pytorch
pytorch copied to clipboard
Published
20 hours ago •
pytorch
pytorch
Code update for vectorized interpolate cpu uint8
Stack from ghstack (oldest at bottom):
- #96848
- -> #96847
- code style update
- use idx_ptr_xmin/idx_ptr_size instead of bounds
- compute wt_max inside _compute_indices_weights_aa (no significant overhead)
- added comments and explanations
- renamed xmin/xmax into ids_min, ids_size
cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10 @datumbox @pmeier
:link: Helpful Links
:test_tube: See artifacts and rendered test results at hud.pytorch.org/pr/96847
- :page_facing_up: Preview Python docs built from this PR
- :page_facing_up: Preview C++ docs built from this PR
- :question: Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours
Note: Links to docs will display an error until the docs builds have been completed.
:white_check_mark: No Failures
As of commit c44323baeed6e2080911d7695cfc5905389d3d36: :green_heart: Looks good so far! There are no failures yet. :green_heart:
This comment was automatically generated by Dr. CI and updates every 15 minutes.
![pytorch-bot[bot] avatar](https://avatars.githubusercontent.com/in/40112?v=4 "Published by a pub.dev verified publisher"){kind=small}
I'll push few updates into the code based on review comments and to avoid accidental merge I set the PR as draft
Results of benchmarks for this PR. There 4 test cases marked with "X" which give unreliable results. I rerun them explicitly and provided results below.
Click here
Description:
- 20230323-164856-pr-96847
Torch version: 2.1.0a0+git7b834d6
Torch config: PyTorch built with:
- GCC 9.4
- C++ Version: 201703
- OpenMP 201511 (a.k.a. OpenMP 4.5)
- CPU capability usage: AVX2
- Build settings: BUILD_TYPE=Release, CXX_COMPILER=/usr/bin/c++, CXX_FLAGS= -D_GLIBCXX_USE_CXX11_ABI=1 -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOCUPTI -DLIBKINETO_NOROCTRACER -DUSE_PYTORCH_QNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_DISABLE_GPU_ASSERTS=ON, TORCH_VERSION=2.1.0, USE_CUDA=0, USE_CUDNN=OFF, USE_EIGEN_FOR_BLAS=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=OFF, USE_MKLDNN=0, USE_MPI=OFF, USE_NCCL=OFF, USE_NNPACK=0, USE_OPENMP=ON, USE_ROCM=OFF,
- 20230323-172413-nightly
Torch version: 2.1.0a0+git5309c44
Torch config: PyTorch built with:
- GCC 9.4
- C++ Version: 201703
- OpenMP 201511 (a.k.a. OpenMP 4.5)
- CPU capability usage: AVX2
- Build settings: BUILD_TYPE=Release, CXX_COMPILER=/usr/bin/c++, CXX_FLAGS= -D_GLIBCXX_USE_CXX11_ABI=1 -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOCUPTI -DLIBKINETO_NOROCTRACER -DUSE_PYTORCH_QNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_DISABLE_GPU_ASSERTS=ON, TORCH_VERSION=2.1.0, USE_CUDA=0, USE_CUDNN=OFF, USE_EIGEN_FOR_BLAS=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=OFF, USE_MKLDNN=0, USE_MPI=OFF, USE_NCCL=OFF, USE_NNPACK=0, USE_OPENMP=ON, USE_ROCM=OFF,
[------------------------------------------------------------------------------------------------------------- Resize ------------------------------------------------------------------------------------------------------------]
| Pillow (9.0.0.post1) | torch (2.1.0a0+git7b834d6) ghstack-PR-96847 | torch (2.1.0a0+git5309c44) nightly | Speed-up: PR-96847 vs nightly
1 threads: ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
3 torch.uint8 channels_last bilinear (256, 256) -> (32, 32) aa=True | 38.719 (+-0.511) | 129.155 (+-1.491) | 132.050 (+-2.890) | 1.022 (+-0.000)
3 torch.uint8 channels_last bilinear (256, 256) -> (32, 32) aa=False | | 112.135 (+-3.153) | 111.513 (+-2.585) | 0.994 (+-0.000)
3 torch.uint8 channels_last bilinear (256, 256) -> (224, 224) aa=True | 127.837 (+-1.687) | 295.055 (+-2.561) | 300.912 (+-1.963) | 1.020 (+-0.000)
3 torch.uint8 channels_last bilinear (256, 256) -> (224, 224) aa=False | | 281.954 (+-1.805) | 287.682 (+-3.179) | 1.020 (+-0.000)
3 torch.uint8 channels_last bilinear (256, 256) -> (320, 320) aa=True | 178.967 (+-1.598) | 424.912 (+-4.052) | 435.762 (+-2.653) | 1.026 (+-0.000)
3 torch.uint8 channels_last bilinear (256, 256) -> (320, 320) aa=False | | 421.486 (+-3.018) | 431.192 (+-3.461) | 1.023 (+-0.000)
3 torch.uint8 channels_last bilinear (520, 520) -> (32, 32) aa=True | 112.960 (+-0.995) | 431.450 (+-4.843) | 451.762 (+-9.525) | 1.047 (+-0.000)
X 3 torch.uint8 channels_last bilinear (520, 520) -> (32, 32) aa=False | | 416.816 (+-53.366) | 363.680 (+-2.959) | 0.873 (+-0.000)
x 3 torch.uint8 channels_last bilinear (520, 520) -> (224, 224) aa=True | 281.495 (+-2.045) | 669.758 (+-2.280) | 811.035 (+-4.295) | 1.211 (+-0.000)
3 torch.uint8 channels_last bilinear (520, 520) -> (224, 224) aa=False | | 623.003 (+-8.695) | 591.464 (+-2.582) | 0.949 (+-0.000)
3 torch.uint8 channels_last bilinear (712, 712) -> (32, 32) aa=True | 186.052 (+-1.111) | 770.460 (+-12.775) | 837.882 (+-21.376) | 1.088 (+-0.000)
X 3 torch.uint8 channels_last bilinear (712, 712) -> (32, 32) aa=False | | 899.304 (+-13.748) | 659.327 (+-20.590) | 0.733 (+-0.000)
X 3 torch.uint8 channels_last bilinear (712, 712) -> (224, 224) aa=True | 409.952 (+-2.552) | 1080.221 (+-124.780) | 1323.303 (+-93.724) | 1.225 (+-0.000)
3 torch.uint8 channels_last bilinear (712, 712) -> (224, 224) aa=False | | 924.238 (+-3.399) | 1110.081 (+-154.447) | 1.201 (+-0.000)
3 torch.uint8 channels_last bilinear (64, 64) -> (224, 224) aa=True | 60.513 (+-0.272) | 157.454 (+-1.262) | 165.964 (+-2.433) | 1.054 (+-0.000)
3 torch.uint8 channels_last bilinear (224, 224) -> (270, 268) aa=True | 132.726 (+-1.190) | 319.613 (+-2.402) | 327.600 (+-2.460) | 1.025 (+-0.000)
3 torch.uint8 channels_last bilinear (256, 256) -> (1024, 1024) aa=True | 947.430 (+-3.370) | 2580.283 (+-20.541) | 2772.801 (+-23.188) | 1.075 (+-0.000)
3 torch.uint8 channels_last bilinear (224, 224) -> (64, 64) aa=True | 52.687 (+-0.310) | 130.011 (+-1.159) | 134.335 (+-1.196) | 1.033 (+-0.000)
3 torch.uint8 channels_last bilinear (270, 268) -> (224, 224) aa=True | 138.507 (+-1.109) | 314.773 (+-2.458) | 318.490 (+-1.922) | 1.012 (+-0.000)
3 torch.uint8 channels_last bilinear (1024, 1024) -> (256, 256) aa=True | 690.144 (+-10.954) | 1981.538 (+-59.000) | 2042.357 (+-20.922) | 1.031 (+-0.000)
3 torch.uint8 channels_last bilinear (64, 64) -> (224, 224) aa=False | | 155.135 (+-0.878) | 165.086 (+-2.331) | 1.064 (+-0.000)
3 torch.uint8 channels_last bilinear (224, 224) -> (270, 268) aa=False | | 316.657 (+-1.607) | 323.979 (+-2.686) | 1.023 (+-0.000)
3 torch.uint8 channels_last bilinear (256, 256) -> (1024, 1024) aa=False | | 2578.702 (+-21.677) | 2784.678 (+-22.486) | 1.080 (+-0.000)
3 torch.uint8 channels_last bilinear (224, 224) -> (64, 64) aa=False | | 111.604 (+-1.021) | 111.504 (+-1.046) | 0.999 (+-0.000)
3 torch.uint8 channels_last bilinear (270, 268) -> (224, 224) aa=False | | 295.114 (+-2.685) | 298.357 (+-2.184) | 1.011 (+-0.000)
3 torch.uint8 channels_last bilinear (1024, 1024) -> (256, 256) aa=False | | 1726.498 (+-15.274) | 1724.288 (+-37.707) | 0.999 (+-0.000)
3 torch.float32 channels_last bilinear (256, 256) -> (32, 32) aa=True | | 317.711 (+-1.736) | 315.736 (+-1.944) | 0.994 (+-0.000)
3 torch.float32 channels_last bilinear (256, 256) -> (32, 32) aa=False | | 24.486 (+-0.118) | 24.696 (+-0.136) | 1.009 (+-0.000)
3 torch.float32 channels_last bilinear (256, 256) -> (224, 224) aa=True | | 1045.402 (+-5.033) | 1042.310 (+-5.169) | 0.997 (+-0.000)
3 torch.float32 channels_last bilinear (256, 256) -> (224, 224) aa=False | | 762.002 (+-2.855) | 759.993 (+-2.310) | 0.997 (+-0.000)
3 torch.float32 channels_last bilinear (256, 256) -> (320, 320) aa=True | | 1766.138 (+-7.833) | 1768.470 (+-6.633) | 1.001 (+-0.000)
3 torch.float32 channels_last bilinear (256, 256) -> (320, 320) aa=False | | 1542.827 (+-9.588) | 1538.301 (+-13.990) | 0.997 (+-0.000)
3 torch.float32 channels_last bilinear (520, 520) -> (32, 32) aa=True | | 1439.554 (+-6.427) | 1400.032 (+-5.496) | 0.973 (+-0.000)
3 torch.float32 channels_last bilinear (520, 520) -> (32, 32) aa=False | | 26.064 (+-0.097) | 26.145 (+-0.176) | 1.003 (+-0.000)
3 torch.float32 channels_last bilinear (520, 520) -> (224, 224) aa=True | | 2058.325 (+-7.637) | 2059.229 (+-7.675) | 1.000 (+-0.000)
3 torch.float32 channels_last bilinear (520, 520) -> (224, 224) aa=False | | 765.381 (+-2.524) | 762.443 (+-1.976) | 0.996 (+-0.000)
3 torch.float32 channels_last bilinear (712, 712) -> (32, 32) aa=True | | 2669.941 (+-15.326) | 2658.132 (+-12.180) | 0.996 (+-0.000)
3 torch.float32 channels_last bilinear (712, 712) -> (32, 32) aa=False | | 26.666 (+-0.130) | 26.628 (+-0.130) | 0.999 (+-0.000)
3 torch.float32 channels_last bilinear (712, 712) -> (224, 224) aa=True | | 3153.329 (+-20.783) | 3160.121 (+-15.037) | 1.002 (+-0.000)
3 torch.float32 channels_last bilinear (712, 712) -> (224, 224) aa=False | | 766.780 (+-1.709) | 768.550 (+-3.093) | 1.002 (+-0.000)
3 torch.float32 channels_last bilinear (64, 64) -> (224, 224) aa=True | | 712.322 (+-2.249) | 710.755 (+-2.209) | 0.998 (+-0.000)
3 torch.float32 channels_last bilinear (224, 224) -> (270, 268) aa=True | | 1273.656 (+-8.191) | 1273.996 (+-6.578) | 1.000 (+-0.000)
3 torch.float32 channels_last bilinear (256, 256) -> (1024, 1024) aa=True | | 13920.875 (+-90.444) | 13959.411 (+-70.105) | 1.003 (+-0.000)
3 torch.float32 channels_last bilinear (224, 224) -> (64, 64) aa=True | | 324.637 (+-1.033) | 322.538 (+-1.938) | 0.994 (+-0.000)
3 torch.float32 channels_last bilinear (270, 268) -> (224, 224) aa=True | | 1100.012 (+-6.499) | 1098.919 (+-7.063) | 0.999 (+-0.000)
3 torch.float32 channels_last bilinear (1024, 1024) -> (256, 256) aa=True | | 6061.335 (+-31.139) | 6032.787 (+-33.749) | 0.995 (+-0.000)
3 torch.float32 channels_last bilinear (64, 64) -> (224, 224) aa=False | | 762.300 (+-2.342) | 763.982 (+-3.370) | 1.002 (+-0.000)
3 torch.float32 channels_last bilinear (224, 224) -> (270, 268) aa=False | | 1093.883 (+-7.295) | 1094.536 (+-8.027) | 1.001 (+-0.000)
3 torch.float32 channels_last bilinear (256, 256) -> (1024, 1024) aa=False | | 15773.643 (+-174.540) | 15761.104 (+-406.822) | 0.999 (+-0.000)
3 torch.float32 channels_last bilinear (224, 224) -> (64, 64) aa=False | | 71.518 (+-1.553) | 70.876 (+-0.218) | 0.991 (+-0.000)
3 torch.float32 channels_last bilinear (270, 268) -> (224, 224) aa=False | | 762.252 (+-3.098) | 760.904 (+-5.392) | 0.998 (+-0.000)
3 torch.float32 channels_last bilinear (1024, 1024) -> (256, 256) aa=False | | 1001.154 (+-3.420) | 1000.705 (+-4.251) | 1.000 (+-0.000)
4 torch.uint8 channels_last bilinear (256, 256) -> (32, 32) aa=True | | 51.957 (+-0.247) | 54.757 (+-0.199) | 1.054 (+-0.000)
4 torch.uint8 channels_last bilinear (256, 256) -> (32, 32) aa=False | | 34.697 (+-0.210) | 34.080 (+-0.174) | 0.982 (+-0.000)
4 torch.uint8 channels_last bilinear (256, 256) -> (224, 224) aa=True | | 144.017 (+-1.476) | 139.483 (+-1.733) | 0.969 (+-0.000)
4 torch.uint8 channels_last bilinear (256, 256) -> (224, 224) aa=False | | 129.584 (+-1.338) | 124.907 (+-1.629) | 0.964 (+-0.000)
4 torch.uint8 channels_last bilinear (256, 256) -> (320, 320) aa=True | | 193.730 (+-1.545) | 185.567 (+-1.365) | 0.958 (+-0.000)
4 torch.uint8 channels_last bilinear (256, 256) -> (320, 320) aa=False | | 190.381 (+-1.394) | 183.485 (+-1.872) | 0.964 (+-0.000)
4 torch.uint8 channels_last bilinear (520, 520) -> (32, 32) aa=True | | 122.909 (+-1.114) | 134.311 (+-1.665) | 1.093 (+-0.000)
4 torch.uint8 channels_last bilinear (520, 520) -> (32, 32) aa=False | | 55.997 (+-0.238) | 54.905 (+-0.237) | 0.981 (+-0.000)
4 torch.uint8 channels_last bilinear (520, 520) -> (224, 224) aa=True | | 285.947 (+-1.955) | 327.057 (+-2.388) | 1.144 (+-0.000)
4 torch.uint8 channels_last bilinear (520, 520) -> (224, 224) aa=False | | 211.831 (+-1.315) | 200.038 (+-1.421) | 0.944 (+-0.000)
4 torch.uint8 channels_last bilinear (712, 712) -> (32, 32) aa=True | | 191.767 (+-2.166) | 207.113 (+-1.347) | 1.080 (+-0.000)
4 torch.uint8 channels_last bilinear (712, 712) -> (32, 32) aa=False | | 73.375 (+-0.224) | 71.915 (+-0.354) | 0.980 (+-0.000)
4 torch.uint8 channels_last bilinear (712, 712) -> (224, 224) aa=True | | 420.096 (+-2.266) | 460.714 (+-2.824) | 1.097 (+-0.000)
4 torch.uint8 channels_last bilinear (712, 712) -> (224, 224) aa=False | | 272.428 (+-2.255) | 253.531 (+-1.630) | 0.931 (+-0.000)
4 torch.uint8 channels_last bilinear (64, 64) -> (224, 224) aa=True | | 74.814 (+-0.361) | 74.655 (+-0.319) | 0.998 (+-0.000)
4 torch.uint8 channels_last bilinear (224, 224) -> (270, 268) aa=True | | 150.514 (+-1.199) | 145.675 (+-1.193) | 0.968 (+-0.000)
4 torch.uint8 channels_last bilinear (256, 256) -> (1024, 1024) aa=True | | 895.240 (+-3.512) | 869.262 (+-2.442) | 0.971 (+-0.000)
4 torch.uint8 channels_last bilinear (224, 224) -> (64, 64) aa=True | | 66.707 (+-0.207) | 70.331 (+-0.370) | 1.054 (+-0.000)
4 torch.uint8 channels_last bilinear (270, 268) -> (224, 224) aa=True | | 154.572 (+-1.198) | 151.285 (+-1.215) | 0.979 (+-0.000)
4 torch.uint8 channels_last bilinear (1024, 1024) -> (256, 256) aa=True | | 667.814 (+-2.682) | 714.797 (+-2.961) | 1.070 (+-0.000)
4 torch.uint8 channels_last bilinear (64, 64) -> (224, 224) aa=False | | 72.933 (+-0.247) | 73.522 (+-0.308) | 1.008 (+-0.000)
4 torch.uint8 channels_last bilinear (224, 224) -> (270, 268) aa=False | | 148.658 (+-1.072) | 144.106 (+-1.085) | 0.969 (+-0.000)
4 torch.uint8 channels_last bilinear (256, 256) -> (1024, 1024) aa=False | | 887.989 (+-4.149) | 871.161 (+-5.846) | 0.981 (+-0.000)
4 torch.uint8 channels_last bilinear (224, 224) -> (64, 64) aa=False | | 48.430 (+-0.190) | 47.099 (+-0.244) | 0.973 (+-0.000)
4 torch.uint8 channels_last bilinear (270, 268) -> (224, 224) aa=False | | 135.406 (+-1.191) | 130.276 (+-1.227) | 0.962 (+-0.000)
4 torch.uint8 channels_last bilinear (1024, 1024) -> (256, 256) aa=False | | 421.934 (+-2.683) | 395.063 (+-2.857) | 0.936 (+-0.000)
3 torch.uint8 channels_first bilinear (256, 256) -> (32, 32) aa=True | 38.256 (+-0.654) | 129.925 (+-3.938) | 133.051 (+-1.148) | 1.024 (+-0.000)
3 torch.uint8 channels_first bilinear (256, 256) -> (32, 32) aa=False | | 112.929 (+-0.949) | 111.913 (+-1.015) | 0.991 (+-0.000)
3 torch.uint8 channels_first bilinear (256, 256) -> (224, 224) aa=True | 127.690 (+-1.114) | 295.840 (+-1.923) | 307.481 (+-4.318) | 1.039 (+-0.000)
3 torch.uint8 channels_first bilinear (256, 256) -> (224, 224) aa=False | | 282.147 (+-2.209) | 292.091 (+-4.447) | 1.035 (+-0.000)
3 torch.uint8 channels_first bilinear (256, 256) -> (320, 320) aa=True | 178.768 (+-1.925) | 426.283 (+-2.372) | 447.584 (+-9.413) | 1.050 (+-0.000)
3 torch.uint8 channels_first bilinear (256, 256) -> (320, 320) aa=False | | 423.556 (+-6.367) | 451.241 (+-8.831) | 1.065 (+-0.000)
3 torch.uint8 channels_first bilinear (520, 520) -> (32, 32) aa=True | 113.096 (+-0.872) | 433.224 (+-2.394) | 443.731 (+-2.782) | 1.024 (+-0.000)
3 torch.uint8 channels_first bilinear (520, 520) -> (32, 32) aa=False | | 364.883 (+-2.438) | 365.349 (+-1.669) | 1.001 (+-0.000)
3 torch.uint8 channels_first bilinear (520, 520) -> (224, 224) aa=True | 281.490 (+-2.263) | 671.439 (+-2.538) | 729.098 (+-5.497) | 1.086 (+-0.000)
3 torch.uint8 channels_first bilinear (520, 520) -> (224, 224) aa=False | | 596.118 (+-2.328) | 601.511 (+-5.652) | 1.009 (+-0.000)
3 torch.uint8 channels_first bilinear (712, 712) -> (32, 32) aa=True | 185.918 (+-1.693) | 773.266 (+-2.932) | 788.039 (+-3.813) | 1.019 (+-0.000)
3 torch.uint8 channels_first bilinear (712, 712) -> (32, 32) aa=False | | 654.207 (+-3.507) | 649.668 (+-2.238) | 0.993 (+-0.000)
3 torch.uint8 channels_first bilinear (712, 712) -> (224, 224) aa=True | 410.836 (+-2.427) | 1076.095 (+-10.237) | 1133.274 (+-14.727) | 1.053 (+-0.000)
3 torch.uint8 channels_first bilinear (712, 712) -> (224, 224) aa=False | | 926.204 (+-13.314) | 929.228 (+-5.367) | 1.003 (+-0.000)
3 torch.uint8 channels_first bilinear (64, 64) -> (224, 224) aa=True | 60.375 (+-0.289) | 157.108 (+-1.625) | 178.420 (+-6.037) | 1.136 (+-0.000)
3 torch.uint8 channels_first bilinear (224, 224) -> (270, 268) aa=True | 132.439 (+-1.120) | 319.372 (+-2.045) | 341.354 (+-6.069) | 1.069 (+-0.000)
3 torch.uint8 channels_first bilinear (256, 256) -> (1024, 1024) aa=True | 944.113 (+-2.681) | 2591.914 (+-38.223) | 2819.866 (+-148.463) | 1.088 (+-0.000)
3 torch.uint8 channels_first bilinear (224, 224) -> (64, 64) aa=True | 52.428 (+-0.214) | 130.977 (+-0.963) | 136.211 (+-1.751) | 1.040 (+-0.000)
3 torch.uint8 channels_first bilinear (270, 268) -> (224, 224) aa=True | 138.626 (+-1.263) | 313.966 (+-2.245) | 337.755 (+-12.921) | 1.076 (+-0.000)
3 torch.uint8 channels_first bilinear (1024, 1024) -> (256, 256) aa=True | 690.426 (+-12.467) | 1979.607 (+-20.048) | 2050.545 (+-23.798) | 1.036 (+-0.000)
3 torch.uint8 channels_first bilinear (64, 64) -> (224, 224) aa=False | | 155.784 (+-1.254) | 178.000 (+-7.008) | 1.143 (+-0.000)
3 torch.uint8 channels_first bilinear (224, 224) -> (270, 268) aa=False | | 316.951 (+-2.666) | 339.688 (+-6.401) | 1.072 (+-0.000)
3 torch.uint8 channels_first bilinear (256, 256) -> (1024, 1024) aa=False | | 2580.831 (+-33.703) | 2943.312 (+-160.264) | 1.140 (+-0.000)
3 torch.uint8 channels_first bilinear (224, 224) -> (64, 64) aa=False | | 111.623 (+-1.048) | 112.409 (+-1.393) | 1.007 (+-0.000)
3 torch.uint8 channels_first bilinear (270, 268) -> (224, 224) aa=False | | 294.691 (+-1.547) | 306.256 (+-4.166) | 1.039 (+-0.000)
3 torch.uint8 channels_first bilinear (1024, 1024) -> (256, 256) aa=False | | 1729.876 (+-14.071) | 1732.068 (+-21.338) | 1.001 (+-0.000)
3 torch.float32 channels_first bilinear (256, 256) -> (32, 32) aa=True | | 305.353 (+-0.805) | 304.727 (+-1.361) | 0.998 (+-0.000)
3 torch.float32 channels_first bilinear (256, 256) -> (32, 32) aa=False | | 153.584 (+-0.698) | 154.150 (+-0.756) | 1.004 (+-0.000)
3 torch.float32 channels_first bilinear (256, 256) -> (224, 224) aa=True | | 621.918 (+-1.981) | 618.706 (+-1.102) | 0.995 (+-0.000)
3 torch.float32 channels_first bilinear (256, 256) -> (224, 224) aa=False | | 1047.633 (+-7.071) | 1049.938 (+-6.541) | 1.002 (+-0.000)
3 torch.float32 channels_first bilinear (256, 256) -> (320, 320) aa=True | | 907.549 (+-1.557) | 903.190 (+-1.170) | 0.995 (+-0.000)
3 torch.float32 channels_first bilinear (256, 256) -> (320, 320) aa=False | | 1996.973 (+-13.645) | 1994.139 (+-11.676) | 0.999 (+-0.000)
3 torch.float32 channels_first bilinear (520, 520) -> (32, 32) aa=True | | 1427.660 (+-7.690) | 1387.957 (+-5.681) | 0.972 (+-0.000)
3 torch.float32 channels_first bilinear (520, 520) -> (32, 32) aa=False | | 517.519 (+-2.274) | 517.975 (+-1.560) | 1.001 (+-0.000)
3 torch.float32 channels_first bilinear (520, 520) -> (224, 224) aa=True | | 1634.611 (+-12.291) | 1686.206 (+-11.108) | 1.032 (+-0.000)
3 torch.float32 channels_first bilinear (520, 520) -> (224, 224) aa=False | | 1414.320 (+-8.553) | 1415.925 (+-13.567) | 1.001 (+-0.000)
3 torch.float32 channels_first bilinear (712, 712) -> (32, 32) aa=True | | 2654.209 (+-10.720) | 2641.642 (+-8.319) | 0.995 (+-0.000)
3 torch.float32 channels_first bilinear (712, 712) -> (32, 32) aa=False | | 941.203 (+-2.867) | 946.496 (+-6.012) | 1.006 (+-0.000)
3 torch.float32 channels_first bilinear (712, 712) -> (224, 224) aa=True | | 2719.297 (+-11.857) | 2744.925 (+-21.621) | 1.009 (+-0.000)
3 torch.float32 channels_first bilinear (712, 712) -> (224, 224) aa=False | | 1933.822 (+-9.872) | 1929.558 (+-13.361) | 0.998 (+-0.000)
3 torch.float32 channels_first bilinear (64, 64) -> (224, 224) aa=True | | 292.733 (+-1.976) | 289.194 (+-0.734) | 0.988 (+-0.000)
3 torch.float32 channels_first bilinear (224, 224) -> (270, 268) aa=True | | 668.280 (+-2.662) | 665.056 (+-2.019) | 0.995 (+-0.000)
3 torch.float32 channels_first bilinear (256, 256) -> (1024, 1024) aa=True | | 5049.993 (+-15.111) | 5045.329 (+-20.733) | 0.999 (+-0.000)
3 torch.float32 channels_first bilinear (224, 224) -> (64, 64) aa=True | | 287.186 (+-1.878) | 290.790 (+-1.505) | 1.013 (+-0.000)
3 torch.float32 channels_first bilinear (270, 268) -> (224, 224) aa=True | | 675.495 (+-1.491) | 672.473 (+-1.233) | 0.996 (+-0.000)
3 torch.float32 channels_first bilinear (1024, 1024) -> (256, 256) aa=True | | 5110.720 (+-21.234) | 5174.141 (+-18.942) | 1.012 (+-0.000)
3 torch.float32 channels_first bilinear (64, 64) -> (224, 224) aa=False | | 939.682 (+-2.535) | 939.575 (+-2.190) | 1.000 (+-0.000)
3 torch.float32 channels_first bilinear (224, 224) -> (270, 268) aa=False | | 1424.547 (+-8.681) | 1424.590 (+-43.012) | 1.000 (+-0.000)
3 torch.float32 channels_first bilinear (256, 256) -> (1024, 1024) aa=False | | 21238.039 (+-834.799) | 21165.362 (+-268.642) | 0.997 (+-0.000)
3 torch.float32 channels_first bilinear (224, 224) -> (64, 64) aa=False | | 182.334 (+-1.036) | 182.746 (+-1.458) | 1.002 (+-0.000)
3 torch.float32 channels_first bilinear (270, 268) -> (224, 224) aa=False | | 1060.442 (+-5.567) | 1059.309 (+-7.065) | 0.999 (+-0.000)
3 torch.float32 channels_first bilinear (1024, 1024) -> (256, 256) aa=False | | 3603.374 (+-14.050) | 3532.147 (+-18.668) | 0.980 (+-0.000)
4 torch.uint8 channels_first bilinear (256, 256) -> (32, 32) aa=True | | 70.337 (+-0.525) | 73.757 (+-0.221) | 1.049 (+-0.000)
4 torch.uint8 channels_first bilinear (256, 256) -> (32, 32) aa=False | | 53.316 (+-0.199) | 53.351 (+-0.232) | 1.001 (+-0.000)
4 torch.uint8 channels_first bilinear (256, 256) -> (224, 224) aa=True | | 250.769 (+-1.181) | 269.099 (+-3.592) | 1.073 (+-0.000)
4 torch.uint8 channels_first bilinear (256, 256) -> (224, 224) aa=False | | 236.643 (+-2.056) | 248.481 (+-5.474) | 1.050 (+-0.000)
4 torch.uint8 channels_first bilinear (256, 256) -> (320, 320) aa=True | | 398.118 (+-2.679) | 410.447 (+-5.681) | 1.031 (+-0.000)
4 torch.uint8 channels_first bilinear (256, 256) -> (320, 320) aa=False | | 394.098 (+-2.674) | 409.395 (+-7.665) | 1.039 (+-0.000)
4 torch.uint8 channels_first bilinear (520, 520) -> (32, 32) aa=True | | 190.539 (+-1.206) | 201.400 (+-1.865) | 1.057 (+-0.000)
4 torch.uint8 channels_first bilinear (520, 520) -> (32, 32) aa=False | | 124.083 (+-0.701) | 123.629 (+-0.985) | 0.996 (+-0.000)
4 torch.uint8 channels_first bilinear (520, 520) -> (224, 224) aa=True | | 441.806 (+-1.937) | 497.115 (+-3.611) | 1.125 (+-0.000)
4 torch.uint8 channels_first bilinear (520, 520) -> (224, 224) aa=False | | 368.386 (+-2.056) | 367.895 (+-3.530) | 0.999 (+-0.000)
4 torch.uint8 channels_first bilinear (712, 712) -> (32, 32) aa=True | | 318.859 (+-1.982) | 337.082 (+-1.912) | 1.057 (+-0.000)
4 torch.uint8 channels_first bilinear (712, 712) -> (32, 32) aa=False | | 198.969 (+-1.168) | 198.284 (+-1.117) | 0.997 (+-0.000)
4 torch.uint8 channels_first bilinear (712, 712) -> (224, 224) aa=True | | 641.299 (+-3.068) | 689.438 (+-4.126) | 1.075 (+-0.000)
4 torch.uint8 channels_first bilinear (712, 712) -> (224, 224) aa=False | | 488.206 (+-2.259) | 514.368 (+-11.502) | 1.054 (+-0.000)
4 torch.uint8 channels_first bilinear (64, 64) -> (224, 224) aa=True | | 167.609 (+-1.254) | 181.377 (+-3.830) | 1.082 (+-0.000)
4 torch.uint8 channels_first bilinear (224, 224) -> (270, 268) aa=True | | 294.348 (+-2.478) | 304.571 (+-4.449) | 1.035 (+-0.000)
4 torch.uint8 channels_first bilinear (256, 256) -> (1024, 1024) aa=True | | 2923.097 (+-18.406) | 3201.017 (+-98.145) | 1.095 (+-0.000)
4 torch.uint8 channels_first bilinear (224, 224) -> (64, 64) aa=True | | 87.094 (+-0.349) | 92.073 (+-0.470) | 1.057 (+-0.000)
4 torch.uint8 channels_first bilinear (270, 268) -> (224, 224) aa=True | | 263.140 (+-1.476) | 283.198 (+-3.817) | 1.076 (+-0.000)
4 torch.uint8 channels_first bilinear (1024, 1024) -> (256, 256) aa=True | | 1251.367 (+-81.495) | 1249.723 (+-34.231) | 0.999 (+-0.000)
4 torch.uint8 channels_first bilinear (64, 64) -> (224, 224) aa=False | | 165.496 (+-1.151) | 177.639 (+-4.100) | 1.073 (+-0.000)
4 torch.uint8 channels_first bilinear (224, 224) -> (270, 268) aa=False | | 291.731 (+-1.534) | 304.440 (+-5.711) | 1.044 (+-0.000)
4 torch.uint8 channels_first bilinear (256, 256) -> (1024, 1024) aa=False | | 2915.894 (+-17.840) | 3241.920 (+-95.356) | 1.112 (+-0.000)
4 torch.uint8 channels_first bilinear (224, 224) -> (64, 64) aa=False | | 68.399 (+-0.315) | 68.497 (+-0.388) | 1.001 (+-0.000)
4 torch.uint8 channels_first bilinear (270, 268) -> (224, 224) aa=False | | 243.896 (+-1.334) | 249.747 (+-3.271) | 1.024 (+-0.000)
4 torch.uint8 channels_first bilinear (1024, 1024) -> (256, 256) aa=False | | 1016.406 (+-33.316) | 929.075 (+-65.485) | 0.914 (+-0.000)
Times are in microseconds (us).
- Rerun suspicious cases
Num threads: 1
PIL version: 9.0.0.post1
[--------------------------------------------------------------------------------- Resize ---------------------------------------------------------------------------------]
| torch (2.1.0a0+git7b834d6) ghstack-PR-96847 | torchvision resize | Pillow (9.0.0.post1)
1 threads: -----------------------------------------------------------------------------------------------------------------------------------------------------------------
3 torch.uint8 channels_last bilinear (520, 520) -> (32, 32) aa=False | 369.551 (+-10.291) | 191.607 (+-0.566) |
3 torch.uint8 channels_last bilinear (520, 520) -> (224, 224) aa=True | 673.710 (+-13.528) | 2347.606 (+-38.950) | 281.294 (+-1.939)
3 torch.uint8 channels_last bilinear (712, 712) -> (32, 32) aa=False | 655.299 (+-8.068) | 322.105 (+-2.164) |
3 torch.uint8 channels_last bilinear (712, 712) -> (224, 224) aa=True | 1074.346 (+-11.005) | 3589.572 (+-13.669) | 410.919 (+-2.393)
[----------------------------------------------------------------------------- Resize ----------------------------------------------------------------------------]
| torch (2.1.0a0+git5309c44) nightly | torchvision resize | Pillow (9.0.0.post1)
1 threads: --------------------------------------------------------------------------------------------------------------------------------------------------------
3 torch.uint8 channels_last bilinear (520, 520) -> (32, 32) aa=False | 363.553 (+-2.929) | 188.454 (+-0.671) |
3 torch.uint8 channels_last bilinear (520, 520) -> (224, 224) aa=True | 719.061 (+-4.441) | 2318.282 (+-65.772) | 282.557 (+-2.095)
3 torch.uint8 channels_last bilinear (712, 712) -> (32, 32) aa=False | 678.180 (+-35.905) | 319.072 (+-1.148) |
3 torch.uint8 channels_last bilinear (712, 712) -> (224, 224) aa=True | 1124.645 (+-16.315) | 3578.279 (+-88.949) | 407.529 (+-2.543)
Times are in microseconds (us).
Merge started
Your change will be merged once all checks pass (ETA 0-4 Hours).
Learn more about merging in the wiki.
Questions? Feedback? Please reach out to the PyTorch DevX TeamAdvanced Debugging
Check the merge workflow status
here
