, |
| 722 | 0 /* bottom padding */, 0 /* left padding */, |
| 723 | 1 /* kernel height */, 1 /* kernel width */, |
| 724 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 725 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 726 | 1 /* groups */, |
| 727 | 96 /* input channels per group */, |
| 728 | 24 /* output_channels_per_group */, |
| 729 | 96 /* input pixel stride */, |
| 730 | 24 /* output pixel stride */, |
| 731 | w126, w127, |
| 732 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 733 | 0 /* flags */, |
| 734 | &op20); |
| 735 | if (status != xnn_status_success) { |
| 736 | std::cerr << "failed to create operation #20" << std::endl; |
| 737 | return ExecutionPlan(); |
| 738 | } |
| 739 | operators.emplace_back(op20, xnn_delete_operator); |
| 740 | |
| 741 | xnn_operator_t op21 = nullptr; |
| 742 | status = xnn_create_convolution2d_nhwc_f32( |
| 743 | 0 /* top padding */, 0 /* right padding */, |
| 744 | 0 /* bottom padding */, 0 /* left padding */, |
| 745 | 1 /* kernel height */, 1 /* kernel width */, |
| 746 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 747 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 748 | 1 /* groups */, |
| 749 | 24 /* input channels per group */, |
| 750 | 96 /* output_channels_per_group */, |
| 751 | 24 /* input pixel stride */, |
| 752 | 96 /* output pixel stride */, |
| 753 | w128, w129, |
| 754 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 755 | 0 /* flags */, |
| 756 | &op21); |
| 757 | if (status != xnn_status_success) { |
| 758 | std::cerr << "failed to create operation #21" << std::endl; |
| 759 | return ExecutionPlan(); |
| 760 | } |
| 761 | operators.emplace_back(op21, xnn_delete_operator); |
| 762 | |
| 763 | xnn_operator_t op22 = nullptr; |
| 764 | status = xnn_create_multiply_nd_f32( |
| 765 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 766 | 0 /* flags */, |
| 767 | &op22); |
| 768 | if (status != xnn_status_success) { |
| 769 | std::cerr << "failed to create operation #22" << std::endl; |
| 770 | return ExecutionPlan(); |
| 771 | } |
| 772 | operators.emplace_back(op22, xnn_delete_operator); |
| 773 | |
| 774 | xnn_operator_t op23 = nullptr; |
| 775 | status = xnn_create_convolution2d_nhwc_f32( |
| 776 | 0 /* top padding */, 0 /* right padding */, |
| 777 | 0 /* bottom padding */, 0 /* left padding */, |
| 778 | 1 /* kernel height */, 1 /* kernel width */, |
| 779 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 780 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 781 | 1 /* groups */, |
| 782 | 96 /* input channels per group */, |
| 783 | 40 /* output_channels_per_group */, |
| 784 | 96 /* input pixel stride */, |
| 785 | 40 /* output pixel stride */, |
| 786 | w130, w131, |
| 787 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 788 | 0 /* flags */, |
| 789 | &op23); |
| 790 | if (status != xnn_status_success) { |
| 791 | std::cerr << "failed to create operation #23" << std::endl; |
| 792 | return ExecutionPlan(); |
| 793 | } |
| 794 | operators.emplace_back(op23, xnn_delete_operator); |
| 795 | |
| 796 | xnn_operator_t op24 = nullptr; |
| 797 | status = xnn_create_convolution2d_nhwc_f32( |
| 798 | 0 /* top padding */, 0 /* right padding */, |
| 799 | 0 /* bottom padding */, 0 /* left padding */, |
| 800 | 1 /* kernel height */, 1 /* kernel width */, |
| 801 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 802 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 803 | 1 /* groups */, |
| 804 | 40 /* input channels per group */, |
| 805 | 240 /* output_channels_per_group */, |
| 806 | 40 /* input pixel stride */, |
| 807 | 240 /* output pixel stride */, |
| 808 | w132, w133, |
| 809 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 810 | 0 /* flags */, |
| 811 | &op24); |
| 812 | if (status != xnn_status_success) { |
| 813 | std::cerr << "failed to create operation #24" << std::endl; |
| 814 | return ExecutionPlan(); |
| 815 | } |
| 816 | operators.emplace_back(op24, xnn_delete_operator); |
| 817 | |
| 818 | xnn_operator_t op25 = nullptr; |
| 819 | status = xnn_create_hardswish_nc_f32( |
| 820 | 240 /* channels */, |
| 821 | 240 /* input stride */, |
| 822 | 240 /* output stride */, |
| 823 | 0 /* flags */, |
| 824 | &op25); |
| 825 | if (status != xnn_status_success) { |
| 826 | std::cerr << "failed to create operation #25" << std::endl; |
| 827 | return ExecutionPlan(); |
| 828 | } |
| 829 | operators.emplace_back(op25, xnn_delete_operator); |
| 830 | |
| 831 | xnn_operator_t op26 = nullptr; |
| 832 | status = xnn_create_convolution2d_nhwc_f32( |
| 833 | 2 /* top padding */, 2 /* right padding */, |
| 834 | 2 /* bottom padding */, 2 /* left padding */, |
| 835 | 5 /* kernel height */, 5 /* kernel width */, |
| 836 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 837 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 838 | 240 /* groups */, |
| 839 | 1 /* input channels per group */, |
| 840 | 1 /* output_channels_per_group */, |
| 841 | 240 /* input pixel stride */, |
| 842 | 240 /* output pixel stride */, |
| 843 | w134, w135, |
| 844 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 845 | 0 /* flags */, |
| 846 | &op26); |
| 847 | if (status != xnn_status_success) { |
| 848 | std::cerr << "failed to create operation #26" << std::endl; |
| 849 | return ExecutionPlan(); |
| 850 | } |
| 851 | operators.emplace_back(op26, xnn_delete_operator); |
| 852 | |
| 853 | xnn_operator_t op27 = nullptr; |
| 854 | status = xnn_create_hardswish_nc_f32( |
| 855 | 240 /* channels */, |
| 856 | 240 /* input stride */, |
| 857 | 240 /* output stride */, |
| 858 | 0 /* flags */, |
| 859 | &op27); |
| 860 | if (status != xnn_status_success) { |
| 861 | std::cerr << "failed to create operation #27" << std::endl; |
| 862 | return ExecutionPlan(); |
| 863 | } |
| 864 | operators.emplace_back(op27, xnn_delete_operator); |
| 865 | |
| 866 | xnn_operator_t op28 = nullptr; |
| 867 | status = xnn_create_global_average_pooling_nwc_f32( |
| 868 | 240 /* channels */, 240 /* input stride */, 240 /* output stride */, |
| 869 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 870 | 0 /* flags */, |
| 871 | &op28); |
| 872 | if (status != xnn_status_success) { |
| 873 | std::cerr << "failed to create operation #28" << std::endl; |
| 874 | return ExecutionPlan(); |
| 875 | } |
| 876 | operators.emplace_back(op28, xnn_delete_operator); |
| 877 | |
| 878 | xnn_operator_t op29 = nullptr; |
| 879 | status = xnn_create_convolution2d_nhwc_f32( |
| 880 | 0 /* top padding */, 0 /* right padding */, |
| 881 | 0 /* bottom padding */, 0 /* left padding */, |
| 882 | 1 /* kernel height */, 1 /* kernel width */, |
| 883 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 884 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 885 | 1 /* groups */, |
| 886 | 240 /* input channels per group */, |
| 887 | 64 /* output_channels_per_group */, |
| 888 | 240 /* input pixel stride */, |
| 889 | 64 /* output pixel stride */, |
| 890 | w136, w137, |
| 891 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 892 | 0 /* flags */, |
| 893 | &op29); |
| 894 | if (status != xnn_status_success) { |
| 895 | std::cerr << "failed to create operation #29" << std::endl; |
| 896 | return ExecutionPlan(); |
| 897 | } |
| 898 | operators.emplace_back(op29, xnn_delete_operator); |
| 899 | |
| 900 | xnn_operator_t op30 = nullptr; |
| 901 | status = xnn_create_convolution2d_nhwc_f32( |
| 902 | 0 /* top padding */, 0 /* right padding */, |
| 903 | 0 /* bottom padding */, 0 /* left padding */, |
| 904 | 1 /* kernel height */, 1 /* kernel width */, |
| 905 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 906 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 907 | 1 /* groups */, |
| 908 | 64 /* input channels per group */, |
| 909 | 240 /* output_channels_per_group */, |
| 910 | 64 /* input pixel stride */, |
| 911 | 240 /* output pixel stride */, |
| 912 | w138, w139, |
| 913 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 914 | 0 /* flags */, |
| 915 | &op30); |
| 916 | if (status != xnn_status_success) { |
| 917 | std::cerr << "failed to create operation #30" << std::endl; |
| 918 | return ExecutionPlan(); |
| 919 | } |
| 920 | operators.emplace_back(op30, xnn_delete_operator); |
| 921 | |
| 922 | xnn_operator_t op31 = nullptr; |
| 923 | status = xnn_create_multiply_nd_f32( |
| 924 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 925 | 0 /* flags */, |
| 926 | &op31); |
| 927 | if (status != xnn_status_success) { |
| 928 | std::cerr << "failed to create operation #31" << std::endl; |
| 929 | return ExecutionPlan(); |
| 930 | } |
| 931 | operators.emplace_back(op31, xnn_delete_operator); |
| 932 | |
| 933 | xnn_operator_t op32 = nullptr; |
| 934 | status = xnn_create_convolution2d_nhwc_f32( |
| 935 | 0 /* top padding */, 0 /* right padding */, |
| 936 | 0 /* bottom padding */, 0 /* left padding */, |
| 937 | 1 /* kernel height */, 1 /* kernel width */, |
| 938 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 939 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 940 | 1 /* groups */, |
| 941 | 240 /* input channels per group */, |
| 942 | 40 /* output_channels_per_group */, |
| 943 | 240 /* input pixel stride */, |
| 944 | 40 /* output pixel stride */, |
| 945 | w140, w141, |
| 946 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 947 | 0 /* flags */, |
| 948 | &op32); |
| 949 | if (status != xnn_status_success) { |
| 950 | std::cerr << "failed to create operation #32" << std::endl; |
| 951 | return ExecutionPlan(); |
| 952 | } |
| 953 | operators.emplace_back(op32, xnn_delete_operator); |
| 954 | |
| 955 | xnn_operator_t op33 = nullptr; |
| 956 | status = xnn_create_add_nc_f32( |
| 957 | 40 /* channels */, |
| 958 | 40 /* a stride */, |
| 959 | 40 /* b stride */, |
| 960 | 40 /* c stride */, |
| 961 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 962 | 0 /* flags */, |
| 963 | &op33); |
| 964 | if (status != xnn_status_success) { |
| 965 | std::cerr << "failed to create operation #33" << std::endl; |
| 966 | return ExecutionPlan(); |
| 967 | } |
| 968 | operators.emplace_back(op33, xnn_delete_operator); |
| 969 | |
| 970 | xnn_operator_t op34 = nullptr; |
| 971 | status = xnn_create_convolution2d_nhwc_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 | 240 /* output_channels_per_group */, |
| 980 | 40 /* input pixel stride */, |
| 981 | 240 /* output pixel stride */, |
| 982 | w142, w143, |
| 983 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 984 | 0 /* flags */, |
| 985 | &op34); |
| 986 | if (status != xnn_status_success) { |
| 987 | std::cerr << "failed to create operation #34" << std::endl; |
| 988 | return ExecutionPlan(); |
| 989 | } |
| 990 | operators.emplace_back(op34, xnn_delete_operator); |
| 991 | |
| 992 | xnn_operator_t op35 = nullptr; |
| 993 | status = xnn_create_hardswish_nc_f32( |
| 994 | 240 /* channels */, |
| 995 | 240 /* input stride */, |
| 996 | 240 /* output stride */, |
| 997 | 0 /* flags */, |
| 998 | &op35); |
| 999 | if (status != xnn_status_success) { |
| 1000 | std::cerr << "failed to create operation #35" << std::endl; |
| 1001 | return ExecutionPlan(); |
| 1002 | } |
| 1003 | operators.emplace_back(op35, xnn_delete_operator); |
| 1004 | |
| 1005 | xnn_operator_t op36 = nullptr; |
| 1006 | status = xnn_create_convolution2d_nhwc_f32( |
| 1007 | 2 /* top padding */, 2 /* right padding */, |
| 1008 | 2 /* bottom padding */, 2 /* left padding */, |
| 1009 | 5 /* kernel height */, 5 /* kernel width */, |
| 1010 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1011 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1012 | 240 /* groups */, |
| 1013 | 1 /* input channels per group */, |
| 1014 | 1 /* output_channels_per_group */, |
| 1015 | 240 /* input pixel stride */, |
| 1016 | 240 /* output pixel stride */, |
| 1017 | w144, w145, |
| 1018 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1019 | 0 /* flags */, |
| 1020 | &op36); |
| 1021 | if (status != xnn_status_success) { |
| 1022 | std::cerr << "failed to create operation #36" << std::endl; |
| 1023 | return ExecutionPlan(); |
| 1024 | } |
| 1025 | operators.emplace_back(op36, xnn_delete_operator); |
| 1026 | |
| 1027 | xnn_operator_t op37 = nullptr; |
| 1028 | status = xnn_create_hardswish_nc_f32( |
| 1029 | 240 /* channels */, |
| 1030 | 240 /* input stride */, |
| 1031 | 240 /* output stride */, |
| 1032 | 0 /* flags */, |
| 1033 | &op37); |
| 1034 | if (status != xnn_status_success) { |
| 1035 | std::cerr << "failed to create operation #37" << std::endl; |
| 1036 | return ExecutionPlan(); |
| 1037 | } |
| 1038 | operators.emplace_back(op37, xnn_delete_operator); |
| 1039 | |
| 1040 | xnn_operator_t op38 = nullptr; |
| 1041 | status = xnn_create_global_average_pooling_nwc_f32( |
| 1042 | 240 /* channels */, 240 /* input stride */, 240 /* output stride */, |
| 1043 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1044 | 0 /* flags */, |
| 1045 | &op38); |
| 1046 | if (status != xnn_status_success) { |
| 1047 | std::cerr << "failed to create operation #38" << std::endl; |
| 1048 | return ExecutionPlan(); |
| 1049 | } |
| 1050 | operators.emplace_back(op38, xnn_delete_operator); |
| 1051 | |
| 1052 | xnn_operator_t op39 = nullptr; |
| 1053 | status = xnn_create_convolution2d_nhwc_f32( |
| 1054 | 0 /* top padding */, 0 /* right padding */, |
| 1055 | 0 /* bottom padding */, 0 /* left padding */, |
| 1056 | 1 /* kernel height */, 1 /* kernel width */, |
| 1057 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1058 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1059 | 1 /* groups */, |
| 1060 | 240 /* input channels per group */, |
| 1061 | 64 /* output_channels_per_group */, |
| 1062 | 240 /* input pixel stride */, |
| 1063 | 64 /* output pixel stride */, |
| 1064 | w146, w147, |
| 1065 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1066 | 0 /* flags */, |
| 1067 | &op39); |
| 1068 | if (status != xnn_status_success) { |
| 1069 | std::cerr << "failed to create operation #39" << std::endl; |
| 1070 | return ExecutionPlan(); |
| 1071 | } |
| 1072 | operators.emplace_back(op39, xnn_delete_operator); |
| 1073 | |
| 1074 | xnn_operator_t op40 = nullptr; |
| 1075 | status = xnn_create_convolution2d_nhwc_f32( |
| 1076 | 0 /* top padding */, 0 /* right padding */, |
| 1077 | 0 /* bottom padding */, 0 /* left padding */, |
| 1078 | 1 /* kernel height */, 1 /* kernel width */, |
| 1079 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1080 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1081 | 1 /* groups */, |
| 1082 | 64 /* input channels per group */, |
| 1083 | 240 /* output_channels_per_group */, |
| 1084 | 64 /* input pixel stride */, |
| 1085 | 240 /* output pixel stride */, |
| 1086 | w148, w149, |
| 1087 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1088 | 0 /* flags */, |
| 1089 | &op40); |
| 1090 | if (status != xnn_status_success) { |
| 1091 | std::cerr << "failed to create operation #40" << std::endl; |
| 1092 | return ExecutionPlan(); |
| 1093 | } |
| 1094 | operators.emplace_back(op40, xnn_delete_operator); |
| 1095 | |
| 1096 | xnn_operator_t op41 = nullptr; |
| 1097 | status = xnn_create_multiply_nd_f32( |
| 1098 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1099 | 0 /* flags */, |
| 1100 | &op41); |
| 1101 | if (status != xnn_status_success) { |
| 1102 | std::cerr << "failed to create operation #41" << std::endl; |
| 1103 | return ExecutionPlan(); |
| 1104 | } |
| 1105 | operators.emplace_back(op41, xnn_delete_operator); |
| 1106 | |
| 1107 | xnn_operator_t op42 = nullptr; |
| 1108 | status = xnn_create_convolution2d_nhwc_f32( |
| 1109 | 0 /* top padding */, 0 /* right padding */, |
| 1110 | 0 /* bottom padding */, 0 /* left padding */, |
| 1111 | 1 /* kernel height */, 1 /* kernel width */, |
| 1112 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1113 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1114 | 1 /* groups */, |
| 1115 | 240 /* input channels per group */, |
| 1116 | 40 /* output_channels_per_group */, |
| 1117 | 240 /* input pixel stride */, |
| 1118 | 40 /* output pixel stride */, |
| 1119 | w150, w151, |
| 1120 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1121 | 0 /* flags */, |
| 1122 | &op42); |
| 1123 | if (status != xnn_status_success) { |
| 1124 | std::cerr << "failed to create operation #42" << std::endl; |
| 1125 | return ExecutionPlan(); |
| 1126 | } |
| 1127 | operators.emplace_back(op42, xnn_delete_operator); |
| 1128 | |
| 1129 | xnn_operator_t op43 = nullptr; |
| 1130 | status = xnn_create_add_nc_f32( |
| 1131 | 40 /* channels */, |
| 1132 | 40 /* a stride */, |
| 1133 | 40 /* b stride */, |
| 1134 | 40 /* c stride */, |
| 1135 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1136 | 0 /* flags */, |
| 1137 | &op43); |
| 1138 | if (status != xnn_status_success) { |
| 1139 | std::cerr << "failed to create operation #43" << std::endl; |
| 1140 | return ExecutionPlan(); |
| 1141 | } |
| 1142 | operators.emplace_back(op43, xnn_delete_operator); |
| 1143 | |
| 1144 | xnn_operator_t op44 = nullptr; |
| 1145 | status = xnn_create_convolution2d_nhwc_f32( |
| 1146 | 0 /* top padding */, 0 /* right padding */, |
| 1147 | 0 /* bottom padding */, 0 /* left padding */, |
| 1148 | 1 /* kernel height */, 1 /* kernel width */, |
| 1149 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1150 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1151 | 1 /* groups */, |
| 1152 | 40 /* input channels per group */, |
| 1153 | 120 /* output_channels_per_group */, |
| 1154 | 40 /* input pixel stride */, |
| 1155 | 120 /* output pixel stride */, |
| 1156 | w152, w153, |
| 1157 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1158 | 0 /* flags */, |
| 1159 | &op44); |
| 1160 | if (status != xnn_status_success) { |
| 1161 | std::cerr << "failed to create operation #44" << std::endl; |
| 1162 | return ExecutionPlan(); |
| 1163 | } |
| 1164 | operators.emplace_back(op44, xnn_delete_operator); |
| 1165 | |
| 1166 | xnn_operator_t op45 = nullptr; |
| 1167 | status = xnn_create_hardswish_nc_f32( |
| 1168 | 120 /* channels */, |
| 1169 | 120 /* input stride */, |
| 1170 | 120 /* output stride */, |
| 1171 | 0 /* flags */, |
| 1172 | &op45); |
| 1173 | if (status != xnn_status_success) { |
| 1174 | std::cerr << "failed to create operation #45" << std::endl; |
| 1175 | return ExecutionPlan(); |
| 1176 | } |
| 1177 | operators.emplace_back(op45, xnn_delete_operator); |
| 1178 | |
| 1179 | xnn_operator_t op46 = nullptr; |
| 1180 | status = xnn_create_convolution2d_nhwc_f32( |
| 1181 | 2 /* top padding */, 2 /* right padding */, |
| 1182 | 2 /* bottom padding */, 2 /* left padding */, |
| 1183 | 5 /* kernel height */, 5 /* kernel width */, |
| 1184 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1185 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1186 | 120 /* groups */, |
| 1187 | 1 /* input channels per group */, |
| 1188 | 1 /* output_channels_per_group */, |
| 1189 | 120 /* input pixel stride */, |
| 1190 | 120 /* output pixel stride */, |
| 1191 | w154, w155, |
| 1192 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1193 | 0 /* flags */, |
| 1194 | &op46); |
| 1195 | if (status != xnn_status_success) { |
| 1196 | std::cerr << "failed to create operation #46" << std::endl; |
| 1197 | return ExecutionPlan(); |
| 1198 | } |
| 1199 | operators.emplace_back(op46, xnn_delete_operator); |
| 1200 | |
| 1201 | xnn_operator_t op47 = nullptr; |
| 1202 | status = xnn_create_hardswish_nc_f32( |
| 1203 | 120 /* channels */, |
| 1204 | 120 /* input stride */, |
| 1205 | 120 /* output stride */, |
| 1206 | 0 /* flags */, |
| 1207 | &op47); |
| 1208 | if (status != xnn_status_success) { |
| 1209 | std::cerr << "failed to create operation #47" << std::endl; |
| 1210 | return ExecutionPlan(); |
| 1211 | } |
| 1212 | operators.emplace_back(op47, xnn_delete_operator); |
| 1213 | |
| 1214 | xnn_operator_t op48 = nullptr; |
| 1215 | status = xnn_create_global_average_pooling_nwc_f32( |
| 1216 | 120 /* channels */, 120 /* input stride */, 120 /* output stride */, |
| 1217 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1218 | 0 /* flags */, |
| 1219 | &op48); |
| 1220 | if (status != xnn_status_success) { |
| 1221 | std::cerr << "failed to create operation #48" << std::endl; |
| 1222 | return ExecutionPlan(); |
| 1223 | } |
| 1224 | operators.emplace_back(op48, xnn_delete_operator); |
| 1225 | |
| 1226 | xnn_operator_t op49 = nullptr; |
| 1227 | status = xnn_create_convolution2d_nhwc_f32( |
| 1228 | 0 /* top padding */, 0 /* right padding */, |
| 1229 | 0 /* bottom padding */, 0 /* left padding */, |
| 1230 | 1 /* kernel height */, 1 /* kernel width */, |
| 1231 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1232 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1233 | 1 /* groups */, |
| 1234 | 120 /* input channels per group */, |
| 1235 | 32 /* output_channels_per_group */, |
| 1236 | 120 /* input pixel stride */, |
| 1237 | 32 /* output pixel stride */, |
| 1238 | w156, w157, |
| 1239 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1240 | 0 /* flags */, |
| 1241 | &op49); |
| 1242 | if (status != xnn_status_success) { |
| 1243 | std::cerr << "failed to create operation #49" << std::endl; |
| 1244 | return ExecutionPlan(); |
| 1245 | } |
| 1246 | operators.emplace_back(op49, xnn_delete_operator); |
| 1247 | |
| 1248 | xnn_operator_t op50 = nullptr; |
| 1249 | status = xnn_create_convolution2d_nhwc_f32( |
| 1250 | 0 /* top padding */, 0 /* right padding */, |
| 1251 | 0 /* bottom padding */, 0 /* left padding */, |
| 1252 | 1 /* kernel height */, 1 /* kernel width */, |
| 1253 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1254 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1255 | 1 /* groups */, |
| 1256 | 32 /* input channels per group */, |
| 1257 | 120 /* output_channels_per_group */, |
| 1258 | 32 /* input pixel stride */, |
| 1259 | 120 /* output pixel stride */, |
| 1260 | w158, w159, |
| 1261 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1262 | 0 /* flags */, |
| 1263 | &op50); |
| 1264 | if (status != xnn_status_success) { |
| 1265 | std::cerr << "failed to create operation #50" << std::endl; |
| 1266 | return ExecutionPlan(); |
| 1267 | } |
| 1268 | operators.emplace_back(op50, xnn_delete_operator); |
| 1269 | |
| 1270 | xnn_operator_t op51 = nullptr; |
| 1271 | status = xnn_create_multiply_nd_f32( |
| 1272 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1273 | 0 /* flags */, |
| 1274 | &op51); |
| 1275 | if (status != xnn_status_success) { |
| 1276 | std::cerr << "failed to create operation #51" << std::endl; |
| 1277 | return ExecutionPlan(); |
| 1278 | } |
| 1279 | operators.emplace_back(op51, xnn_delete_operator); |
| 1280 | |
| 1281 | xnn_operator_t op52 = nullptr; |
| 1282 | status = xnn_create_convolution2d_nhwc_f32( |
| 1283 | 0 /* top padding */, 0 /* right padding */, |
| 1284 | 0 /* bottom padding */, 0 /* left padding */, |
| 1285 | 1 /* kernel height */, 1 /* kernel width */, |
| 1286 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1287 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1288 | 1 /* groups */, |
| 1289 | 120 /* input channels per group */, |
| 1290 | 48 /* output_channels_per_group */, |
| 1291 | 120 /* input pixel stride */, |
| 1292 | 48 /* output pixel stride */, |
| 1293 | w160, w161, |
| 1294 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1295 | 0 /* flags */, |
| 1296 | &op52); |
| 1297 | if (status != xnn_status_success) { |
| 1298 | std::cerr << "failed to create operation #52" << std::endl; |
| 1299 | return ExecutionPlan(); |
| 1300 | } |
| 1301 | operators.emplace_back(op52, xnn_delete_operator); |
| 1302 | |
| 1303 | xnn_operator_t op53 = nullptr; |
| 1304 | status = xnn_create_convolution2d_nhwc_f32( |
| 1305 | 0 /* top padding */, 0 /* right padding */, |
| 1306 | 0 /* bottom padding */, 0 /* left padding */, |
| 1307 | 1 /* kernel height */, 1 /* kernel width */, |
| 1308 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1309 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1310 | 1 /* groups */, |
| 1311 | 48 /* input channels per group */, |
| 1312 | 144 /* output_channels_per_group */, |
| 1313 | 48 /* input pixel stride */, |
| 1314 | 144 /* output pixel stride */, |
| 1315 | w162, w163, |
| 1316 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1317 | 0 /* flags */, |
| 1318 | &op53); |
| 1319 | if (status != xnn_status_success) { |
| 1320 | std::cerr << "failed to create operation #53" << std::endl; |
| 1321 | return ExecutionPlan(); |
| 1322 | } |
| 1323 | operators.emplace_back(op53, xnn_delete_operator); |
| 1324 | |
| 1325 | xnn_operator_t op54 = nullptr; |
| 1326 | status = xnn_create_hardswish_nc_f32( |
| 1327 | 144 /* channels */, |
| 1328 | 144 /* input stride */, |
| 1329 | 144 /* output stride */, |
| 1330 | 0 /* flags */, |
| 1331 | &op54); |
| 1332 | if (status != xnn_status_success) { |
| 1333 | std::cerr << "failed to create operation #54" << std::endl; |
| 1334 | return ExecutionPlan(); |
| 1335 | } |
| 1336 | operators.emplace_back(op54, xnn_delete_operator); |
| 1337 | |
| 1338 | xnn_operator_t op55 = nullptr; |
| 1339 | status = xnn_create_convolution2d_nhwc_f32( |
| 1340 | 2 /* top padding */, 2 /* right padding */, |
| 1341 | 2 /* bottom padding */, 2 /* left padding */, |
| 1342 | 5 /* kernel height */, 5 /* kernel width */, |
| 1343 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1344 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1345 | 144 /* groups */, |
| 1346 | 1 /* input channels per group */, |
| 1347 | 1 /* output_channels_per_group */, |
| 1348 | 144 /* input pixel stride */, |
| 1349 | 144 /* output pixel stride */, |
| 1350 | w164, w165, |
| 1351 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1352 | 0 /* flags */, |
| 1353 | &op55); |
| 1354 | if (status != xnn_status_success) { |
| 1355 | std::cerr << "failed to create operation #55" << std::endl; |
| 1356 | return ExecutionPlan(); |
| 1357 | } |
| 1358 | operators.emplace_back(op55, xnn_delete_operator); |
| 1359 | |
| 1360 | xnn_operator_t op56 = nullptr; |
| 1361 | status = xnn_create_hardswish_nc_f32( |
| 1362 | 144 /* channels */, |
| 1363 | 144 /* input stride */, |
| 1364 | 144 /* output stride */, |
| 1365 | 0 /* flags */, |
| 1366 | &op56); |
| 1367 | if (status != xnn_status_success) { |
| 1368 | std::cerr << "failed to create operation #56" << std::endl; |
| 1369 | return ExecutionPlan(); |
| 1370 | } |
| 1371 | operators.emplace_back(op56, xnn_delete_operator); |
| 1372 | |
| 1373 | xnn_operator_t op57 = nullptr; |
| 1374 | status = xnn_create_global_average_pooling_nwc_f32( |
| 1375 | 144 /* channels */, 144 /* input stride */, 144 /* output stride */, |
| 1376 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1377 | 0 /* flags */, |
| 1378 | &op57); |
| 1379 | if (status != xnn_status_success) { |
| 1380 | std::cerr << "failed to create operation #57" << std::endl; |
| 1381 | return ExecutionPlan(); |
| 1382 | } |
| 1383 | operators.emplace_back(op57, xnn_delete_operator); |
| 1384 | |
| 1385 | xnn_operator_t op58 = nullptr; |
| 1386 | status = xnn_create_convolution2d_nhwc_f32( |
| 1387 | 0 /* top padding */, 0 /* right padding */, |
| 1388 | 0 /* bottom padding */, 0 /* left padding */, |
| 1389 | 1 /* kernel height */, 1 /* kernel width */, |
| 1390 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1391 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1392 | 1 /* groups */, |
| 1393 | 144 /* input channels per group */, |
| 1394 | 40 /* output_channels_per_group */, |
| 1395 | 144 /* input pixel stride */, |
| 1396 | 40 /* output pixel stride */, |
| 1397 | w166, w167, |
| 1398 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1399 | 0 /* flags */, |
| 1400 | &op58); |
| 1401 | if (status != xnn_status_success) { |
| 1402 | std::cerr << "failed to create operation #58" << std::endl; |
| 1403 | return ExecutionPlan(); |
| 1404 | } |
| 1405 | operators.emplace_back(op58, xnn_delete_operator); |
| 1406 | |
| 1407 | xnn_operator_t op59 = nullptr; |
| 1408 | status = xnn_create_convolution2d_nhwc_f32( |
| 1409 | 0 /* top padding */, 0 /* right padding */, |
| 1410 | 0 /* bottom padding */, 0 /* left padding */, |
| 1411 | 1 /* kernel height */, 1 /* kernel width */, |
| 1412 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1413 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1414 | 1 /* groups */, |
| 1415 | 40 /* input channels per group */, |
| 1416 | 144 /* output_channels_per_group */, |
| 1417 | 40 /* input pixel stride */, |
| 1418 | 144 /* output pixel stride */, |
| 1419 | w168, w169, |
| 1420 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1421 | 0 /* flags */, |
| 1422 | &op59); |
| 1423 | if (status != xnn_status_success) { |
| 1424 | std::cerr << "failed to create operation #59" << std::endl; |
| 1425 | return ExecutionPlan(); |
| 1426 | } |
| 1427 | operators.emplace_back(op59, xnn_delete_operator); |
| 1428 | |
| 1429 | xnn_operator_t op60 = nullptr; |
| 1430 | status = xnn_create_multiply_nd_f32( |
| 1431 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1432 | 0 /* flags */, |
| 1433 | &op60); |
| 1434 | if (status != xnn_status_success) { |
| 1435 | std::cerr << "failed to create operation #60" << std::endl; |
| 1436 | return ExecutionPlan(); |
| 1437 | } |
| 1438 | operators.emplace_back(op60, xnn_delete_operator); |
| 1439 | |
| 1440 | xnn_operator_t op61 = nullptr; |
| 1441 | status = xnn_create_convolution2d_nhwc_f32( |
| 1442 | 0 /* top padding */, 0 /* right padding */, |
| 1443 | 0 /* bottom padding */, 0 /* left padding */, |
| 1444 | 1 /* kernel height */, 1 /* kernel width */, |
| 1445 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1446 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1447 | 1 /* groups */, |
| 1448 | 144 /* input channels per group */, |
| 1449 | 48 /* output_channels_per_group */, |
| 1450 | 144 /* input pixel stride */, |
| 1451 | 48 /* output pixel stride */, |
| 1452 | w170, w171, |
| 1453 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1454 | 0 /* flags */, |
| 1455 | &op61); |
| 1456 | if (status != xnn_status_success) { |
| 1457 | std::cerr << "failed to create operation #61" << std::endl; |
| 1458 | return ExecutionPlan(); |
| 1459 | } |
| 1460 | operators.emplace_back(op61, xnn_delete_operator); |
| 1461 | |
| 1462 | xnn_operator_t op62 = nullptr; |
| 1463 | status = xnn_create_add_nc_f32( |
| 1464 | 48 /* channels */, |
| 1465 | 48 /* a stride */, |
| 1466 | 48 /* b stride */, |
| 1467 | 48 /* c stride */, |
| 1468 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1469 | 0 /* flags */, |
| 1470 | &op62); |
| 1471 | if (status != xnn_status_success) { |
| 1472 | std::cerr << "failed to create operation #62" << std::endl; |
| 1473 | return ExecutionPlan(); |
| 1474 | } |
| 1475 | operators.emplace_back(op62, xnn_delete_operator); |
| 1476 | |
| 1477 | xnn_operator_t op63 = nullptr; |
| 1478 | status = xnn_create_convolution2d_nhwc_f32( |
| 1479 | 0 /* top padding */, 0 /* right padding */, |
| 1480 | 0 /* bottom padding */, 0 /* left padding */, |
| 1481 | 1 /* kernel height */, 1 /* kernel width */, |
| 1482 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1483 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1484 | 1 /* groups */, |
| 1485 | 48 /* input channels per group */, |
| 1486 | 288 /* output_channels_per_group */, |
| 1487 | 48 /* input pixel stride */, |
| 1488 | 288 /* output pixel stride */, |
| 1489 | w172, w173, |
| 1490 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1491 | 0 /* flags */, |
| 1492 | &op63); |
| 1493 | if (status != xnn_status_success) { |
| 1494 | std::cerr << "failed to create operation #63" << std::endl; |
| 1495 | return ExecutionPlan(); |
| 1496 | } |
| 1497 | operators.emplace_back(op63, xnn_delete_operator); |
| 1498 | |
| 1499 | xnn_operator_t op64 = nullptr; |
| 1500 | status = xnn_create_hardswish_nc_f32( |
| 1501 | 288 /* channels */, |
| 1502 | 288 /* input stride */, |
| 1503 | 288 /* output stride */, |
| 1504 | 0 /* flags */, |
| 1505 | &op64); |
| 1506 | if (status != xnn_status_success) { |
| 1507 | std::cerr << "failed to create operation #64" << std::endl; |
| 1508 | return ExecutionPlan(); |
| 1509 | } |
| 1510 | operators.emplace_back(op64, xnn_delete_operator); |
| 1511 | |
| 1512 | xnn_operator_t op65 = nullptr; |
| 1513 | status = xnn_create_convolution2d_nhwc_f32( |
| 1514 | 1 /* top padding */, 2 /* right padding */, |
| 1515 | 2 /* bottom padding */, 1 /* left padding */, |
| 1516 | 5 /* kernel height */, 5 /* kernel width */, |
| 1517 | 2 /* subsampling height */, 2 /* subsampling width */, |
| 1518 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1519 | 288 /* groups */, |
| 1520 | 1 /* input channels per group */, |
| 1521 | 1 /* output_channels_per_group */, |
| 1522 | 288 /* input pixel stride */, |
| 1523 | 288 /* output pixel stride */, |
| 1524 | w174, w175, |
| 1525 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1526 | 0 /* flags */, |
| 1527 | &op65); |
| 1528 | if (status != xnn_status_success) { |
| 1529 | std::cerr << "failed to create operation #65" << std::endl; |
| 1530 | return ExecutionPlan(); |
| 1531 | } |
| 1532 | operators.emplace_back(op65, xnn_delete_operator); |
| 1533 | |
| 1534 | xnn_operator_t op66 = nullptr; |
| 1535 | status = xnn_create_hardswish_nc_f32( |
| 1536 | 288 /* channels */, |
| 1537 | 288 /* input stride */, |
| 1538 | 288 /* output stride */, |
| 1539 | 0 /* flags */, |
| 1540 | &op66); |
| 1541 | if (status != xnn_status_success) { |
| 1542 | std::cerr << "failed to create operation #66" << std::endl; |
| 1543 | return ExecutionPlan(); |
| 1544 | } |
| 1545 | operators.emplace_back(op66, xnn_delete_operator); |
| 1546 | |
| 1547 | xnn_operator_t op67 = nullptr; |
| 1548 | status = xnn_create_global_average_pooling_nwc_f32( |
| 1549 | 288 /* channels */, 288 /* input stride */, 288 /* output stride */, |
| 1550 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1551 | 0 /* flags */, |
| 1552 | &op67); |
| 1553 | if (status != xnn_status_success) { |
| 1554 | std::cerr << "failed to create operation #67" << std::endl; |
| 1555 | return ExecutionPlan(); |
| 1556 | } |
| 1557 | operators.emplace_back(op67, xnn_delete_operator); |
| 1558 | |
| 1559 | xnn_operator_t op68 = nullptr; |
| 1560 | status = xnn_create_convolution2d_nhwc_f32( |
| 1561 | 0 /* top padding */, 0 /* right padding */, |
| 1562 | 0 /* bottom padding */, 0 /* left padding */, |
| 1563 | 1 /* kernel height */, 1 /* kernel width */, |
| 1564 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1565 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1566 | 1 /* groups */, |
| 1567 | 288 /* input channels per group */, |
| 1568 | 72 /* output_channels_per_group */, |
| 1569 | 288 /* input pixel stride */, |
| 1570 | 72 /* output pixel stride */, |
| 1571 | w176, w177, |
| 1572 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1573 | 0 /* flags */, |
| 1574 | &op68); |
| 1575 | if (status != xnn_status_success) { |
| 1576 | std::cerr << "failed to create operation #68" << std::endl; |
| 1577 | return ExecutionPlan(); |
| 1578 | } |
| 1579 | operators.emplace_back(op68, xnn_delete_operator); |
| 1580 | |
| 1581 | xnn_operator_t op69 = nullptr; |
| 1582 | status = xnn_create_convolution2d_nhwc_f32( |
| 1583 | 0 /* top padding */, 0 /* right padding */, |
| 1584 | 0 /* bottom padding */, 0 /* left padding */, |
| 1585 | 1 /* kernel height */, 1 /* kernel width */, |
| 1586 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1587 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1588 | 1 /* groups */, |
| 1589 | 72 /* input channels per group */, |
| 1590 | 288 /* output_channels_per_group */, |
| 1591 | 72 /* input pixel stride */, |
| 1592 | 288 /* output pixel stride */, |
| 1593 | w178, w179, |
| 1594 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1595 | 0 /* flags */, |
| 1596 | &op69); |
| 1597 | if (status != xnn_status_success) { |
| 1598 | std::cerr << "failed to create operation #69" << std::endl; |
| 1599 | return ExecutionPlan(); |
| 1600 | } |
| 1601 | operators.emplace_back(op69, xnn_delete_operator); |
| 1602 | |
| 1603 | xnn_operator_t op70 = nullptr; |
| 1604 | status = xnn_create_multiply_nd_f32( |
| 1605 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1606 | 0 /* flags */, |
| 1607 | &op70); |
| 1608 | if (status != xnn_status_success) { |
| 1609 | std::cerr << "failed to create operation #70" << std::endl; |
| 1610 | return ExecutionPlan(); |
| 1611 | } |
| 1612 | operators.emplace_back(op70, xnn_delete_operator); |
| 1613 | |
| 1614 | xnn_operator_t op71 = nullptr; |
| 1615 | status = xnn_create_convolution2d_nhwc_f32( |
| 1616 | 0 /* top padding */, 0 /* right padding */, |
| 1617 | 0 /* bottom padding */, 0 /* left padding */, |
| 1618 | 1 /* kernel height */, 1 /* kernel width */, |
| 1619 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1620 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1621 | 1 /* groups */, |
| 1622 | 288 /* input channels per group */, |
| 1623 | 96 /* output_channels_per_group */, |
| 1624 | 288 /* input pixel stride */, |
| 1625 | 96 /* output pixel stride */, |
| 1626 | w180, w181, |
| 1627 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1628 | 0 /* flags */, |
| 1629 | &op71); |
| 1630 | if (status != xnn_status_success) { |
| 1631 | std::cerr << "failed to create operation #71" << std::endl; |
| 1632 | return ExecutionPlan(); |
| 1633 | } |
| 1634 | operators.emplace_back(op71, xnn_delete_operator); |
| 1635 | |
| 1636 | xnn_operator_t op72 = nullptr; |
| 1637 | status = xnn_create_convolution2d_nhwc_f32( |
| 1638 | 0 /* top padding */, 0 /* right padding */, |
| 1639 | 0 /* bottom padding */, 0 /* left padding */, |
| 1640 | 1 /* kernel height */, 1 /* kernel width */, |
| 1641 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1642 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1643 | 1 /* groups */, |
| 1644 | 96 /* input channels per group */, |
| 1645 | 576 /* output_channels_per_group */, |
| 1646 | 96 /* input pixel stride */, |
| 1647 | 576 /* output pixel stride */, |
| 1648 | w182, w183, |
| 1649 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1650 | 0 /* flags */, |
| 1651 | &op72); |
| 1652 | if (status != xnn_status_success) { |
| 1653 | std::cerr << "failed to create operation #72" << std::endl; |
| 1654 | return ExecutionPlan(); |
| 1655 | } |
| 1656 | operators.emplace_back(op72, xnn_delete_operator); |
| 1657 | |
| 1658 | xnn_operator_t op73 = nullptr; |
| 1659 | status = xnn_create_hardswish_nc_f32( |
| 1660 | 576 /* channels */, |
| 1661 | 576 /* input stride */, |
| 1662 | 576 /* output stride */, |
| 1663 | 0 /* flags */, |
| 1664 | &op73); |
| 1665 | if (status != xnn_status_success) { |
| 1666 | std::cerr << "failed to create operation #73" << std::endl; |
| 1667 | return ExecutionPlan(); |
| 1668 | } |
| 1669 | operators.emplace_back(op73, xnn_delete_operator); |
| 1670 | |
| 1671 | xnn_operator_t op74 = nullptr; |
| 1672 | status = xnn_create_convolution2d_nhwc_f32( |
| 1673 | 2 /* top padding */, 2 /* right padding */, |
| 1674 | 2 /* bottom padding */, 2 /* left padding */, |
| 1675 | 5 /* kernel height */, 5 /* kernel width */, |
| 1676 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1677 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1678 | 576 /* groups */, |
| 1679 | 1 /* input channels per group */, |
| 1680 | 1 /* output_channels_per_group */, |
| 1681 | 576 /* input pixel stride */, |
| 1682 | 576 /* output pixel stride */, |
| 1683 | w184, w185, |
| 1684 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1685 | 0 /* flags */, |
| 1686 | &op74); |
| 1687 | if (status != xnn_status_success) { |
| 1688 | std::cerr << "failed to create operation #74" << std::endl; |
| 1689 | return ExecutionPlan(); |
| 1690 | } |
| 1691 | operators.emplace_back(op74, xnn_delete_operator); |
| 1692 | |
| 1693 | xnn_operator_t op75 = nullptr; |
| 1694 | status = xnn_create_hardswish_nc_f32( |
| 1695 | 576 /* channels */, |
| 1696 | 576 /* input stride */, |
| 1697 | 576 /* output stride */, |
| 1698 | 0 /* flags */, |
| 1699 | &op75); |
| 1700 | if (status != xnn_status_success) { |
| 1701 | std::cerr << "failed to create operation #75" << std::endl; |
| 1702 | return ExecutionPlan(); |
| 1703 | } |
| 1704 | operators.emplace_back(op75, xnn_delete_operator); |
| 1705 | |
| 1706 | xnn_operator_t op76 = nullptr; |
| 1707 | status = xnn_create_global_average_pooling_nwc_f32( |
| 1708 | 576 /* channels */, 576 /* input stride */, 576 /* output stride */, |
| 1709 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1710 | 0 /* flags */, |
| 1711 | &op76); |
| 1712 | if (status != xnn_status_success) { |
| 1713 | std::cerr << "failed to create operation #76" << std::endl; |
| 1714 | return ExecutionPlan(); |
| 1715 | } |
| 1716 | operators.emplace_back(op76, xnn_delete_operator); |
| 1717 | |
| 1718 | xnn_operator_t op77 = nullptr; |
| 1719 | status = xnn_create_convolution2d_nhwc_f32( |
| 1720 | 0 /* top padding */, 0 /* right padding */, |
| 1721 | 0 /* bottom padding */, 0 /* left padding */, |
| 1722 | 1 /* kernel height */, 1 /* kernel width */, |
| 1723 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1724 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1725 | 1 /* groups */, |
| 1726 | 576 /* input channels per group */, |
| 1727 | 144 /* output_channels_per_group */, |
| 1728 | 576 /* input pixel stride */, |
| 1729 | 144 /* output pixel stride */, |
| 1730 | w186, w187, |
| 1731 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1732 | 0 /* flags */, |
| 1733 | &op77); |
| 1734 | if (status != xnn_status_success) { |
| 1735 | std::cerr << "failed to create operation #77" << std::endl; |
| 1736 | return ExecutionPlan(); |
| 1737 | } |
| 1738 | operators.emplace_back(op77, xnn_delete_operator); |
| 1739 | |
| 1740 | xnn_operator_t op78 = nullptr; |
| 1741 | status = xnn_create_convolution2d_nhwc_f32( |
| 1742 | 0 /* top padding */, 0 /* right padding */, |
| 1743 | 0 /* bottom padding */, 0 /* left padding */, |
| 1744 | 1 /* kernel height */, 1 /* kernel width */, |
| 1745 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1746 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1747 | 1 /* groups */, |
| 1748 | 144 /* input channels per group */, |
| 1749 | 576 /* output_channels_per_group */, |
| 1750 | 144 /* input pixel stride */, |
| 1751 | 576 /* output pixel stride */, |
| 1752 | w188, w189, |
| 1753 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1754 | 0 /* flags */, |
| 1755 | &op78); |
| 1756 | if (status != xnn_status_success) { |
| 1757 | std::cerr << "failed to create operation #78" << std::endl; |
| 1758 | return ExecutionPlan(); |
| 1759 | } |
| 1760 | operators.emplace_back(op78, xnn_delete_operator); |
| 1761 | |
| 1762 | xnn_operator_t op79 = nullptr; |
| 1763 | status = xnn_create_multiply_nd_f32( |
| 1764 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1765 | 0 /* flags */, |
| 1766 | &op79); |
| 1767 | if (status != xnn_status_success) { |
| 1768 | std::cerr << "failed to create operation #79" << std::endl; |
| 1769 | return ExecutionPlan(); |
| 1770 | } |
| 1771 | operators.emplace_back(op79, xnn_delete_operator); |
| 1772 | |
| 1773 | xnn_operator_t op80 = nullptr; |
| 1774 | status = xnn_create_convolution2d_nhwc_f32( |
| 1775 | 0 /* top padding */, 0 /* right padding */, |
| 1776 | 0 /* bottom padding */, 0 /* left padding */, |
| 1777 | 1 /* kernel height */, 1 /* kernel width */, |
| 1778 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1779 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1780 | 1 /* groups */, |
| 1781 | 576 /* input channels per group */, |
| 1782 | 96 /* output_channels_per_group */, |
| 1783 | 576 /* input pixel stride */, |
| 1784 | 96 /* output pixel stride */, |
| 1785 | w190, w191, |
| 1786 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1787 | 0 /* flags */, |
| 1788 | &op80); |
| 1789 | if (status != xnn_status_success) { |
| 1790 | std::cerr << "failed to create operation #80" << std::endl; |
| 1791 | return ExecutionPlan(); |
| 1792 | } |
| 1793 | operators.emplace_back(op80, xnn_delete_operator); |
| 1794 | |
| 1795 | xnn_operator_t op81 = nullptr; |
| 1796 | status = xnn_create_add_nc_f32( |
| 1797 | 96 /* channels */, |
| 1798 | 96 /* a stride */, |
| 1799 | 96 /* b stride */, |
| 1800 | 96 /* c stride */, |
| 1801 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1802 | 0 /* flags */, |
| 1803 | &op81); |
| 1804 | if (status != xnn_status_success) { |
| 1805 | std::cerr << "failed to create operation #81" << std::endl; |
| 1806 | return ExecutionPlan(); |
| 1807 | } |
| 1808 | operators.emplace_back(op81, xnn_delete_operator); |
| 1809 | |
| 1810 | xnn_operator_t op82 = nullptr; |
| 1811 | status = xnn_create_convolution2d_nhwc_f32( |
| 1812 | 0 /* top padding */, 0 /* right padding */, |
| 1813 | 0 /* bottom padding */, 0 /* left padding */, |
| 1814 | 1 /* kernel height */, 1 /* kernel width */, |
| 1815 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1816 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1817 | 1 /* groups */, |
| 1818 | 96 /* input channels per group */, |
| 1819 | 576 /* output_channels_per_group */, |
| 1820 | 96 /* input pixel stride */, |
| 1821 | 576 /* output pixel stride */, |
| 1822 | w192, w193, |
| 1823 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1824 | 0 /* flags */, |
| 1825 | &op82); |
| 1826 | if (status != xnn_status_success) { |
| 1827 | std::cerr << "failed to create operation #82" << std::endl; |
| 1828 | return ExecutionPlan(); |
| 1829 | } |
| 1830 | operators.emplace_back(op82, xnn_delete_operator); |
| 1831 | |
| 1832 | xnn_operator_t op83 = nullptr; |
| 1833 | status = xnn_create_hardswish_nc_f32( |
| 1834 | 576 /* channels */, |
| 1835 | 576 /* input stride */, |
| 1836 | 576 /* output stride */, |
| 1837 | 0 /* flags */, |
| 1838 | &op83); |
| 1839 | if (status != xnn_status_success) { |
| 1840 | std::cerr << "failed to create operation #83" << std::endl; |
| 1841 | return ExecutionPlan(); |
| 1842 | } |
| 1843 | operators.emplace_back(op83, xnn_delete_operator); |
| 1844 | |
| 1845 | xnn_operator_t op84 = nullptr; |
| 1846 | status = xnn_create_convolution2d_nhwc_f32( |
| 1847 | 2 /* top padding */, 2 /* right padding */, |
| 1848 | 2 /* bottom padding */, 2 /* left padding */, |
| 1849 | 5 /* kernel height */, 5 /* kernel width */, |
| 1850 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1851 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1852 | 576 /* groups */, |
| 1853 | 1 /* input channels per group */, |
| 1854 | 1 /* output_channels_per_group */, |
| 1855 | 576 /* input pixel stride */, |
| 1856 | 576 /* output pixel stride */, |
| 1857 | w194, w195, |
| 1858 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1859 | 0 /* flags */, |
| 1860 | &op84); |
| 1861 | if (status != xnn_status_success) { |
| 1862 | std::cerr << "failed to create operation #84" << std::endl; |
| 1863 | return ExecutionPlan(); |
| 1864 | } |
| 1865 | operators.emplace_back(op84, xnn_delete_operator); |
| 1866 | |
| 1867 | xnn_operator_t op85 = nullptr; |
| 1868 | status = xnn_create_hardswish_nc_f32( |
| 1869 | 576 /* channels */, |
| 1870 | 576 /* input stride */, |
| 1871 | 576 /* output stride */, |
| 1872 | 0 /* flags */, |
| 1873 | &op85); |
| 1874 | if (status != xnn_status_success) { |
| 1875 | std::cerr << "failed to create operation #85" << std::endl; |
| 1876 | return ExecutionPlan(); |
| 1877 | } |
| 1878 | operators.emplace_back(op85, xnn_delete_operator); |
| 1879 | |
| 1880 | xnn_operator_t op86 = nullptr; |
| 1881 | status = xnn_create_global_average_pooling_nwc_f32( |
| 1882 | 576 /* channels */, 576 /* input stride */, 576 /* output stride */, |
| 1883 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 1884 | 0 /* flags */, |
| 1885 | &op86); |
| 1886 | if (status != xnn_status_success) { |
| 1887 | std::cerr << "failed to create operation #86" << std::endl; |
| 1888 | return ExecutionPlan(); |
| 1889 | } |
| 1890 | operators.emplace_back(op86, xnn_delete_operator); |
| 1891 | |
| 1892 | xnn_operator_t op87 = nullptr; |
| 1893 | status = xnn_create_convolution2d_nhwc_f32( |
| 1894 | 0 /* top padding */, 0 /* right padding */, |
| 1895 | 0 /* bottom padding */, 0 /* left padding */, |
| 1896 | 1 /* kernel height */, 1 /* kernel width */, |
| 1897 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1898 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1899 | 1 /* groups */, |
| 1900 | 576 /* input channels per group */, |
| 1901 | 144 /* output_channels_per_group */, |
| 1902 | 576 /* input pixel stride */, |
| 1903 | 144 /* output pixel stride */, |
| 1904 | w196, w197, |
| 1905 | 0.0f /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1906 | 0 /* flags */, |
| 1907 | &op87); |
| 1908 | if (status != xnn_status_success) { |
| 1909 | std::cerr << "failed to create operation #87" << std::endl; |
| 1910 | return ExecutionPlan(); |
| 1911 | } |
| 1912 | operators.emplace_back(op87, xnn_delete_operator); |
| 1913 | |
| 1914 | xnn_operator_t op88 = nullptr; |
| 1915 | status = xnn_create_convolution2d_nhwc_f32( |
| 1916 | 0 /* top padding */, 0 /* right padding */, |
| 1917 | 0 /* bottom padding */, 0 /* left padding */, |
| 1918 | 1 /* kernel height */, 1 /* kernel width */, |
| 1919 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1920 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1921 | 1 /* groups */, |
| 1922 | 144 /* input channels per group */, |
| 1923 | 576 /* output_channels_per_group */, |
| 1924 | 144 /* input pixel stride */, |
| 1925 | 576 /* output pixel stride */, |
| 1926 | w198, w199, |
| 1927 | 0.0f /* output min */, +0x1.00014Fp+0 /* output max */, |
| 1928 | 0 /* flags */, |
| 1929 | &op88); |
| 1930 | if (status != xnn_status_success) { |
| 1931 | std::cerr << "failed to create operation #88" << std::endl; |
| 1932 | return ExecutionPlan(); |
| 1933 | } |
| 1934 | operators.emplace_back(op88, xnn_delete_operator); |
| 1935 | |
| 1936 | xnn_operator_t op89 = nullptr; |
| 1937 | status = xnn_create_multiply_nd_f32( |
| 1938 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1939 | 0 /* flags */, |
| 1940 | &op89); |
| 1941 | if (status != xnn_status_success) { |
| 1942 | std::cerr << "failed to create operation #89" << std::endl; |
| 1943 | return ExecutionPlan(); |
| 1944 | } |
| 1945 | operators.emplace_back(op89, xnn_delete_operator); |
| 1946 | |
| 1947 | xnn_operator_t op90 = nullptr; |
| 1948 | status = xnn_create_convolution2d_nhwc_f32( |
| 1949 | 0 /* top padding */, 0 /* right padding */, |
| 1950 | 0 /* bottom padding */, 0 /* left padding */, |
| 1951 | 1 /* kernel height */, 1 /* kernel width */, |
| 1952 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1953 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1954 | 1 /* groups */, |
| 1955 | 576 /* input channels per group */, |
| 1956 | 96 /* output_channels_per_group */, |
| 1957 | 576 /* input pixel stride */, |
| 1958 | 96 /* output pixel stride */, |
| 1959 | w200, w201, |
| 1960 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1961 | 0 /* flags */, |
| 1962 | &op90); |
| 1963 | if (status != xnn_status_success) { |
| 1964 | std::cerr << "failed to create operation #90" << std::endl; |
| 1965 | return ExecutionPlan(); |
| 1966 | } |
| 1967 | operators.emplace_back(op90, xnn_delete_operator); |
| 1968 | |
| 1969 | xnn_operator_t op91 = nullptr; |
| 1970 | status = xnn_create_add_nc_f32( |
| 1971 | 96 /* channels */, |
| 1972 | 96 /* a stride */, |
| 1973 | 96 /* b stride */, |
| 1974 | 96 /* c stride */, |
| 1975 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1976 | 0 /* flags */, |
| 1977 | &op91); |
| 1978 | if (status != xnn_status_success) { |
| 1979 | std::cerr << "failed to create operation #91" << std::endl; |
| 1980 | return ExecutionPlan(); |
| 1981 | } |
| 1982 | operators.emplace_back(op91, xnn_delete_operator); |
| 1983 | |
| 1984 | xnn_operator_t op92 = nullptr; |
| 1985 | status = xnn_create_convolution2d_nhwc_f32( |
| 1986 | 0 /* top padding */, 0 /* right padding */, |
| 1987 | 0 /* bottom padding */, 0 /* left padding */, |
| 1988 | 1 /* kernel height */, 1 /* kernel width */, |
| 1989 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 1990 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 1991 | 1 /* groups */, |
| 1992 | 96 /* input channels per group */, |
| 1993 | 576 /* output_channels_per_group */, |
| 1994 | 96 /* input pixel stride */, |
| 1995 | 576 /* output pixel stride */, |
| 1996 | w202, w203, |
| 1997 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 1998 | 0 /* flags */, |
| 1999 | &op92); |
| 2000 | if (status != xnn_status_success) { |
| 2001 | std::cerr << "failed to create operation #92" << std::endl; |
| 2002 | return ExecutionPlan(); |
| 2003 | } |
| 2004 | operators.emplace_back(op92, xnn_delete_operator); |
| 2005 | |
| 2006 | xnn_operator_t op93 = nullptr; |
| 2007 | status = xnn_create_hardswish_nc_f32( |
| 2008 | 576 /* channels */, |
| 2009 | 576 /* input stride */, |
| 2010 | 576 /* output stride */, |
| 2011 | 0 /* flags */, |
| 2012 | &op93); |
| 2013 | if (status != xnn_status_success) { |
| 2014 | std::cerr << "failed to create operation #93" << std::endl; |
| 2015 | return ExecutionPlan(); |
| 2016 | } |
| 2017 | operators.emplace_back(op93, xnn_delete_operator); |
| 2018 | |
| 2019 | xnn_operator_t op94 = nullptr; |
| 2020 | status = xnn_create_global_average_pooling_nwc_f32( |
| 2021 | 576 /* channels */, 576 /* input stride */, 576 /* output stride */, |
| 2022 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 2023 | 0 /* flags */, |
| 2024 | &op94); |
| 2025 | if (status != xnn_status_success) { |
| 2026 | std::cerr << "failed to create operation #94" << std::endl; |
| 2027 | return ExecutionPlan(); |
| 2028 | } |
| 2029 | operators.emplace_back(op94, xnn_delete_operator); |
| 2030 | |
| 2031 | xnn_operator_t op95 = nullptr; |
| 2032 | status = xnn_create_convolution2d_nhwc_f32( |
| 2033 | 0 /* top padding */, 0 /* right padding */, |
| 2034 | 0 /* bottom padding */, 0 /* left padding */, |
| 2035 | 1 /* kernel height */, 1 /* kernel width */, |
| 2036 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2037 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2038 | 1 /* groups */, |
| 2039 | 576 /* input channels per group */, |
| 2040 | 1024 /* output_channels_per_group */, |
| 2041 | 576 /* input pixel stride */, |
| 2042 | 1024 /* output pixel stride */, |
| 2043 | w204, w205, |
| 2044 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2045 | 0 /* flags */, |
| 2046 | &op95); |
| 2047 | if (status != xnn_status_success) { |
| 2048 | std::cerr << "failed to create operation #95" << std::endl; |
| 2049 | return ExecutionPlan(); |
| 2050 | } |
| 2051 | operators.emplace_back(op95, xnn_delete_operator); |
| 2052 | |
| 2053 | xnn_operator_t op96 = nullptr; |
| 2054 | status = xnn_create_hardswish_nc_f32( |
| 2055 | 1024 /* channels */, |
| 2056 | 1024 /* input stride */, |
| 2057 | 1024 /* output stride */, |
| 2058 | 0 /* flags */, |
| 2059 | &op96); |
| 2060 | if (status != xnn_status_success) { |
| 2061 | std::cerr << "failed to create operation #96" << std::endl; |
| 2062 | return ExecutionPlan(); |
| 2063 | } |
| 2064 | operators.emplace_back(op96, xnn_delete_operator); |
| 2065 | |
| 2066 | xnn_operator_t op97 = nullptr; |
| 2067 | status = xnn_create_global_average_pooling_nwc_f32( |
| 2068 | 1024 /* channels */, 1024 /* input stride */, 1024 /* output stride */, |
| 2069 | -std::numeric_limits<float>::infinity(), std::numeric_limits<float>::infinity(), |
| 2070 | 0 /* flags */, |
| 2071 | &op97); |
| 2072 | if (status != xnn_status_success) { |
| 2073 | std::cerr << "failed to create operation #97" << std::endl; |
| 2074 | return ExecutionPlan(); |
| 2075 | } |
| 2076 | operators.emplace_back(op97, xnn_delete_operator); |
| 2077 | |
| 2078 | xnn_operator_t op98 = nullptr; |
| 2079 | status = xnn_create_convolution2d_nhwc_f32( |
| 2080 | 0 /* top padding */, 0 /* right padding */, |
| 2081 | 0 /* bottom padding */, 0 /* left padding */, |
| 2082 | 1 /* kernel height */, 1 /* kernel width */, |
| 2083 | 1 /* subsampling height */, 1 /* subsampling width */, |
| 2084 | 1 /* dilation_height */, 1 /* dilation_width */, |
| 2085 | 1 /* groups */, |
| 2086 | 1024 /* input channels per group */, |
| 2087 | 1001 /* output_channels_per_group */, |
| 2088 | 1024 /* input pixel stride */, |
| 2089 | 1001 /* output pixel stride */, |
| 2090 | w206, w207, |
| 2091 | -std::numeric_limits<float>::infinity() /* output min */, std::numeric_limits<float>::infinity() /* output max */, |
| 2092 | 0 /* flags */, |
| 2093 | &op98); |
| 2094 | if (status != xnn_status_success) { |
| 2095 | std::cerr << "failed to create operation #98" << std::endl; |
| 2096 | return ExecutionPlan(); |
| 2097 | } |
| 2098 | operators.emplace_back(op98, xnn_delete_operator); |
| 2099 | |
| 2100 | |
| 2101 | |
| 2102 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2103 | op0, |
| 2104 | 1 /* batch size */, 224 /* input height */, 224 /* input width */, |
| 2105 | &v0[0] /* input */, &v1[0] /* output */, |
| 2106 | threadpool /* threadpool */); |
| 2107 | if (status != xnn_status_success) { |
| 2108 | std::cerr << "failed to setup operation #0" << std::endl; |
| 2109 | return ExecutionPlan(); |
| 2110 | } |
| 2111 | |
| 2112 | status = xnn_setup_hardswish_nc_f32( |
| 2113 | op1, |
| 2114 | 12544 /* batch size */, |
| 2115 | &v1[0] /* input */, &v2[0] /* output */, |
| 2116 | threadpool /* threadpool */); |
| 2117 | if (status != xnn_status_success) { |
| 2118 | std::cerr << "failed to setup operation #1" << std::endl; |
| 2119 | return ExecutionPlan(); |
| 2120 | } |
| 2121 | |
| 2122 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2123 | op2, |
| 2124 | 1 /* batch size */, 112 /* input height */, 112 /* input width */, |
| 2125 | &v2[0] /* input */, &v3[0] /* output */, |
| 2126 | threadpool /* threadpool */); |
| 2127 | if (status != xnn_status_success) { |
| 2128 | std::cerr << "failed to setup operation #2" << std::endl; |
| 2129 | return ExecutionPlan(); |
| 2130 | } |
| 2131 | |
| 2132 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 2133 | op3, |
| 2134 | 1 /* batch size */, 3136 /* width */, |
| 2135 | &v3[0] /* input */, &v4[0] /* output */, |
| 2136 | threadpool /* threadpool */); |
| 2137 | if (status != xnn_status_success) { |
| 2138 | std::cerr << "failed to setup operation #3" << std::endl; |
| 2139 | return ExecutionPlan(); |
| 2140 | } |
| 2141 | |
| 2142 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2143 | op4, |
| 2144 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2145 | &v4[0] /* input */, &v5[0] /* output */, |
| 2146 | threadpool /* threadpool */); |
| 2147 | if (status != xnn_status_success) { |
| 2148 | std::cerr << "failed to setup operation #4" << std::endl; |
| 2149 | return ExecutionPlan(); |
| 2150 | } |
| 2151 | |
| 2152 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2153 | op5, |
| 2154 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2155 | &v5[0] /* input */, &v6[0] /* output */, |
| 2156 | threadpool /* threadpool */); |
| 2157 | if (status != xnn_status_success) { |
| 2158 | std::cerr << "failed to setup operation #5" << std::endl; |
| 2159 | return ExecutionPlan(); |
| 2160 | } |
| 2161 | |
| 2162 | { |
| 2163 | const size_t a_shape[] = { 1, 56, 56, 16 }; |
| 2164 | const size_t b_shape[] = { 1, 1, 1, 16 }; |
| 2165 | status = xnn_setup_multiply_nd_f32( |
| 2166 | op6, |
| 2167 | 4, a_shape, 4, b_shape, |
| 2168 | &v3[0] /* a */, &v6[0] /* b */, &v7[0] /* output */, |
| 2169 | threadpool /* threadpool */); |
| 2170 | } |
| 2171 | if (status != xnn_status_success) { |
| 2172 | std::cerr << "failed to setup operation #6" << std::endl; |
| 2173 | return ExecutionPlan(); |
| 2174 | } |
| 2175 | |
| 2176 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2177 | op7, |
| 2178 | 1 /* batch size */, 56 /* input height */, 56 /* input width */, |
| 2179 | &v7[0] /* input */, &v8[0] /* output */, |
| 2180 | threadpool /* threadpool */); |
| 2181 | if (status != xnn_status_success) { |
| 2182 | std::cerr << "failed to setup operation #7" << std::endl; |
| 2183 | return ExecutionPlan(); |
| 2184 | } |
| 2185 | |
| 2186 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2187 | op8, |
| 2188 | 1 /* batch size */, 56 /* input height */, 56 /* input width */, |
| 2189 | &v8[0] /* input */, &v9[0] /* output */, |
| 2190 | threadpool /* threadpool */); |
| 2191 | if (status != xnn_status_success) { |
| 2192 | std::cerr << "failed to setup operation #8" << std::endl; |
| 2193 | return ExecutionPlan(); |
| 2194 | } |
| 2195 | |
| 2196 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2197 | op9, |
| 2198 | 1 /* batch size */, 56 /* input height */, 56 /* input width */, |
| 2199 | &v9[0] /* input */, &v10[0] /* output */, |
| 2200 | threadpool /* threadpool */); |
| 2201 | if (status != xnn_status_success) { |
| 2202 | std::cerr << "failed to setup operation #9" << std::endl; |
| 2203 | return ExecutionPlan(); |
| 2204 | } |
| 2205 | |
| 2206 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2207 | op10, |
| 2208 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2209 | &v10[0] /* input */, &v11[0] /* output */, |
| 2210 | threadpool /* threadpool */); |
| 2211 | if (status != xnn_status_success) { |
| 2212 | std::cerr << "failed to setup operation #10" << std::endl; |
| 2213 | return ExecutionPlan(); |
| 2214 | } |
| 2215 | |
| 2216 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2217 | op11, |
| 2218 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2219 | &v11[0] /* input */, &v12[0] /* output */, |
| 2220 | threadpool /* threadpool */); |
| 2221 | if (status != xnn_status_success) { |
| 2222 | std::cerr << "failed to setup operation #11" << std::endl; |
| 2223 | return ExecutionPlan(); |
| 2224 | } |
| 2225 | |
| 2226 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2227 | op12, |
| 2228 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2229 | &v12[0] /* input */, &v13[0] /* output */, |
| 2230 | threadpool /* threadpool */); |
| 2231 | if (status != xnn_status_success) { |
| 2232 | std::cerr << "failed to setup operation #12" << std::endl; |
| 2233 | return ExecutionPlan(); |
| 2234 | } |
| 2235 | |
| 2236 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2237 | op13, |
| 2238 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2239 | &v13[0] /* input */, &v14[0] /* output */, |
| 2240 | threadpool /* threadpool */); |
| 2241 | if (status != xnn_status_success) { |
| 2242 | std::cerr << "failed to setup operation #13" << std::endl; |
| 2243 | return ExecutionPlan(); |
| 2244 | } |
| 2245 | |
| 2246 | status = xnn_setup_add_nc_f32( |
| 2247 | op14, |
| 2248 | 784 /* batch size */, |
| 2249 | &v14[0] /* a */, &v11[0] /* b */, &v15[0] /* sum */, |
| 2250 | threadpool /* threadpool */); |
| 2251 | if (status != xnn_status_success) { |
| 2252 | std::cerr << "failed to setup operation #14" << std::endl; |
| 2253 | return ExecutionPlan(); |
| 2254 | } |
| 2255 | |
| 2256 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2257 | op15, |
| 2258 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2259 | &v15[0] /* input */, &v16[0] /* output */, |
| 2260 | threadpool /* threadpool */); |
| 2261 | if (status != xnn_status_success) { |
| 2262 | std::cerr << "failed to setup operation #15" << std::endl; |
| 2263 | return ExecutionPlan(); |
| 2264 | } |
| 2265 | |
| 2266 | status = xnn_setup_hardswish_nc_f32( |
| 2267 | op16, |
| 2268 | 784 /* batch size */, |
| 2269 | &v16[0] /* input */, &v17[0] /* output */, |
| 2270 | threadpool /* threadpool */); |
| 2271 | if (status != xnn_status_success) { |
| 2272 | std::cerr << "failed to setup operation #16" << std::endl; |
| 2273 | return ExecutionPlan(); |
| 2274 | } |
| 2275 | |
| 2276 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2277 | op17, |
| 2278 | 1 /* batch size */, 28 /* input height */, 28 /* input width */, |
| 2279 | &v17[0] /* input */, &v18[0] /* output */, |
| 2280 | threadpool /* threadpool */); |
| 2281 | if (status != xnn_status_success) { |
| 2282 | std::cerr << "failed to setup operation #17" << std::endl; |
| 2283 | return ExecutionPlan(); |
| 2284 | } |
| 2285 | |
| 2286 | status = xnn_setup_hardswish_nc_f32( |
| 2287 | op18, |
| 2288 | 196 /* batch size */, |
| 2289 | &v18[0] /* input */, &v19[0] /* output */, |
| 2290 | threadpool /* threadpool */); |
| 2291 | if (status != xnn_status_success) { |
| 2292 | std::cerr << "failed to setup operation #18" << std::endl; |
| 2293 | return ExecutionPlan(); |
| 2294 | } |
| 2295 | |
| 2296 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 2297 | op19, |
| 2298 | 1 /* batch size */, 196 /* width */, |
| 2299 | &v19[0] /* input */, &v20[0] /* output */, |
| 2300 | threadpool /* threadpool */); |
| 2301 | if (status != xnn_status_success) { |
| 2302 | std::cerr << "failed to setup operation #19" << std::endl; |
| 2303 | return ExecutionPlan(); |
| 2304 | } |
| 2305 | |
| 2306 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2307 | op20, |
| 2308 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2309 | &v20[0] /* input */, &v21[0] /* output */, |
| 2310 | threadpool /* threadpool */); |
| 2311 | if (status != xnn_status_success) { |
| 2312 | std::cerr << "failed to setup operation #20" << std::endl; |
| 2313 | return ExecutionPlan(); |
| 2314 | } |
| 2315 | |
| 2316 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2317 | op21, |
| 2318 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2319 | &v21[0] /* input */, &v22[0] /* output */, |
| 2320 | threadpool /* threadpool */); |
| 2321 | if (status != xnn_status_success) { |
| 2322 | std::cerr << "failed to setup operation #21" << std::endl; |
| 2323 | return ExecutionPlan(); |
| 2324 | } |
| 2325 | |
| 2326 | { |
| 2327 | const size_t a_shape[] = { 1, 14, 14, 96 }; |
| 2328 | const size_t b_shape[] = { 1, 1, 1, 96 }; |
| 2329 | status = xnn_setup_multiply_nd_f32( |
| 2330 | op22, |
| 2331 | 4, a_shape, 4, b_shape, |
| 2332 | &v19[0] /* a */, &v22[0] /* b */, &v23[0] /* output */, |
| 2333 | threadpool /* threadpool */); |
| 2334 | } |
| 2335 | if (status != xnn_status_success) { |
| 2336 | std::cerr << "failed to setup operation #22" << std::endl; |
| 2337 | return ExecutionPlan(); |
| 2338 | } |
| 2339 | |
| 2340 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2341 | op23, |
| 2342 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2343 | &v23[0] /* input */, &v24[0] /* output */, |
| 2344 | threadpool /* threadpool */); |
| 2345 | if (status != xnn_status_success) { |
| 2346 | std::cerr << "failed to setup operation #23" << std::endl; |
| 2347 | return ExecutionPlan(); |
| 2348 | } |
| 2349 | |
| 2350 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2351 | op24, |
| 2352 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2353 | &v24[0] /* input */, &v25[0] /* output */, |
| 2354 | threadpool /* threadpool */); |
| 2355 | if (status != xnn_status_success) { |
| 2356 | std::cerr << "failed to setup operation #24" << std::endl; |
| 2357 | return ExecutionPlan(); |
| 2358 | } |
| 2359 | |
| 2360 | status = xnn_setup_hardswish_nc_f32( |
| 2361 | op25, |
| 2362 | 196 /* batch size */, |
| 2363 | &v25[0] /* input */, &v26[0] /* output */, |
| 2364 | threadpool /* threadpool */); |
| 2365 | if (status != xnn_status_success) { |
| 2366 | std::cerr << "failed to setup operation #25" << std::endl; |
| 2367 | return ExecutionPlan(); |
| 2368 | } |
| 2369 | |
| 2370 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2371 | op26, |
| 2372 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2373 | &v26[0] /* input */, &v27[0] /* output */, |
| 2374 | threadpool /* threadpool */); |
| 2375 | if (status != xnn_status_success) { |
| 2376 | std::cerr << "failed to setup operation #26" << std::endl; |
| 2377 | return ExecutionPlan(); |
| 2378 | } |
| 2379 | |
| 2380 | status = xnn_setup_hardswish_nc_f32( |
| 2381 | op27, |
| 2382 | 196 /* batch size */, |
| 2383 | &v27[0] /* input */, &v28[0] /* output */, |
| 2384 | threadpool /* threadpool */); |
| 2385 | if (status != xnn_status_success) { |
| 2386 | std::cerr << "failed to setup operation #27" << std::endl; |
| 2387 | return ExecutionPlan(); |
| 2388 | } |
| 2389 | |
| 2390 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 2391 | op28, |
| 2392 | 1 /* batch size */, 196 /* width */, |
| 2393 | &v28[0] /* input */, &v29[0] /* output */, |
| 2394 | threadpool /* threadpool */); |
| 2395 | if (status != xnn_status_success) { |
| 2396 | std::cerr << "failed to setup operation #28" << std::endl; |
| 2397 | return ExecutionPlan(); |
| 2398 | } |
| 2399 | |
| 2400 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2401 | op29, |
| 2402 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2403 | &v29[0] /* input */, &v30[0] /* output */, |
| 2404 | threadpool /* threadpool */); |
| 2405 | if (status != xnn_status_success) { |
| 2406 | std::cerr << "failed to setup operation #29" << std::endl; |
| 2407 | return ExecutionPlan(); |
| 2408 | } |
| 2409 | |
| 2410 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2411 | op30, |
| 2412 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2413 | &v30[0] /* input */, &v31[0] /* output */, |
| 2414 | threadpool /* threadpool */); |
| 2415 | if (status != xnn_status_success) { |
| 2416 | std::cerr << "failed to setup operation #30" << std::endl; |
| 2417 | return ExecutionPlan(); |
| 2418 | } |
| 2419 | |
| 2420 | { |
| 2421 | const size_t a_shape[] = { 1, 14, 14, 240 }; |
| 2422 | const size_t b_shape[] = { 1, 1, 1, 240 }; |
| 2423 | status = xnn_setup_multiply_nd_f32( |
| 2424 | op31, |
| 2425 | 4, a_shape, 4, b_shape, |
| 2426 | &v28[0] /* a */, &v31[0] /* b */, &v32[0] /* output */, |
| 2427 | threadpool /* threadpool */); |
| 2428 | } |
| 2429 | if (status != xnn_status_success) { |
| 2430 | std::cerr << "failed to setup operation #31" << std::endl; |
| 2431 | return ExecutionPlan(); |
| 2432 | } |
| 2433 | |
| 2434 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2435 | op32, |
| 2436 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2437 | &v32[0] /* input */, &v33[0] /* output */, |
| 2438 | threadpool /* threadpool */); |
| 2439 | if (status != xnn_status_success) { |
| 2440 | std::cerr << "failed to setup operation #32" << std::endl; |
| 2441 | return ExecutionPlan(); |
| 2442 | } |
| 2443 | |
| 2444 | status = xnn_setup_add_nc_f32( |
| 2445 | op33, |
| 2446 | 196 /* batch size */, |
| 2447 | &v33[0] /* a */, &v24[0] /* b */, &v34[0] /* sum */, |
| 2448 | threadpool /* threadpool */); |
| 2449 | if (status != xnn_status_success) { |
| 2450 | std::cerr << "failed to setup operation #33" << std::endl; |
| 2451 | return ExecutionPlan(); |
| 2452 | } |
| 2453 | |
| 2454 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2455 | op34, |
| 2456 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2457 | &v34[0] /* input */, &v35[0] /* output */, |
| 2458 | threadpool /* threadpool */); |
| 2459 | if (status != xnn_status_success) { |
| 2460 | std::cerr << "failed to setup operation #34" << std::endl; |
| 2461 | return ExecutionPlan(); |
| 2462 | } |
| 2463 | |
| 2464 | status = xnn_setup_hardswish_nc_f32( |
| 2465 | op35, |
| 2466 | 196 /* batch size */, |
| 2467 | &v35[0] /* input */, &v36[0] /* output */, |
| 2468 | threadpool /* threadpool */); |
| 2469 | if (status != xnn_status_success) { |
| 2470 | std::cerr << "failed to setup operation #35" << std::endl; |
| 2471 | return ExecutionPlan(); |
| 2472 | } |
| 2473 | |
| 2474 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2475 | op36, |
| 2476 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2477 | &v36[0] /* input */, &v37[0] /* output */, |
| 2478 | threadpool /* threadpool */); |
| 2479 | if (status != xnn_status_success) { |
| 2480 | std::cerr << "failed to setup operation #36" << std::endl; |
| 2481 | return ExecutionPlan(); |
| 2482 | } |
| 2483 | |
| 2484 | status = xnn_setup_hardswish_nc_f32( |
| 2485 | op37, |
| 2486 | 196 /* batch size */, |
| 2487 | &v37[0] /* input */, &v38[0] /* output */, |
| 2488 | threadpool /* threadpool */); |
| 2489 | if (status != xnn_status_success) { |
| 2490 | std::cerr << "failed to setup operation #37" << std::endl; |
| 2491 | return ExecutionPlan(); |
| 2492 | } |
| 2493 | |
| 2494 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 2495 | op38, |
| 2496 | 1 /* batch size */, 196 /* width */, |
| 2497 | &v38[0] /* input */, &v39[0] /* output */, |
| 2498 | threadpool /* threadpool */); |
| 2499 | if (status != xnn_status_success) { |
| 2500 | std::cerr << "failed to setup operation #38" << std::endl; |
| 2501 | return ExecutionPlan(); |
| 2502 | } |
| 2503 | |
| 2504 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2505 | op39, |
| 2506 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2507 | &v39[0] /* input */, &v40[0] /* output */, |
| 2508 | threadpool /* threadpool */); |
| 2509 | if (status != xnn_status_success) { |
| 2510 | std::cerr << "failed to setup operation #39" << std::endl; |
| 2511 | return ExecutionPlan(); |
| 2512 | } |
| 2513 | |
| 2514 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2515 | op40, |
| 2516 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2517 | &v40[0] /* input */, &v41[0] /* output */, |
| 2518 | threadpool /* threadpool */); |
| 2519 | if (status != xnn_status_success) { |
| 2520 | std::cerr << "failed to setup operation #40" << std::endl; |
| 2521 | return ExecutionPlan(); |
| 2522 | } |
| 2523 | |
| 2524 | { |
| 2525 | const size_t a_shape[] = { 1, 14, 14, 240 }; |
| 2526 | const size_t b_shape[] = { 1, 1, 1, 240 }; |
| 2527 | status = xnn_setup_multiply_nd_f32( |
| 2528 | op41, |
| 2529 | 4, a_shape, 4, b_shape, |
| 2530 | &v38[0] /* a */, &v41[0] /* b */, &v42[0] /* output */, |
| 2531 | threadpool /* threadpool */); |
| 2532 | } |
| 2533 | if (status != xnn_status_success) { |
| 2534 | std::cerr << "failed to setup operation #41" << std::endl; |
| 2535 | return ExecutionPlan(); |
| 2536 | } |
| 2537 | |
| 2538 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2539 | op42, |
| 2540 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2541 | &v42[0] /* input */, &v43[0] /* output */, |
| 2542 | threadpool /* threadpool */); |
| 2543 | if (status != xnn_status_success) { |
| 2544 | std::cerr << "failed to setup operation #42" << std::endl; |
| 2545 | return ExecutionPlan(); |
| 2546 | } |
| 2547 | |
| 2548 | status = xnn_setup_add_nc_f32( |
| 2549 | op43, |
| 2550 | 196 /* batch size */, |
| 2551 | &v43[0] /* a */, &v34[0] /* b */, &v44[0] /* sum */, |
| 2552 | threadpool /* threadpool */); |
| 2553 | if (status != xnn_status_success) { |
| 2554 | std::cerr << "failed to setup operation #43" << std::endl; |
| 2555 | return ExecutionPlan(); |
| 2556 | } |
| 2557 | |
| 2558 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2559 | op44, |
| 2560 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2561 | &v44[0] /* input */, &v45[0] /* output */, |
| 2562 | threadpool /* threadpool */); |
| 2563 | if (status != xnn_status_success) { |
| 2564 | std::cerr << "failed to setup operation #44" << std::endl; |
| 2565 | return ExecutionPlan(); |
| 2566 | } |
| 2567 | |
| 2568 | status = xnn_setup_hardswish_nc_f32( |
| 2569 | op45, |
| 2570 | 196 /* batch size */, |
| 2571 | &v45[0] /* input */, &v46[0] /* output */, |
| 2572 | threadpool /* threadpool */); |
| 2573 | if (status != xnn_status_success) { |
| 2574 | std::cerr << "failed to setup operation #45" << std::endl; |
| 2575 | return ExecutionPlan(); |
| 2576 | } |
| 2577 | |
| 2578 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2579 | op46, |
| 2580 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2581 | &v46[0] /* input */, &v47[0] /* output */, |
| 2582 | threadpool /* threadpool */); |
| 2583 | if (status != xnn_status_success) { |
| 2584 | std::cerr << "failed to setup operation #46" << std::endl; |
| 2585 | return ExecutionPlan(); |
| 2586 | } |
| 2587 | |
| 2588 | status = xnn_setup_hardswish_nc_f32( |
| 2589 | op47, |
| 2590 | 196 /* batch size */, |
| 2591 | &v47[0] /* input */, &v48[0] /* output */, |
| 2592 | threadpool /* threadpool */); |
| 2593 | if (status != xnn_status_success) { |
| 2594 | std::cerr << "failed to setup operation #47" << std::endl; |
| 2595 | return ExecutionPlan(); |
| 2596 | } |
| 2597 | |
| 2598 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 2599 | op48, |
| 2600 | 1 /* batch size */, 196 /* width */, |
| 2601 | &v48[0] /* input */, &v49[0] /* output */, |
| 2602 | threadpool /* threadpool */); |
| 2603 | if (status != xnn_status_success) { |
| 2604 | std::cerr << "failed to setup operation #48" << std::endl; |
| 2605 | return ExecutionPlan(); |
| 2606 | } |
| 2607 | |
| 2608 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2609 | op49, |
| 2610 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2611 | &v49[0] /* input */, &v50[0] /* output */, |
| 2612 | threadpool /* threadpool */); |
| 2613 | if (status != xnn_status_success) { |
| 2614 | std::cerr << "failed to setup operation #49" << std::endl; |
| 2615 | return ExecutionPlan(); |
| 2616 | } |
| 2617 | |
| 2618 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2619 | op50, |
| 2620 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2621 | &v50[0] /* input */, &v51[0] /* output */, |
| 2622 | threadpool /* threadpool */); |
| 2623 | if (status != xnn_status_success) { |
| 2624 | std::cerr << "failed to setup operation #50" << std::endl; |
| 2625 | return ExecutionPlan(); |
| 2626 | } |
| 2627 | |
| 2628 | { |
| 2629 | const size_t a_shape[] = { 1, 14, 14, 120 }; |
| 2630 | const size_t b_shape[] = { 1, 1, 1, 120 }; |
| 2631 | status = xnn_setup_multiply_nd_f32( |
| 2632 | op51, |
| 2633 | 4, a_shape, 4, b_shape, |
| 2634 | &v48[0] /* a */, &v51[0] /* b */, &v52[0] /* output */, |
| 2635 | threadpool /* threadpool */); |
| 2636 | } |
| 2637 | if (status != xnn_status_success) { |
| 2638 | std::cerr << "failed to setup operation #51" << std::endl; |
| 2639 | return ExecutionPlan(); |
| 2640 | } |
| 2641 | |
| 2642 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2643 | op52, |
| 2644 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2645 | &v52[0] /* input */, &v53[0] /* output */, |
| 2646 | threadpool /* threadpool */); |
| 2647 | if (status != xnn_status_success) { |
| 2648 | std::cerr << "failed to setup operation #52" << std::endl; |
| 2649 | return ExecutionPlan(); |
| 2650 | } |
| 2651 | |
| 2652 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2653 | op53, |
| 2654 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2655 | &v53[0] /* input */, &v54[0] /* output */, |
| 2656 | threadpool /* threadpool */); |
| 2657 | if (status != xnn_status_success) { |
| 2658 | std::cerr << "failed to setup operation #53" << std::endl; |
| 2659 | return ExecutionPlan(); |
| 2660 | } |
| 2661 | |
| 2662 | status = xnn_setup_hardswish_nc_f32( |
| 2663 | op54, |
| 2664 | 196 /* batch size */, |
| 2665 | &v54[0] /* input */, &v55[0] /* output */, |
| 2666 | threadpool /* threadpool */); |
| 2667 | if (status != xnn_status_success) { |
| 2668 | std::cerr << "failed to setup operation #54" << std::endl; |
| 2669 | return ExecutionPlan(); |
| 2670 | } |
| 2671 | |
| 2672 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2673 | op55, |
| 2674 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2675 | &v55[0] /* input */, &v56[0] /* output */, |
| 2676 | threadpool /* threadpool */); |
| 2677 | if (status != xnn_status_success) { |
| 2678 | std::cerr << "failed to setup operation #55" << std::endl; |
| 2679 | return ExecutionPlan(); |
| 2680 | } |
| 2681 | |
| 2682 | status = xnn_setup_hardswish_nc_f32( |
| 2683 | op56, |
| 2684 | 196 /* batch size */, |
| 2685 | &v56[0] /* input */, &v57[0] /* output */, |
| 2686 | threadpool /* threadpool */); |
| 2687 | if (status != xnn_status_success) { |
| 2688 | std::cerr << "failed to setup operation #56" << std::endl; |
| 2689 | return ExecutionPlan(); |
| 2690 | } |
| 2691 | |
| 2692 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 2693 | op57, |
| 2694 | 1 /* batch size */, 196 /* width */, |
| 2695 | &v57[0] /* input */, &v58[0] /* output */, |
| 2696 | threadpool /* threadpool */); |
| 2697 | if (status != xnn_status_success) { |
| 2698 | std::cerr << "failed to setup operation #57" << std::endl; |
| 2699 | return ExecutionPlan(); |
| 2700 | } |
| 2701 | |
| 2702 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2703 | op58, |
| 2704 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2705 | &v58[0] /* input */, &v59[0] /* output */, |
| 2706 | threadpool /* threadpool */); |
| 2707 | if (status != xnn_status_success) { |
| 2708 | std::cerr << "failed to setup operation #58" << std::endl; |
| 2709 | return ExecutionPlan(); |
| 2710 | } |
| 2711 | |
| 2712 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2713 | op59, |
| 2714 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2715 | &v59[0] /* input */, &v60[0] /* output */, |
| 2716 | threadpool /* threadpool */); |
| 2717 | if (status != xnn_status_success) { |
| 2718 | std::cerr << "failed to setup operation #59" << std::endl; |
| 2719 | return ExecutionPlan(); |
| 2720 | } |
| 2721 | |
| 2722 | { |
| 2723 | const size_t a_shape[] = { 1, 14, 14, 144 }; |
| 2724 | const size_t b_shape[] = { 1, 1, 1, 144 }; |
| 2725 | status = xnn_setup_multiply_nd_f32( |
| 2726 | op60, |
| 2727 | 4, a_shape, 4, b_shape, |
| 2728 | &v57[0] /* a */, &v60[0] /* b */, &v61[0] /* output */, |
| 2729 | threadpool /* threadpool */); |
| 2730 | } |
| 2731 | if (status != xnn_status_success) { |
| 2732 | std::cerr << "failed to setup operation #60" << std::endl; |
| 2733 | return ExecutionPlan(); |
| 2734 | } |
| 2735 | |
| 2736 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2737 | op61, |
| 2738 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2739 | &v61[0] /* input */, &v62[0] /* output */, |
| 2740 | threadpool /* threadpool */); |
| 2741 | if (status != xnn_status_success) { |
| 2742 | std::cerr << "failed to setup operation #61" << std::endl; |
| 2743 | return ExecutionPlan(); |
| 2744 | } |
| 2745 | |
| 2746 | status = xnn_setup_add_nc_f32( |
| 2747 | op62, |
| 2748 | 196 /* batch size */, |
| 2749 | &v62[0] /* a */, &v53[0] /* b */, &v63[0] /* sum */, |
| 2750 | threadpool /* threadpool */); |
| 2751 | if (status != xnn_status_success) { |
| 2752 | std::cerr << "failed to setup operation #62" << std::endl; |
| 2753 | return ExecutionPlan(); |
| 2754 | } |
| 2755 | |
| 2756 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2757 | op63, |
| 2758 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2759 | &v63[0] /* input */, &v64[0] /* output */, |
| 2760 | threadpool /* threadpool */); |
| 2761 | if (status != xnn_status_success) { |
| 2762 | std::cerr << "failed to setup operation #63" << std::endl; |
| 2763 | return ExecutionPlan(); |
| 2764 | } |
| 2765 | |
| 2766 | status = xnn_setup_hardswish_nc_f32( |
| 2767 | op64, |
| 2768 | 196 /* batch size */, |
| 2769 | &v64[0] /* input */, &v65[0] /* output */, |
| 2770 | threadpool /* threadpool */); |
| 2771 | if (status != xnn_status_success) { |
| 2772 | std::cerr << "failed to setup operation #64" << std::endl; |
| 2773 | return ExecutionPlan(); |
| 2774 | } |
| 2775 | |
| 2776 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2777 | op65, |
| 2778 | 1 /* batch size */, 14 /* input height */, 14 /* input width */, |
| 2779 | &v65[0] /* input */, &v66[0] /* output */, |
| 2780 | threadpool /* threadpool */); |
| 2781 | if (status != xnn_status_success) { |
| 2782 | std::cerr << "failed to setup operation #65" << std::endl; |
| 2783 | return ExecutionPlan(); |
| 2784 | } |
| 2785 | |
| 2786 | status = xnn_setup_hardswish_nc_f32( |
| 2787 | op66, |
| 2788 | 49 /* batch size */, |
| 2789 | &v66[0] /* input */, &v67[0] /* output */, |
| 2790 | threadpool /* threadpool */); |
| 2791 | if (status != xnn_status_success) { |
| 2792 | std::cerr << "failed to setup operation #66" << std::endl; |
| 2793 | return ExecutionPlan(); |
| 2794 | } |
| 2795 | |
| 2796 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 2797 | op67, |
| 2798 | 1 /* batch size */, 49 /* width */, |
| 2799 | &v67[0] /* input */, &v68[0] /* output */, |
| 2800 | threadpool /* threadpool */); |
| 2801 | if (status != xnn_status_success) { |
| 2802 | std::cerr << "failed to setup operation #67" << std::endl; |
| 2803 | return ExecutionPlan(); |
| 2804 | } |
| 2805 | |
| 2806 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2807 | op68, |
| 2808 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2809 | &v68[0] /* input */, &v69[0] /* output */, |
| 2810 | threadpool /* threadpool */); |
| 2811 | if (status != xnn_status_success) { |
| 2812 | std::cerr << "failed to setup operation #68" << std::endl; |
| 2813 | return ExecutionPlan(); |
| 2814 | } |
| 2815 | |
| 2816 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2817 | op69, |
| 2818 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2819 | &v69[0] /* input */, &v70[0] /* output */, |
| 2820 | threadpool /* threadpool */); |
| 2821 | if (status != xnn_status_success) { |
| 2822 | std::cerr << "failed to setup operation #69" << std::endl; |
| 2823 | return ExecutionPlan(); |
| 2824 | } |
| 2825 | |
| 2826 | { |
| 2827 | const size_t a_shape[] = { 1, 7, 7, 288 }; |
| 2828 | const size_t b_shape[] = { 1, 1, 1, 288 }; |
| 2829 | status = xnn_setup_multiply_nd_f32( |
| 2830 | op70, |
| 2831 | 4, a_shape, 4, b_shape, |
| 2832 | &v67[0] /* a */, &v70[0] /* b */, &v71[0] /* output */, |
| 2833 | threadpool /* threadpool */); |
| 2834 | } |
| 2835 | if (status != xnn_status_success) { |
| 2836 | std::cerr << "failed to setup operation #70" << std::endl; |
| 2837 | return ExecutionPlan(); |
| 2838 | } |
| 2839 | |
| 2840 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2841 | op71, |
| 2842 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 2843 | &v71[0] /* input */, &v72[0] /* output */, |
| 2844 | threadpool /* threadpool */); |
| 2845 | if (status != xnn_status_success) { |
| 2846 | std::cerr << "failed to setup operation #71" << std::endl; |
| 2847 | return ExecutionPlan(); |
| 2848 | } |
| 2849 | |
| 2850 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2851 | op72, |
| 2852 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 2853 | &v72[0] /* input */, &v73[0] /* output */, |
| 2854 | threadpool /* threadpool */); |
| 2855 | if (status != xnn_status_success) { |
| 2856 | std::cerr << "failed to setup operation #72" << std::endl; |
| 2857 | return ExecutionPlan(); |
| 2858 | } |
| 2859 | |
| 2860 | status = xnn_setup_hardswish_nc_f32( |
| 2861 | op73, |
| 2862 | 49 /* batch size */, |
| 2863 | &v73[0] /* input */, &v74[0] /* output */, |
| 2864 | threadpool /* threadpool */); |
| 2865 | if (status != xnn_status_success) { |
| 2866 | std::cerr << "failed to setup operation #73" << std::endl; |
| 2867 | return ExecutionPlan(); |
| 2868 | } |
| 2869 | |
| 2870 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2871 | op74, |
| 2872 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 2873 | &v74[0] /* input */, &v75[0] /* output */, |
| 2874 | threadpool /* threadpool */); |
| 2875 | if (status != xnn_status_success) { |
| 2876 | std::cerr << "failed to setup operation #74" << std::endl; |
| 2877 | return ExecutionPlan(); |
| 2878 | } |
| 2879 | |
| 2880 | status = xnn_setup_hardswish_nc_f32( |
| 2881 | op75, |
| 2882 | 49 /* batch size */, |
| 2883 | &v75[0] /* input */, &v76[0] /* output */, |
| 2884 | threadpool /* threadpool */); |
| 2885 | if (status != xnn_status_success) { |
| 2886 | std::cerr << "failed to setup operation #75" << std::endl; |
| 2887 | return ExecutionPlan(); |
| 2888 | } |
| 2889 | |
| 2890 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 2891 | op76, |
| 2892 | 1 /* batch size */, 49 /* width */, |
| 2893 | &v76[0] /* input */, &v77[0] /* output */, |
| 2894 | threadpool /* threadpool */); |
| 2895 | if (status != xnn_status_success) { |
| 2896 | std::cerr << "failed to setup operation #76" << std::endl; |
| 2897 | return ExecutionPlan(); |
| 2898 | } |
| 2899 | |
| 2900 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2901 | op77, |
| 2902 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2903 | &v77[0] /* input */, &v78[0] /* output */, |
| 2904 | threadpool /* threadpool */); |
| 2905 | if (status != xnn_status_success) { |
| 2906 | std::cerr << "failed to setup operation #77" << std::endl; |
| 2907 | return ExecutionPlan(); |
| 2908 | } |
| 2909 | |
| 2910 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2911 | op78, |
| 2912 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 2913 | &v78[0] /* input */, &v79[0] /* output */, |
| 2914 | threadpool /* threadpool */); |
| 2915 | if (status != xnn_status_success) { |
| 2916 | std::cerr << "failed to setup operation #78" << std::endl; |
| 2917 | return ExecutionPlan(); |
| 2918 | } |
| 2919 | |
| 2920 | { |
| 2921 | const size_t a_shape[] = { 1, 7, 7, 576 }; |
| 2922 | const size_t b_shape[] = { 1, 1, 1, 576 }; |
| 2923 | status = xnn_setup_multiply_nd_f32( |
| 2924 | op79, |
| 2925 | 4, a_shape, 4, b_shape, |
| 2926 | &v76[0] /* a */, &v79[0] /* b */, &v80[0] /* output */, |
| 2927 | threadpool /* threadpool */); |
| 2928 | } |
| 2929 | if (status != xnn_status_success) { |
| 2930 | std::cerr << "failed to setup operation #79" << std::endl; |
| 2931 | return ExecutionPlan(); |
| 2932 | } |
| 2933 | |
| 2934 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2935 | op80, |
| 2936 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 2937 | &v80[0] /* input */, &v81[0] /* output */, |
| 2938 | threadpool /* threadpool */); |
| 2939 | if (status != xnn_status_success) { |
| 2940 | std::cerr << "failed to setup operation #80" << std::endl; |
| 2941 | return ExecutionPlan(); |
| 2942 | } |
| 2943 | |
| 2944 | status = xnn_setup_add_nc_f32( |
| 2945 | op81, |
| 2946 | 49 /* batch size */, |
| 2947 | &v81[0] /* a */, &v72[0] /* b */, &v82[0] /* sum */, |
| 2948 | threadpool /* threadpool */); |
| 2949 | if (status != xnn_status_success) { |
| 2950 | std::cerr << "failed to setup operation #81" << std::endl; |
| 2951 | return ExecutionPlan(); |
| 2952 | } |
| 2953 | |
| 2954 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2955 | op82, |
| 2956 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 2957 | &v82[0] /* input */, &v83[0] /* output */, |
| 2958 | threadpool /* threadpool */); |
| 2959 | if (status != xnn_status_success) { |
| 2960 | std::cerr << "failed to setup operation #82" << std::endl; |
| 2961 | return ExecutionPlan(); |
| 2962 | } |
| 2963 | |
| 2964 | status = xnn_setup_hardswish_nc_f32( |
| 2965 | op83, |
| 2966 | 49 /* batch size */, |
| 2967 | &v83[0] /* input */, &v84[0] /* output */, |
| 2968 | threadpool /* threadpool */); |
| 2969 | if (status != xnn_status_success) { |
| 2970 | std::cerr << "failed to setup operation #83" << std::endl; |
| 2971 | return ExecutionPlan(); |
| 2972 | } |
| 2973 | |
| 2974 | status = xnn_setup_convolution2d_nhwc_f32( |
| 2975 | op84, |
| 2976 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 2977 | &v84[0] /* input */, &v85[0] /* output */, |
| 2978 | threadpool /* threadpool */); |
| 2979 | if (status != xnn_status_success) { |
| 2980 | std::cerr << "failed to setup operation #84" << std::endl; |
| 2981 | return ExecutionPlan(); |
| 2982 | } |
| 2983 | |
| 2984 | status = xnn_setup_hardswish_nc_f32( |
| 2985 | op85, |
| 2986 | 49 /* batch size */, |
| 2987 | &v85[0] /* input */, &v86[0] /* output */, |
| 2988 | threadpool /* threadpool */); |
| 2989 | if (status != xnn_status_success) { |
| 2990 | std::cerr << "failed to setup operation #85" << std::endl; |
| 2991 | return ExecutionPlan(); |
| 2992 | } |
| 2993 | |
| 2994 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 2995 | op86, |
| 2996 | 1 /* batch size */, 49 /* width */, |
| 2997 | &v86[0] /* input */, &v87[0] /* output */, |
| 2998 | threadpool /* threadpool */); |
| 2999 | if (status != xnn_status_success) { |
| 3000 | std::cerr << "failed to setup operation #86" << std::endl; |
| 3001 | return ExecutionPlan(); |
| 3002 | } |
| 3003 | |
| 3004 | status = xnn_setup_convolution2d_nhwc_f32( |
| 3005 | op87, |
| 3006 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3007 | &v87[0] /* input */, &v88[0] /* output */, |
| 3008 | threadpool /* threadpool */); |
| 3009 | if (status != xnn_status_success) { |
| 3010 | std::cerr << "failed to setup operation #87" << std::endl; |
| 3011 | return ExecutionPlan(); |
| 3012 | } |
| 3013 | |
| 3014 | status = xnn_setup_convolution2d_nhwc_f32( |
| 3015 | op88, |
| 3016 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3017 | &v88[0] /* input */, &v89[0] /* output */, |
| 3018 | threadpool /* threadpool */); |
| 3019 | if (status != xnn_status_success) { |
| 3020 | std::cerr << "failed to setup operation #88" << std::endl; |
| 3021 | return ExecutionPlan(); |
| 3022 | } |
| 3023 | |
| 3024 | { |
| 3025 | const size_t a_shape[] = { 1, 7, 7, 576 }; |
| 3026 | const size_t b_shape[] = { 1, 1, 1, 576 }; |
| 3027 | status = xnn_setup_multiply_nd_f32( |
| 3028 | op89, |
| 3029 | 4, a_shape, 4, b_shape, |
| 3030 | &v86[0] /* a */, &v89[0] /* b */, &v90[0] /* output */, |
| 3031 | threadpool /* threadpool */); |
| 3032 | } |
| 3033 | if (status != xnn_status_success) { |
| 3034 | std::cerr << "failed to setup operation #89" << std::endl; |
| 3035 | return ExecutionPlan(); |
| 3036 | } |
| 3037 | |
| 3038 | status = xnn_setup_convolution2d_nhwc_f32( |
| 3039 | op90, |
| 3040 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3041 | &v90[0] /* input */, &v91[0] /* output */, |
| 3042 | threadpool /* threadpool */); |
| 3043 | if (status != xnn_status_success) { |
| 3044 | std::cerr << "failed to setup operation #90" << std::endl; |
| 3045 | return ExecutionPlan(); |
| 3046 | } |
| 3047 | |
| 3048 | status = xnn_setup_add_nc_f32( |
| 3049 | op91, |
| 3050 | 49 /* batch size */, |
| 3051 | &v91[0] /* a */, &v82[0] /* b */, &v92[0] /* sum */, |
| 3052 | threadpool /* threadpool */); |
| 3053 | if (status != xnn_status_success) { |
| 3054 | std::cerr << "failed to setup operation #91" << std::endl; |
| 3055 | return ExecutionPlan(); |
| 3056 | } |
| 3057 | |
| 3058 | status = xnn_setup_convolution2d_nhwc_f32( |
| 3059 | op92, |
| 3060 | 1 /* batch size */, 7 /* input height */, 7 /* input width */, |
| 3061 | &v92[0] /* input */, &v93[0] /* output */, |
| 3062 | threadpool /* threadpool */); |
| 3063 | if (status != xnn_status_success) { |
| 3064 | std::cerr << "failed to setup operation #92" << std::endl; |
| 3065 | return ExecutionPlan(); |
| 3066 | } |
| 3067 | |
| 3068 | status = xnn_setup_hardswish_nc_f32( |
| 3069 | op93, |
| 3070 | 49 /* batch size */, |
| 3071 | &v93[0] /* input */, &v94[0] /* output */, |
| 3072 | threadpool /* threadpool */); |
| 3073 | if (status != xnn_status_success) { |
| 3074 | std::cerr << "failed to setup operation #93" << std::endl; |
| 3075 | return ExecutionPlan(); |
| 3076 | } |
| 3077 | |
| 3078 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 3079 | op94, |
| 3080 | 1 /* batch size */, 49 /* width */, |
| 3081 | &v94[0] /* input */, &v95[0] /* output */, |
| 3082 | threadpool /* threadpool */); |
| 3083 | if (status != xnn_status_success) { |
| 3084 | std::cerr << "failed to setup operation #94" << std::endl; |
| 3085 | return ExecutionPlan(); |
| 3086 | } |
| 3087 | |
| 3088 | status = xnn_setup_convolution2d_nhwc_f32( |
| 3089 | op95, |
| 3090 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3091 | &v95[0] /* input */, &v96[0] /* output */, |
| 3092 | threadpool /* threadpool */); |
| 3093 | if (status != xnn_status_success) { |
| 3094 | std::cerr << "failed to setup operation #95" << std::endl; |
| 3095 | return ExecutionPlan(); |
| 3096 | } |
| 3097 | |
| 3098 | status = xnn_setup_hardswish_nc_f32( |
| 3099 | op96, |
| 3100 | 1 /* batch size */, |
| 3101 | &v96[0] /* input */, &v97[0] /* output */, |
| 3102 | threadpool /* threadpool */); |
| 3103 | if (status != xnn_status_success) { |
| 3104 | std::cerr << "failed to setup operation #96" << std::endl; |
| 3105 | return ExecutionPlan(); |
| 3106 | } |
| 3107 | |
| 3108 | status = xnn_setup_global_average_pooling_nwc_f32( |
| 3109 | op97, |
| 3110 | 1 /* batch size */, 1 /* width */, |
| 3111 | &v97[0] /* input */, &v98[0] /* output */, |
| 3112 | threadpool /* threadpool */); |
| 3113 | if (status != xnn_status_success) { |
| 3114 | std::cerr << "failed to setup operation #97" << std::endl; |
| 3115 | return ExecutionPlan(); |
| 3116 | } |
| 3117 | |
| 3118 | status = xnn_setup_convolution2d_nhwc_f32( |
| 3119 | op98, |
| 3120 | 1 /* batch size */, 1 /* input height */, 1 /* input width */, |
| 3121 | &v98[0] /* input */, &v99[0] /* output */, |
| 3122 | threadpool /* threadpool */); |
| 3123 | if (status != xnn_status_success) { |
| 3124 | std::cerr << "failed to setup operation #98" << std::endl; |
| 3125 | return ExecutionPlan(); |
| 3126 | } |
| 3127 | |
| 3128 | #pragma clang diagnostic push |
| 3129 | #pragma clang diagnostic ignored "-Wpessimizing-move" |
| 3130 | return operators; |
| 3131 | #pragma clang diagnostic pop |
| 3132 | } |
| 3133 | |
| 3134 | } // namespace models |