GitLab

* enable cuda for unittest

* add one check

* rename

* minor fix

* add more check

* add cuda envs

* add more cuda envs

* add driver update

* update vs integration

* get cuda version

* merge

* add envs in install.sh

* reduce some change and add comments

* minor improve

* fix typo

* exit if driver update failed

* fix lint

* fix lint

* Avoid catching exception to show error message

* try 11.1

* try 11.1

* "try 11.1"

* set the downloaded gpu driver dlls only for 10.2

* check torch.cuda.is_available()

* add display.driver

* Revert "Avoid catching exception to show error message"

This reverts commit 1ce58c50de9fefcc3438d29ff398b941f9b6a757.

* Update .circleci/unittest/windows/scripts/set_cuda_envs.sh

* Update .circleci/unittest/windows/scripts/set_cuda_envs.sh

* Update .circleci/unittest/windows/scripts/set_cuda_envs.sh
Co-authored-by: Nicolas Hug <nicolashug@fb.com>
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>
Co-authored-by: Nikita Shulga <nikita.shulga@gmail.com>
Co-authored-by: Nikita Shulga <nshulga@fb.com>

`import torchvision` fails if numpy is not installed, so it should have
an explicit dependency on it (which is already the case for wheel
package)

* add cuda envs

* packaging doesn't need cuda env now

* fix cmake_windows_gpu

* minor fix

* rename suffix

* minor change

* update vs integration

* if there's dll, overwrite it

* temp change

* revert last
Co-authored-by: Nikita Shulga <nshulga@fb.com>

* add pre-commit hooks

* ignore yamls in packaging/*

* add pre-commit to contributing guide lines

* Update CONTRIBUTING.md
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* remove some hooks

* fix docstrings

* fix end of files
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Add deprecation warnings on vision::models

* Change the C++ example.

* Chage readme.

* Update deprecation warning.

* Unpinned the libjpeg version and fixed jpeg_mem_dest's size type Windows BC issue

* Temporarily get back pinned jpeg lib in order to run tests on CircleCI

* Revert "Temporarily get back pinned jpeg lib in order to run tests on CircleCI"

This reverts commit ab18a354150f750256b61911e54aa48621ef24cd.

* Used using instead of typedef and Fixed comment typo
Co-authored-by: Zhongkai Zhu <zhongkai@fb.com>
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Add CUDA-11.3 builds to torchvision

Also mark  `.cicrleci/config.yml` as autogenerated in `.gitattributes`

* Add CUDA-11.3 install steps for Windows

* And here as well

* fix MacOS cmake workflow

* try only mkl

* only pin mkl on MacOs

* fix
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

Co-authored-by: Nicolas Hug <nicolashug@fb.com>

* Do not install lame when building torchvision

It is needed only for ffmpeg integration, which is disabled anyway

* Also upgrade setuptool

To update pkg_resources.packaging to version 20.0+

* packaging: Remove special versioning priveleges

Adds a version suffix for CUDA 10.2 for ease of installing when
installing from download.pytorch.org
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

* also include windows cmath fix
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

* remove comment
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

* fix indent
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

* actually add the script
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

* Fix inconsistent version pinning across yaml files.

* Pin pillow to >=5.3.0

* Pin pillow>=5.3.0 everywhere

* hello

* Try with DYLD_FALLBACK_LIBRARY_PATH
Co-authored-by: Edgar Andrés Margffoy Tuay <andfoy@gmail.com>

* packaging: Remove pin for jpeg, numpy

These may no longer be necessary due to the default anaconda channel
having the necessary packages now.
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

* Update packaging/torchvision/meta.yaml
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

* remove some files

* Also remove entire packaging/conda folder
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

* trying stuff

* Put flags back?

* remove second one just to see

* It worked but it's because it's not built by CI. Trying another one

* Try using TORCH_CUDA_ARCH_LIST instead

* oops

* Add new env variable to build/script_env

* set TORCH_CUDA_ARCH_LIST for the rest of the CUDA versions

* don't pass NVCC_FLAGS venv to conda, let's see if it works
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

* add ROCm binary wheels

* fix lint error
Co-authored-by: Eli Uriegas <1700823+seemethere@users.noreply.github.com>
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

(cherry picked from commit 2f40a483d73018ae6e1488a484c5927f2b309969)
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

* Speedup test_ucf101 (#2623 

* Speedup Cmake builds (#3186)

* Speedup test_autoaugment (#3190)

* Speedup DeformConvTester (#3191)

* Speedup InceptionV3 and GoogleNet on Windows (#3196)

Replace tabs with spaces, add newlines to files and replace whitelist with allowlist

This was a mistake since version suffix should always be appended
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
(cherry picked from commit 6631b74db7cf8aebee8af196b4c65b1fce220baa)
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

ROCM builds were being picked up for CUDA 10.2 builds since they do not
specify a version suffix. We can utilize the CU_VERSION to be specific
about which index to actually pull from.
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>

* Fix external DLL loading on wheels

* Use SetDefaultDllDirectoriess and AddDllDirectory

* Add previous paths

* Trigger debug

* Do not call SetDefaultDllDirectories.

* Do not call loadlibrary if the extensions were not compiled

* Fix lint issues

* [DEBUG] Check wheel relocation issues

* Call delocate on Mac

* Use yum instead of conda

* Do not copy ffmpeg dylibs

* Linux dry run

* Do not download FFmpeg on Mac

* Install bzip2

* Restore FFmpeg on Windows

* Remove ffmpeg temporarily

* Do not remove dependencies

* Disable FFmpeg temporarily on Linux wheels

* Test relocation in Linux

* Add docstring

* Call relocation script

* Minor error correction

* Import auditwheel only on Linux

* Restore ffmpeg again on Windows

* Return *device

* Try fix Windows

* Fix clang-format

* Start windows patchwork

* Copy all DLLs

* Disable FFmpeg on Windows for now

* Bugfix
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

* Add PyTorch FFmpeg to wheel and conda distributions

* Try to install wget from conda

* Add yq flag on Mac

* Correct copy instructions

* Use cURL on Windows

* Call bzip2 directly due to msys2/MSYS2-packages#1548

* Copy ffmpeg binaries to system-wide directories

* Try to use std:c++17 on Windows

* Try to define ssize_t on Windows

* Use C++14

* Declare AVRational structs explicitly

* Initialize AVRational explicitly

* Replace macro to prevent errors on Windows

* Replace AV_TIME_BASE_Q

* Add library paths for video extension

* Force ffmpeg from pytorch channels?

* Fix clang style warnings

* Update CONDA_CHANNEL_FLAGS

* Fix clang style issues

* Update unittest

* Use FFmpeg 4.2

* Install correct version on Mac

* Pin av version to 8.0.0

* Fix string formatting issue

* Fix pip pinning

* Try with 8.0.1

* Use av 8.0.2

* Remove trailling whitespaces

* Disable test_io_opt.py

* Disable test_datasets_video_utils
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>

* Run C++ torchvision example as part of cmake tests

* Enable C++14 support

* Minor error correction

* Add GPU forward on example

* Use arrow operator

* Add CMake build pipeline

* Add CMake build workflow

* Add executable permissions to script

* Install cmake on Windows/MacOS

* Install conda-build before setting up MSVC

* Install PyTorch from nightly

* Do not use conda-build variables

* Add path to CMake

* Install libpng and libjpeg

* Perform make

* Call msbuild on Windows

* Add missing yq

* Use vc_env_helper

* Use string instruction

* Escape configuration option

* Remove configuration flag

* Try to pass -p

* Use caret to escape equal sign

* Escape string option in Windows

* Try to call other bat

* Remove Windows/GPU CMake

* Add tracing cpp test

* Script model instead of tracing it

* Try to register operators manually

* Use manylinux-cuda102

* Activate conda env on Linux

* Build and run sample tracing test

* Add empty echo

* Remove unnecessary register

* Copy headers on Mac

* Revert to 2xlarge

* Include /usr/local/include on Mac

* Install pillow on Windows

* Install future

* Install torchvision on Windows

* Set include flag

* Add torchlib to PATH

* Normalize path via cygpath

* Register ops on Windows

* Minor error correction

* Register CPU/GPU ops on DLL library and register ops via reference

* Install dataclasses

* Install dataclasses using pip

* Address clang formatting issue

* Try to use an actual GPU instance on Linux

* Remove extra environment section

* Declare environment explicitly

* Regenerate

* Pass env variables to Dokcer

* Regenerate circleci

* Test tracing on GPU

* Use GPU medium

* Regenerate

* Use cuda101

* Regenerate

* Do not use pre-trained weights

Avoids having to download pretrained files, which could cause flaky tests
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>