请问一下怎么将 yolov5-lite 的模型转换为 正确的 onnx 模式

[Singformeasong]

我知道这个问题很小白，但是我用 models/export.py 导出来的onnx模型 最终output 为

name: output
type: float32[1,3,80,80,85]
name: 591
type: float32[1,3,40,40,85]
name: 605
type: float32[1,3,20,20,85]

而我看其他人正确导出的模型

name: outputs
type: float32[1,25200,85]

我要怎么做才能正确导出....

[ppogg]

你转的是对的
一般转YOLOX才是你下面的形式，也就是tensor耦合在一起了

[Singformeasong]

你转的是对的 一般转YOLOX才是你下面的形式，也就是tensor耦合在一起了

我使用的是 master分支版本的 v5lite

我按照 #53 的教程，将大佬你存在网盘中的 v5lite-s.pt 转换成 ncnn格式后
运行 cpp_demo/ncnn/v5lite-s.cpp 报错：

而我在 https://github.com/ppogg/ncnn-android-v5lite/tree/master/app/src/main/assets 中下载的模型在 v5lite-s.cpp 中运行是正确的...

我修改后的 v5lite-s-fp16.param 如下:

7767517
206 232
Input images 0 1 images
Convolution Conv_0 1 1 images 173 0=32 1=3 3=2 4=1 5=1 6=864 9=1
Pooling MaxPool_2 1 1 173 174 1=3 2=2 3=1 5=1
Split splitncnn_0 1 2 174 174_splitncnn_0 174_splitncnn_1
ConvolutionDepthWise Conv_3 1 1 174_splitncnn_1 175 0=32 1=3 3=2 4=1 5=1 6=288 7=32
Convolution Conv_4 1 1 175 177 0=60 1=1 5=1 6=1920 9=1
Convolution Conv_6 1 1 174_splitncnn_0 179 0=60 1=1 5=1 6=1920 9=1
ConvolutionDepthWise Conv_8 1 1 179 180 0=60 1=3 3=2 4=1 5=1 6=540 7=60
Convolution Conv_9 1 1 180 182 0=60 1=1 5=1 6=3600 9=1
Concat Concat_11 2 1 177 182 183
ShuffleChannel Reshape_16 1 1 183 188 0=2
Split splitncnn_1 1 2 188 188_splitncnn_0 188_splitncnn_1
Crop Slice_27 1 1 188_splitncnn_1 199 -23309=1,0 -23310=1,60 -23311=1,0
Crop Slice_30 1 1 188_splitncnn_0 202 -23309=1,60 -23310=1,120 -23311=1,0
Convolution Conv_31 1 1 202 204 0=60 1=1 5=1 6=3600 9=1
ConvolutionDepthWise Conv_33 1 1 204 205 0=60 1=3 4=1 5=1 6=540 7=60
Convolution Conv_34 1 1 205 207 0=60 1=1 5=1 6=3600 9=1
Concat Concat_36 2 1 199 207 208
ShuffleChannel Reshape_41 1 1 208 213 0=2
Split splitncnn_2 1 2 213 213_splitncnn_0 213_splitncnn_1
Crop Slice_52 1 1 213_splitncnn_1 224 -23309=1,0 -23310=1,60 -23311=1,0
Crop Slice_55 1 1 213_splitncnn_0 227 -23309=1,60 -23310=1,120 -23311=1,0
Convolution Conv_56 1 1 227 229 0=60 1=1 5=1 6=3600 9=1
ConvolutionDepthWise Conv_58 1 1 229 230 0=60 1=3 4=1 5=1 6=540 7=60
Convolution Conv_59 1 1 230 232 0=60 1=1 5=1 6=3600 9=1
Concat Concat_61 2 1 224 232 233
ShuffleChannel Reshape_66 1 1 233 238 0=2
Split splitncnn_3 1 2 238 238_splitncnn_0 238_splitncnn_1
Crop Slice_77 1 1 238_splitncnn_1 249 -23309=1,0 -23310=1,60 -23311=1,0
Crop Slice_80 1 1 238_splitncnn_0 252 -23309=1,60 -23310=1,120 -23311=1,0
Convolution Conv_81 1 1 252 254 0=60 1=1 5=1 6=3600 9=1
ConvolutionDepthWise Conv_83 1 1 254 255 0=60 1=3 4=1 5=1 6=540 7=60
Convolution Conv_84 1 1 255 257 0=60 1=1 5=1 6=3600 9=1
Concat Concat_86 2 1 249 257 258
ShuffleChannel Reshape_91 1 1 258 263 0=2
Split splitncnn_4 1 3 263 263_splitncnn_0 263_splitncnn_1 263_splitncnn_2
ConvolutionDepthWise Conv_92 1 1 263_splitncnn_2 264 0=120 1=3 3=2 4=1 5=1 6=1080 7=120
Convolution Conv_93 1 1 264 266 0=116 1=1 5=1 6=13920 9=1
Convolution Conv_95 1 1 263_splitncnn_1 268 0=116 1=1 5=1 6=13920 9=1
ConvolutionDepthWise Conv_97 1 1 268 269 0=116 1=3 3=2 4=1 5=1 6=1044 7=116
Convolution Conv_98 1 1 269 271 0=116 1=1 5=1 6=13456 9=1
Concat Concat_100 2 1 266 271 272
ShuffleChannel Reshape_105 1 1 272 277 0=2
Split splitncnn_5 1 2 277 277_splitncnn_0 277_splitncnn_1
Crop Slice_116 1 1 277_splitncnn_1 288 -23309=1,0 -23310=1,116 -23311=1,0
Crop Slice_119 1 1 277_splitncnn_0 291 -23309=1,116 -23310=1,232 -23311=1,0
Convolution Conv_120 1 1 291 293 0=116 1=1 5=1 6=13456 9=1
ConvolutionDepthWise Conv_122 1 1 293 294 0=116 1=3 4=1 5=1 6=1044 7=116
Convolution Conv_123 1 1 294 296 0=116 1=1 5=1 6=13456 9=1
Concat Concat_125 2 1 288 296 297
ShuffleChannel Reshape_130 1 1 297 302 0=2
Split splitncnn_6 1 2 302 302_splitncnn_0 302_splitncnn_1
Crop Slice_141 1 1 302_splitncnn_1 313 -23309=1,0 -23310=1,116 -23311=1,0
Crop Slice_144 1 1 302_splitncnn_0 316 -23309=1,116 -23310=1,232 -23311=1,0
Convolution Conv_145 1 1 316 318 0=116 1=1 5=1 6=13456 9=1
ConvolutionDepthWise Conv_147 1 1 318 319 0=116 1=3 4=1 5=1 6=1044 7=116
Convolution Conv_148 1 1 319 321 0=116 1=1 5=1 6=13456 9=1
Concat Concat_150 2 1 313 321 322
ShuffleChannel Reshape_155 1 1 322 327 0=2
Split splitncnn_7 1 2 327 327_splitncnn_0 327_splitncnn_1
Crop Slice_166 1 1 327_splitncnn_1 338 -23309=1,0 -23310=1,116 -23311=1,0
Crop Slice_169 1 1 327_splitncnn_0 341 -23309=1,116 -23310=1,232 -23311=1,0
Convolution Conv_170 1 1 341 343 0=116 1=1 5=1 6=13456 9=1
ConvolutionDepthWise Conv_172 1 1 343 344 0=116 1=3 4=1 5=1 6=1044 7=116
Convolution Conv_173 1 1 344 346 0=116 1=1 5=1 6=13456 9=1
Concat Concat_175 2 1 338 346 347
ShuffleChannel Reshape_180 1 1 347 352 0=2
Split splitncnn_8 1 2 352 352_splitncnn_0 352_splitncnn_1
Crop Slice_191 1 1 352_splitncnn_1 363 -23309=1,0 -23310=1,116 -23311=1,0
Crop Slice_194 1 1 352_splitncnn_0 366 -23309=1,116 -23310=1,232 -23311=1,0
Convolution Conv_195 1 1 366 368 0=116 1=1 5=1 6=13456 9=1
ConvolutionDepthWise Conv_197 1 1 368 369 0=116 1=3 4=1 5=1 6=1044 7=116
Convolution Conv_198 1 1 369 371 0=116 1=1 5=1 6=13456 9=1
Concat Concat_200 2 1 363 371 372
ShuffleChannel Reshape_205 1 1 372 377 0=2
Split splitncnn_9 1 2 377 377_splitncnn_0 377_splitncnn_1
Crop Slice_216 1 1 377_splitncnn_1 388 -23309=1,0 -23310=1,116 -23311=1,0
Crop Slice_219 1 1 377_splitncnn_0 391 -23309=1,116 -23310=1,232 -23311=1,0
Convolution Conv_220 1 1 391 393 0=116 1=1 5=1 6=13456 9=1
ConvolutionDepthWise Conv_222 1 1 393 394 0=116 1=3 4=1 5=1 6=1044 7=116
Convolution Conv_223 1 1 394 396 0=116 1=1 5=1 6=13456 9=1
Concat Concat_225 2 1 388 396 397
ShuffleChannel Reshape_230 1 1 397 402 0=2
Split splitncnn_10 1 2 402 402_splitncnn_0 402_splitncnn_1
Crop Slice_241 1 1 402_splitncnn_1 413 -23309=1,0 -23310=1,116 -23311=1,0
Crop Slice_244 1 1 402_splitncnn_0 416 -23309=1,116 -23310=1,232 -23311=1,0
Convolution Conv_245 1 1 416 418 0=116 1=1 5=1 6=13456 9=1
ConvolutionDepthWise Conv_247 1 1 418 419 0=116 1=3 4=1 5=1 6=1044 7=116
Convolution Conv_248 1 1 419 421 0=116 1=1 5=1 6=13456 9=1
Concat Concat_250 2 1 413 421 422
ShuffleChannel Reshape_255 1 1 422 427 0=2
Split splitncnn_11 1 2 427 427_splitncnn_0 427_splitncnn_1
Crop Slice_266 1 1 427_splitncnn_1 438 -23309=1,0 -23310=1,116 -23311=1,0
Crop Slice_269 1 1 427_splitncnn_0 441 -23309=1,116 -23310=1,232 -23311=1,0
Convolution Conv_270 1 1 441 443 0=116 1=1 5=1 6=13456 9=1
ConvolutionDepthWise Conv_272 1 1 443 444 0=116 1=3 4=1 5=1 6=1044 7=116
Convolution Conv_273 1 1 444 446 0=116 1=1 5=1 6=13456 9=1
Concat Concat_275 2 1 438 446 447
ShuffleChannel Reshape_280 1 1 447 452 0=2
Split splitncnn_12 1 3 452 452_splitncnn_0 452_splitncnn_1 452_splitncnn_2
ConvolutionDepthWise Conv_281 1 1 452_splitncnn_2 453 0=232 1=3 3=2 4=1 5=1 6=2088 7=232
Convolution Conv_282 1 1 453 455 0=232 1=1 5=1 6=53824 9=1
Convolution Conv_284 1 1 452_splitncnn_1 457 0=232 1=1 5=1 6=53824 9=1
ConvolutionDepthWise Conv_286 1 1 457 458 0=232 1=3 3=2 4=1 5=1 6=2088 7=232
Convolution Conv_287 1 1 458 460 0=232 1=1 5=1 6=53824 9=1
Concat Concat_289 2 1 455 460 461
ShuffleChannel Reshape_294 1 1 461 466 0=2
Split splitncnn_13 1 2 466 466_splitncnn_0 466_splitncnn_1
Crop Slice_305 1 1 466_splitncnn_1 477 -23309=1,0 -23310=1,232 -23311=1,0
Crop Slice_308 1 1 466_splitncnn_0 480 -23309=1,232 -23310=1,464 -23311=1,0
Convolution Conv_309 1 1 480 482 0=232 1=1 5=1 6=53824 9=1
ConvolutionDepthWise Conv_311 1 1 482 483 0=232 1=3 4=1 5=1 6=2088 7=232
Convolution Conv_312 1 1 483 485 0=232 1=1 5=1 6=53824 9=1
Concat Concat_314 2 1 477 485 486
ShuffleChannel Reshape_319 1 1 486 491 0=2
Split splitncnn_14 1 2 491 491_splitncnn_0 491_splitncnn_1
Crop Slice_330 1 1 491_splitncnn_1 502 -23309=1,0 -23310=1,232 -23311=1,0
Crop Slice_333 1 1 491_splitncnn_0 505 -23309=1,232 -23310=1,464 -23311=1,0
Convolution Conv_334 1 1 505 507 0=232 1=1 5=1 6=53824 9=1
ConvolutionDepthWise Conv_336 1 1 507 508 0=232 1=3 4=1 5=1 6=2088 7=232
Convolution Conv_337 1 1 508 510 0=232 1=1 5=1 6=53824 9=1
Concat Concat_339 2 1 502 510 511
ShuffleChannel Reshape_344 1 1 511 516 0=2
Split splitncnn_15 1 2 516 516_splitncnn_0 516_splitncnn_1
Crop Slice_355 1 1 516_splitncnn_1 527 -23309=1,0 -23310=1,232 -23311=1,0
Crop Slice_358 1 1 516_splitncnn_0 530 -23309=1,232 -23310=1,464 -23311=1,0
Convolution Conv_359 1 1 530 532 0=232 1=1 5=1 6=53824 9=1
ConvolutionDepthWise Conv_361 1 1 532 533 0=232 1=3 4=1 5=1 6=2088 7=232
Convolution Conv_362 1 1 533 535 0=232 1=1 5=1 6=53824 9=1
Concat Concat_364 2 1 527 535 536
ShuffleChannel Reshape_369 1 1 536 541 0=2
Convolution Conv_370 1 1 541 542 0=128 1=1 5=1 6=59392
Swish Mul_372 1 1 542 544
Split splitncnn_16 1 2 544 544_splitncnn_0 544_splitncnn_1
Interp Resize_374 1 1 544_splitncnn_1 549 0=1 1=2.000000e+00 2=2.000000e+00
Concat Concat_375 2 1 549 452_splitncnn_0 550
Split splitncnn_17 1 2 550 550_splitncnn_0 550_splitncnn_1
Convolution Conv_376 1 1 550_splitncnn_1 551 0=64 1=1 5=1 6=23040
Swish Mul_378 1 1 551 553
Convolution Conv_379 1 1 553 554 0=64 1=1 5=1 6=4096
Swish Mul_381 1 1 554 556
Convolution Conv_382 1 1 556 557 0=64 1=3 4=1 5=1 6=36864
Swish Mul_384 1 1 557 559
Convolution Conv_385 1 1 550_splitncnn_0 560 0=64 1=1 5=1 6=23040
Swish Mul_387 1 1 560 562
Concat Concat_388 2 1 559 562 563
Convolution Conv_389 1 1 563 564 0=128 1=1 5=1 6=16384
Swish Mul_391 1 1 564 566
Convolution Conv_392 1 1 566 567 0=64 1=1 5=1 6=8192
Swish Mul_394 1 1 567 569
Split splitncnn_18 1 2 569 569_splitncnn_0 569_splitncnn_1
Interp Resize_396 1 1 569_splitncnn_1 574 0=1 1=2.000000e+00 2=2.000000e+00
Concat Concat_397 2 1 574 263_splitncnn_0 575
Split splitncnn_19 1 2 575 575_splitncnn_0 575_splitncnn_1
Convolution Conv_398 1 1 575_splitncnn_1 576 0=32 1=1 5=1 6=5888
Swish Mul_400 1 1 576 578
Convolution Conv_401 1 1 578 579 0=32 1=1 5=1 6=1024
Swish Mul_403 1 1 579 581
Convolution Conv_404 1 1 581 582 0=32 1=3 4=1 5=1 6=9216
Swish Mul_406 1 1 582 584
Convolution Conv_407 1 1 575_splitncnn_0 585 0=32 1=1 5=1 6=5888
Swish Mul_409 1 1 585 587
Concat Concat_410 2 1 584 587 588
Convolution Conv_411 1 1 588 589 0=64 1=1 5=1 6=4096
Swish Mul_413 1 1 589 591
Split splitncnn_20 1 2 591 591_splitncnn_0 591_splitncnn_1
Convolution Conv_414 1 1 591_splitncnn_1 592 0=64 1=3 3=2 4=1 5=1 6=36864
Swish Mul_416 1 1 592 594
Concat Concat_417 2 1 594 569_splitncnn_0 595
Split splitncnn_21 1 2 595 595_splitncnn_0 595_splitncnn_1
Convolution Conv_418 1 1 595_splitncnn_1 596 0=64 1=1 5=1 6=8192
Swish Mul_420 1 1 596 598
Convolution Conv_421 1 1 598 599 0=64 1=1 5=1 6=4096
Swish Mul_423 1 1 599 601
Convolution Conv_424 1 1 601 602 0=64 1=3 4=1 5=1 6=36864
Swish Mul_426 1 1 602 604
Convolution Conv_427 1 1 595_splitncnn_0 605 0=64 1=1 5=1 6=8192
Swish Mul_429 1 1 605 607
Concat Concat_430 2 1 604 607 608
Convolution Conv_431 1 1 608 609 0=128 1=1 5=1 6=16384
Swish Mul_433 1 1 609 611
Split splitncnn_22 1 2 611 611_splitncnn_0 611_splitncnn_1
Convolution Conv_434 1 1 611_splitncnn_1 612 0=128 1=3 3=2 4=1 5=1 6=147456
Swish Mul_436 1 1 612 614
Concat Concat_437 2 1 614 544_splitncnn_0 615
Split splitncnn_23 1 2 615 615_splitncnn_0 615_splitncnn_1
Convolution Conv_438 1 1 615_splitncnn_1 616 0=128 1=1 5=1 6=32768
Swish Mul_440 1 1 616 618
Convolution Conv_441 1 1 618 619 0=128 1=1 5=1 6=16384
Swish Mul_443 1 1 619 621
Convolution Conv_444 1 1 621 622 0=128 1=3 4=1 5=1 6=147456
Swish Mul_446 1 1 622 624
Convolution Conv_447 1 1 615_splitncnn_0 625 0=128 1=1 5=1 6=32768
Swish Mul_449 1 1 625 627
Concat Concat_450 2 1 624 627 628
Convolution Conv_451 1 1 628 629 0=256 1=1 5=1 6=65536
Swish Mul_453 1 1 629 631
Convolution Conv_454 1 1 591_splitncnn_0 632 0=255 1=1 5=1 6=16320
Reshape Reshape_468 1 1 632 650 0=-1 1=85 2=3
Permute Transpose_469 1 1 650 output 0=1
Convolution Conv_470 1 1 611_splitncnn_0 652 0=255 1=1 5=1 6=32640
Reshape Reshape_484 1 1 652 670 0=-1 1=85 2=3
Permute Transpose_485 1 1 670 671 0=1
Convolution Conv_486 1 1 631 672 0=255 1=1 5=1 6=65280
Reshape Reshape_500 1 1 672 690 0=-1 1=85 2=3
Permute Transpose_501 1 1 690 691 0=1

大佬能帮忙看看吗？目前我自己的模型训练好了,但是没法部署在C++中，这个模型转换卡了我好久了...