GitLab

There is a mistake in FGVCAircraft dataset documentation. If you load dataset and sum up 'trainval' and 'test' splits, there are not 10 200, but 10 000 images. Also mistake is in the number of classes: if we choose `annotation_level` to be `variant`, then unique number of classes is going to be not 102, but 100, and with `manufacturer` level it is not 41, but 30.
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Add SWAG model that only the linear classifier head is finetuned with frozen trunk weight

* Add accuracy from experiments

* Change name from SWAG_LC to SWAG_LINEAR

* Add comment on SWAG_LINEAR weight

* Remove the comment docs (moved to PR description), and add the PR url as recipe. Also change name of previous swag model to SWAG_E2E_V1

* Update _register_onnx_ops.py

* Update misc.py

* [proto] Added crop_bounding_box op

* Removed "pass"

* Updated comment

* Removed unused args from signature

* refactor prototype datasets to inherit from IterDataPipe (#5448)

* refactor prototype datasets to inherit from IterDataPipe

* depend on new architecture

* fix missing file detection

* remove unrelated file

* reinstante decorator for mock registering

* options -> config

* remove passing of info to mock data functions

* refactor categories file generation

* fix imagenet

* fix prototype datasets data loading tests (#5711)

* reenable serialization test

* cleanup

* fix dill test

* trigger CI

* patch DILL_AVAILABLE for pickle serialization

* revert CI changes

* remove dill test and traversable test

* add data loader test

* parametrize over only_datapipe

* draw one sample rather than exhaust data loader

* cleanup

* trigger CI

* migrate VOC prototype dataset (#5743)

* migrate VOC prototype dataset

* cleanup

* revert unrelated mock data changes

* remove categories annotations

* move properties to constructor

* readd homepage

* migrate CIFAR prototype datasets (#5751)

* migrate country211 prototype dataset (#5753)

* migrate CLEVR prototype datsaet (#5752)

* migrate coco prototype (#5473)

* migrate coco prototype

* revert unrelated change

* add kwargs to super constructor call

* remove unneeded changes

* fix docstring position

* make kwargs explicit

* add dependencies to docstring

* fix missing dependency message

* Migrate PCAM prototype dataset (#5745)

* Port PCAM

* skip_integrity_check

* Update torchvision/prototype/datasets/_builtin/pcam.py
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* Address comments
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* Migrate DTD prototype dataset (#5757)

* Migrate DTD prototype dataset

* Docstring

* Apply suggestions from code review
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* Migrate GTSRB prototype dataset (#5746)

* Migrate GTSRB prototype dataset

* ufmt

* Address comments

* Apparently mypy doesn't know that __len__ returns ints. How cute.

* why is the CI not triggered??

* Update torchvision/prototype/datasets/_builtin/gtsrb.py
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* migrate CelebA prototype dataset (#5750)

* migrate CelebA prototype dataset

* inline split_id

* Migrate Food101 prototype dataset (#5758)

* Migrate Food101 dataset

* Added length

* Update torchvision/prototype/datasets/_builtin/food101.py
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* Migrate Fer2013 prototype dataset (#5759)

* Migrate Fer2013 prototype dataset

* Update torchvision/prototype/datasets/_builtin/fer2013.py
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* Migrate EuroSAT prototype dataset (#5760)

* Migrate Semeion prototype dataset (#5761)

* migrate caltech prototype datasets (#5749)

* migrate caltech prototype datasets

* resolve third party dependencies

* Migrate Oxford Pets prototype dataset (#5764)

* Migrate Oxford Pets prototype dataset

* Update torchvision/prototype/datasets/_builtin/oxford_iiit_pet.py
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* migrate mnist prototype datasets (#5480)

* migrate MNIST prototype datasets

* Update torchvision/prototype/datasets/_builtin/mnist.py
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Migrate Stanford Cars prototype dataset (#5767)

* Migrate Stanford Cars prototype dataset

* Address comments

* fix category file generation (#5770)

* fix category file generation

* revert unrelated change

* revert unrelated change

* migrate cub200 prototype dataset (#5765)

* migrate cub200 prototype dataset

* address comments

* fix category-file-generation

* Migrate USPS prototype dataset (#5771)

* migrate SBD prototype dataset (#5772)

* migrate SBD prototype dataset

* reuse categories

* Migrate SVHN prototype dataset (#5769)

* add test to enforce __len__ is working on prototype datasets (#5742)

* reactivate special dataset tests

* add missing annotation

* Cleanup prototype dataset implementation (#5774)

* Remove Dataset2 class

* Move read_categories_file out of DatasetInfo

* Remove FrozenBunch and FrozenMapping

* Remove test_prototype_datasets_api.py and move missing dep test somewhere else

* ufmt

* Let read_categories_file accept names instead of paths

* Mypy

* flake8

* fix category file reading
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* update prototype dataset README (#5777)

* update prototype dataset README

* fix header level

* Apply suggestions from code review
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Add MaskRCNN improved weights

* Adding recipe URL

* Add FasterRCNN improved weights

* Add recipe URL

* Update publication_year field

* Add RetinaNet improved weights

* Add weights.

* Change publication date.

* Use frozen BN only if pre-trained.

* Add LSJ and ability to from scratch training.

* Fixing formatter

* Adding `--opt` and `--norm-weight-decay` support in Detection.

* Fix error message

* Make ScaleJitter proportional.

* Adding more norm layers in split_normalization_params.

* Add FixedSizeCrop

* Temporary fix for fill values on PIL

* Fix the bug on fill.

* Add RandomShortestSize.

* Skip resize when an augmentation method is used.

* multiscale in [480, 800]

* Add missing star

* Add new RetinaNet variant.

* Add tests.

* Update expected file for old retina

* Fixing tests

* Add FrozenBN to retinav2

* Fix network initialization issues

* Adding BN support in MaskRCNNHeads and FPN

* Adding support of FasterRCNNHeads

* Introduce norm_layers in backbone utils.

* Bigger RPN head + 2x rcnn v2 models.

* Adding gIoU support to retinanet

* Fix assert

* Add back nesterov momentum

* Rename and extend `FastRCNNConvFCHead` to support arbitrary FCs

* Fix linter

* Add vit_b_16_swag

* Better handling idiom for image_size, edit test_extended_model to handle case where number of param differ from default due to different image size input

* Update the accuracy to the experiment result on torchvision model

* Fix typo missing underscore

* raise exception instead of torch._assert, add back publication year (accidentally deleted)

* Add license information on meta and readme

* Improve wording and fix typo for pretrained model license in readme

* Add vit_l_16 weight

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

* Update the accuracy meta on vit_l_16_swag model to result from our experiment

* Add vit_h_14_swag model

* Add accuracy from experiments

* Add to vit_h_16 model to hubconf.py

* Add docs and expected pkl file for test

* Remove legacy compatibility for ViT_H_14 model
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Test vit_h_14 with smaller image_size to speedup the test
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* port image type conversion transforms to prototype API

* implement proposal for image type conversion

* add deprecation warnings

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

* Add regnet_y_128gh_swag weight

* Add default weight for regnet_y_128gf

* Add the accuracy from experiments

* Added functional affine_bounding_box op with tests

* Updated comments and added another test case

* Update _geometry.py

* Added affine_segmentation_mask with tests

* Fixed device mismatch issue
Added a cude/cpu test
Reduced the number of test samples

* Added test_correctness_affine_segmentation_mask_on_fixed_input

* Updates according to the review

* Replaced [None, ...] by [None, :]

* Adressed review comments

* Fixed formatting and more updates according to the review

* Fixed bad merge

* WIP

* Fixed tests

* Updated warning message

* [FEAT] Start implementing Food101 using the new datasets API. WIP.

* [FEAT] Generate Food101 categories and start the test mock.

* [FEAT] food101 dataset code seems to work now.

* [TEST] food101 mock update.

* [FIX] Some fixes thanks to running food101 tests.

* [FIX] Fix mypy checks for the food101 file.

* [FIX] Remove unused numpy.

* [FIX] Some changes thanks to code review.

* [ENH] More idomatic dataset code thanks to code review.

* [FIX] Remove unused cast.

* [ENH] Set decompress and extract to True for some performance gains.

* [FEAT] Use the preprocess=decompress keyword.

* [ENH] Use the train and test.txt file instead of the .json variants and simplify code + update mock data.

* [ENH] Better food101 mock data generation.

* [FIX] Remove a useless print.
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* port Grayccale and RandomGrayscale transforms to prototype API

* revert some changes

* move gray output channels to transform rather than functional

* indicate md5 checksum is not used for security

* add explanation

* Add regnet_y_16gf and regnet_y_32gf from swag

* Format with ufmt

* Add the experiment accuracy
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* support confirming no virus scan on GDrive download

* put gen_bar_updater back

* Update torchvision/datasets/utils.py
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Fix shape mismatch error

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

* Add vit_b_16_swag

* Better handling idiom for image_size, edit test_extended_model to handle case where number of param differ from default due to different image size input

* Update the accuracy to the experiment result on torchvision model

* Fix typo missing underscore

* raise exception instead of torch._assert, add back publication year (accidentally deleted)

* Add license information on meta and readme

* Improve wording and fix typo for pretrained model license in readme

* Add vit_l_16 weight

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

* Update the accuracy meta on vit_l_16_swag model to result from our experiment
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* Detection recipe enhancements

* Add back nesterov momentum

* update hint_shuffling

* trigger CI

* Update docstring of adjust_hue
Signed-off-by: Behrooz <3968947+drbeh@users.noreply.github.com>

* Update docstring of ColorJitter
Signed-off-by: Behrooz <3968947+drbeh@users.noreply.github.com>

* Minor corrections
Signed-off-by: Behrooz <3968947+drbeh@users.noreply.github.com>

* reverting some recently introduced exceptions

* Update torchvision/ops/poolers.py
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* address PR comments

* replace one more assert with torch._assert:

* address PR comments

* make type checker happy

* Fix bug

* fix bug

* fix for wrong asserts

* attempt to make tests pass

* Fix test_ops tests

* Fix expected exception in tests

* fix typo

* fix tests and format

* fix flake8

* remove one last exception

* fix error

* remove unused immport

* replace fake returns by else
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* enforce shuffling before sharding

* revert test changes and add comment

* call _upcast to consider overflow

giou loss is weak at overflow problem because it computes area of box.

* cast datatype to float

* add ":" to if

* lint Test for membership should be 'not in' (E713)

* fix for FFMPEG 5.0

* Attempt to fix memory leak

* early stop when mem alloc fails

* fix c lint error

* fix bug

* revert changes

* remove incorrect av_packet_free

* add `av_packet_free` to subtitle stream

* Addressing subtitle stream comments

* addressing decoder changes

* nit

* addressing datumbox concernsz

* averror push

* addressing comments

* Apply suggestions from code review

* add new error code to documentation

* re-introducing timeout

* fix c++ syntax
Co-authored-by: Bruno Korbar <bkorbar@quansight.com>
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>
Co-authored-by: Joao Gomes <jdsgomes@fb.com>
Co-authored-by: Joao Gomes <joaopsgomes@gmail.com>

* fix PCAM prototype dataset

* update Zip and Tar archive loader datapipes

* only fail on warnings from the pytorch ecosystem

* Revert "only fail on warnings from the pytorch ecosystem"

This reverts commit 2bf3aa6f2d875a4055f7f3ed0b468316fc60d4f4.

* Fixed rotate with expand inconsistency between torch vs PIL on odd-sized images

* Update functional_tensor.py

* added usps dataset

* fixed type issues

* fix mobilnet norm layer test (#5643)

* xfail mobilnet norm layer test

* fix test

* More robust check in tests for 16 bits images (#5652)

* Prefer nvidia channel for conda builds (#5648)

To mitigate missing `libcupti.so` dependency

* fix torchdata CI installation (#5657)

* update urls for kinetics dataset (#5578)

* update urls for kinetics dataset

* update urls for kinetics dataset

* remove errors

* update the changes and add test option to split

* added test to valid values for split arg

* change .txt to .csv for annotation url of k600
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Port Multi-weight support from prototype to main (#5618)

* Moving basefiles outside of prototype and porting Alexnet, ConvNext, Densenet and EfficientNet.

* Porting googlenet

* Porting inception

* Porting mnasnet

* Porting mobilenetv2

* Porting mobilenetv3

* Porting regnet

* Porting resnet

* Porting shufflenetv2

* Porting squeezenet

* Porting vgg

* Porting vit

* Fix docstrings

* Fixing imports

* Adding missing import

* Fix mobilenet imports

* Fix tests

* Fix prototype tests

* Exclude get_weight from models on test

* Fix init files

* Porting googlenet

* Porting inception

* porting mobilenetv2

* porting mobilenetv3

* porting resnet

* porting shufflenetv2

* Fix test and linter

* Fixing docs.

* Porting Detection models (#5617)

* fix inits

* fix docs

* Port faster_rcnn

* Port fcos

* Port keypoint_rcnn

* Port mask_rcnn

* Port retinanet

* Port ssd

* Port ssdlite

* Fix linter

* Fixing tests

* Fixing tests

* Fixing vgg test

* Porting Optical Flow, Segmentation, Video models (#5619)

* Porting raft

* Porting video resnet

* Porting deeplabv3

* Porting fcn and lraspp

* Fixing the tests and linter

* Porting docs, examples, tutorials and galleries (#5620)

* Fix examples, tutorials and gallery

* Update gallery/plot_optical_flow.py
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Fix import

* Revert hardcoded normalization

* fix uncommitted changes

* Fix bug

* Fix more bugs

* Making resize optional for segmentation

* Fixing preset

* Fix mypy

* Fixing documentation strings

* Fix flake8

* minor refactoring
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>

* Resolve conflict

* Porting model tests (#5622)

* Porting tests

* Remove unnecessary variable

* Fix linter

* Move prototype to extended tests

* Fix download models job

* Update CI on Multiweight branch to use the new weight download approach (#5628)

* port Pad to prototype transforms (#5621)

* port Pad to prototype transforms

* use literal

* Bump up LibTorchvision version number for Podspec to release Cocoapods (#5624)
Co-authored-by: Anton Thomma <anton@pri.co.nz>
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

* pre-download model weights in CI docs build (#5625)

* pre-download model weights in CI docs build

* move changes into template

* change docs image

* Regenerated config.yml
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Anton Thomma <11010310+thommaa@users.noreply.github.com>
Co-authored-by: Anton Thomma <anton@pri.co.nz>

* Porting reference scripts and updating presets (#5629)

* Making _preset.py classes

* Remove support of targets on presets.

* Rewriting the video preset

* Adding tests to check that the bundled transforms are JIT scriptable

* Rename all presets from *Eval to *Inference

* Minor refactoring

* Remove --prototype and --pretrained from reference scripts

* remove  pretained_backbone refs

* Corrections and simplifications

* Fixing bug

* Fixing linter

* Fix flake8

* restore documentation example

* minor fixes

* fix optical flow missing param

* Fixing commands

* Adding weights_backbone support in detection and segmentation

* Updating the commands for InceptionV3

* Setting `weights_backbone` to its fully BC value (#5653)

* Replace default `weights_backbone=None` with its BC values.

* Fixing tests

* Fix linter

* Update docs.

* Update preprocessing on reference scripts.

* Change qat/ptq to their full values.

* Refactoring preprocessing

* Fix video preset

* No initialization on VGG if pretrained

* Fix warning messages for backbone utils.

* Adding star to all preset constructors.

* Fix mypy.
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Anton Thomma <11010310+thommaa@users.noreply.github.com>
Co-authored-by: Anton Thomma <anton@pri.co.nz>

* Apply suggestions from code review
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* use decompressor for extracting bz2

* Apply suggestions from code review
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* Apply suggestions from code review
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* fixed lint fails

* added tests for USPS

* check image shape

* fix tests

* check shape on image directly

* Apply suggestions from code review
Co-authored-by: Philip Meier <github.pmeier@posteo.de>

* removed test and comments

* Update test/test_prototype_builtin_datasets.py
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Co-authored-by: Nicolas Hug <contact@nicolas-hug.com>
Co-authored-by: Nikita Shulga <nshulga@fb.com>
Co-authored-by: Sahil Goyal <sahilgoyal6801@gmail.com>
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>
Co-authored-by: Anton Thomma <11010310+thommaa@users.noreply.github.com>
Co-authored-by: Anton Thomma <anton@pri.co.nz>