GitLab

* [Windows] Workaround for loading bundled DLLs

Python-3.8+ adds `add_dll_directory` call, see https://docs.python.org/3/whatsnew/3.8.html#ctypes
Simulate this behaviour on older versions of Python runtime by calling
`LoadLibraryExW` with the appropriate flags

Fixes https://github.com/pytorch/vision/issues/4787

As titled.

* Add API usage calls to utils

* Update to the new api

* add from_checkpoint method for vit

* remove useless change

* Making interpolate_embeddings a utility function

* remove logging

* fix type hint

* fix return type check

* ad  retuurns in docsting & unify type hint

* remove useless import

* fix issue: 'type' object is not subscriptable

* Fixing typing issues

* Making interpolation mode configurable

* formatting
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Add CUDA-11.5 builds to torchvision

* Try to fix cudatoolkit issue

* Trying to use cuda11.5 instead of the toolkit

* Trying to rollback last two changes

* Testing

* Using conda-forge only for cuda115 builds

* Fix typo

* revamp log api usage method

Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* support random seed used to shuffle the sampler

* fix bug for RA samper
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Adding test that verifies meta schema.

* Adding missing interpolation to keypoint model.

* Change assertion.

* Adding map metric in meta schema for detection and renaming metrics for kp and mask rcnn.

* add download functionality to prototype datasets

* fix annotation

* fix test

* remove iopath

* add comments

* Revert "[FBcode->GH] remove unused requests functionality (#5014)"

This reverts commit 33123bee.

* Revert "replace requests with urllib (#4973)"

This reverts commit 8d25de7b.

* add requests as hard dependency

* install library stubs in CI

* fix syntax

* add requests to conda dependencies

* fix mypy CI

* Adding repaeted data-augument sampler

* rebase on top of latest main

* fix formatting

* rename file

* adding coode source

* Adding logging calls for raft and vit

* Linter fix

* simplify model builders

* cleanup

* refactor kwonly to pos or kw handling

* put weight verification back

* revert num categories checks

* fix default weights

* cleanup

* remove manual parameter map

* refactor decorator interface

* address review comments

* cleanup

* refactor callable default

* fix type annotation

* process ungrouped models

* cleanup

* mroe cleanup

* use decorator for detection models

* add decorator for quantization models

* add decorator for segmentation  models

* add decorator for video  models

* remove old helpers

* fix resnet50

* Adding verification back on InceptionV3

* Add kwargs in DeeplabeV3

* Add kwargs on FCN

* Fix typing on Deeplab

* Fix typing on FCN
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* adding PR template

* update

* Update PULL_REQUEST_TEMPLATE.md

* Update PULL_REQUEST_TEMPLATE.md

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

* exclude windows from mmap

* add comment for windows

* appease mypy

* support amp training for video classification models

* Removed extra empty line and used scaler instead of args.amp as function argument

* apply formating to pass lint tests
Co-authored-by: Konstantinos Bozas <kbz@kbz-mbp.broadband>

According to the benchmark link https://github.com/pytorch/vision/issues/1311#issuecomment-781329339

, for GPU the threshold should be higher (which is why detectron2 used 40k).
This PR changes the threshold to be device dependent.
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Francisco Massa <fvsmassa@gmail.com>