GitLab

* Fix extensions in wheels build

* pass pre-script from caller

* add pre-script to run relocate on wheel

* install auditwheel for relocate

* reorder numpy install

* use jpeg80 for py3.9 and py3.10 mac builds

* revert back to jpeg9

* printing the found jpeg libs during build time and reintro circle for comparison

* more debug logs, remove extra linux builds

* typo fix

* brew uninstall conflicting jpeg libraries on mac machines

* fix path for unittests

* ufmt

* [Nova] Disable Linux Wheels Builds from CircleCI

* remove old comment

* remove prototype datasets from CI

* move encoded features to prototype datasets namespace

* remove decoding transforms

* [REVERT ME] reinstate prototype datasets CI

* Revert "[REVERT ME] reinstate prototype datasets CI"

This reverts commit 215fb185cf6be5be7adf0388116c77acc9a5d3f3.

* [Nova] Caller for Conda M1 builds

* rename job

* [Nova] Add Linux Wheels Build Workflow

* Add smoke tests script

* [Nova] Add Caller workflow for Linux Conda Builds

* Add smoke tests script

* [Nova] GHA Linux GPU Job

* increase timeout since jobs timeout and cancel after 30 mins

* bigger instance and longer timeout

* use instance with more gpu memory and only run py38 and cu116 on PR CI

This reverts commit d0de55db.

* [proto][WIP] Enable GPU tests on prototype

* Update prototype-tests.yml

* tests on gpu as separate file

* Removed matrix setup

* Update prototype-tests-gpu.yml

* Update prototype-tests-gpu.yml

* Added --gpus=all flag

* Added xfail for cuda vs cpu tolerance issue

* Update prototype-tests-gpu.yml

* [Nova] Migrate Linux CPU job to Generic Job

* branch ref for composite action job

* move checkout step to separate job

* added runs-on

* nit fixes

* no need to run conda sheel script thing

* Channel is set inside the script

* add remaining env vars

* nit env var fix

* cleanup

* simplify unneeded jobs

* name of the conda env should be the path

* remove main ref to use PR

* pin pyav to <10

* pin av in GHA workflows as well

* also pin in M1 workflow

* [Nova][WIP] Add Linux CPU Unittests for Torchvision

* use conda-builder image since conda installation is needed

* install torch dep with conda instead

* use circleCI command to run tests

* larger instance to avoid OOM issues

* proper syntax for self-hosted runners

* 4xlarge instance

* 8xlarge

* 12xlarge

* use setup-miniconda job

* add back PATH change to help setup py detect conda

* run conda shell script

* install other deps up front

* git config and undo path change

* revert to local conda install

* conda-builder image

* support for whole python version matrix

* clean up the conda env once we are done with the job

* prevent feature wrapping if the feature is not the primary operand

* explicitly add feature tests to CI

* add coverage to prototype tests CI

* line continuation

* reduce verbosity of prototype tests and split them into multiple steps

* rename builtin datasets module to fit common scheme

* shlex -> shutil

* [SKIP CI]

* [skip ci]
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Added mid-level ops and feature-based ops

* Fixing deadlock in dataloader with circular imports

* Added non-scalar fill support workaround for pad

* Removed comments

* int/float support for fill in pad op

* Updated type hints and removed bypass option from mid-level methods

* Minor nit fixes

* Make sure we are building against test chanell for release

* Cleanup

* Cleanup

* Refactoring logic

* Remove tagged tests

s#ref/heads/#refs/heads/#

I should have noticed it while copy-n-pasting the condition. Unfortunately there are no way to test is other than in prod, but nightly builds are still not getting pushed, see https://github.com/pytorch/vision/runs/6860407007?check_suite_focus=true for example

Replace `steps.extract_branch.outputs.branch` (which were probably taken from
https://stackoverflow.com/questions/58033366/how-to-get-the-current-branch-within-github-actions ) with straightforward `github.event.ref`

Tested in
https://github.com/malfet/deleteme/runs/6822647720?check_suite_focus=true and https://github.com/malfet/deleteme/runs/6822691158?check_suite_focus=true

* Adding tagged builds

* Testing

* Testing

* Testing

* Testing

* Adding conda builds

By passing `--pre` option to `pip install`, otherwise torchvision were always tested against last PyTorch release

Clean up Conda build folder before every run
Enable artifact upload to GitHub for every workflow run, but upload to Conda/S3 only on nightly pushes

Test plan: `conda install -c pytorch-nightly torchvision; python -c "import torchvision;print(torchvision.io.read_image('hummingbird.jpg').shape)"`

* Add M1 testing job

* libjpeg -> jpeg<=9b in test-m1.yml

* Added export PATH=~/miniconda3/bin... from 6122

* Tests were OK, let's see if we can remove the pinning

Explicitly set PATH to point to `conda` binary, otherwise libjpeg detection logic does not work
Pin libjpeg to the same version on x86 and m1
Add simple tests that jpeg can be decoded by a generated wheel