* feature (roi_align_v2): use roi_align v2 to support aligned=True

* fix (setup.py): fix install bug

* recover configs to the original ones

* remove cpu version for now

* rename files

* reformat (roi_align): clean code and reformat

* support legacy roi_align

* support legacy roi_align

* reformat (roi_align): reformat c/cuda/python code

* fix reformat request

* add unsaved change

* pass yapf

* merge roi_align function

* update note

Co-authored-by: beansi <zhangwenwei@sensetime.com>

* Allow find_unused_parameters to be set in config

* Fix typo

* Fix wrong keyword argument 'nms_cfg' in htc

* fix the bug of simple_test with proposals
remove the redundant codes in fast_rcnn

* correct the wrong comment

* minor update

* fix typos

* update fast test pipeline

* fix yapf error

* change fields from tuple to list

Co-authored-by: wangchuanyuan <Trayvon@wangchuanyuandeMacBook.local>
Co-authored-by: Cao Yuhang <yhcao6@gmail.com>

* fix (dcn): fix config bugs and group bugs of dcn

* keep groups in backbone rather than in dcn configs

* fix (htc dcn): rm groups in htc dcn config

* add act_cfg

* add more act_cfg options

* delel GELU

* reorder args, save act_cfg, update doc

* update doc

* add norm_cfg in ResNet pytorch backbone

* delete redundent fields

* fix __repr__ bug of no attribute `contrast_range` in PhotoMetricDistortion

* fix check of too long code line

* remove trailing whitespace

* Fixed wrong attribute name

* Update transforms.py

* add affine_grid

* missing setup

* remove import

* reformat

* rename and reformat

* reformat cpp

* reimplement cityscapes

* fixed gt bbox mode

* convert cityscapes to coco style, add cityscapes eval

* add cityscapes convert script

* add doc

* Update INSTALL.md

* Update INSTALL.md

* update fater rcnn

* fix cityscapes eval

* support format only in cityscapes

* add docs

* remove redundancy

* resolve eval

* update cityscapes md

* more doc and rename

* update doc and cfg

* change to test set

* update pre-commit hook config

* fix styles conflicting with pre-commit hooks

* add grid sampler

* add doc

* remove comment

* fp16_speedup

* speed up fp16 for torch >= 1.3.0

* speed up fp16 for torch >= 1.3.0

* fix trailing whitespace

* no weight decay in depthwise convolution

* fix name of var

* fix train.py:196:80 (99 > 79 characters)

* fix trailing whitespace

* no_wd_in_dw is independ setting now

* add Hellock's suggestions

Co-authored-by: Kai Chen <chenkaidev@gmail.com>

Need to specify --format-only rather than --out in order to actually get .json output

* add optimizer registry

* move under core, add doc

* unexpose TORCH_OPTIMIZERS

* update version to 1.1

* fix typo

* remove the version limit of ubelt

* Fix IOU assigners when ignore_of_thr > 0 and no pred boxes

* remove extra brakets

* fix test linter

* reset worker_seed

* fix isort

* minor fix

* fix comment

* Fix issue with list of metrics in CustomDataset.evaluate

Changes to be committed:
modified:   mmdet/datasets/custom.py

* Updated to evaluate only one of mAP or recall

Changes to be committed:
modified:   mmdet/datasets/custom.py

* add carafe ops

* rename carafe benchmark

* grad check fix

* update grad check

* update grad check output

* add fpn carafe & mask head carafe

* add ReadMe

* update readme

* add carafe setup

* update naive carafe

* update readme and setup

* readme typo fix

* fix flake8 error

* fix flake 8 error

* fix flake 8

* fix flake 8 more

* flake 8 fix plus

* flake 8 fix

* fix flake 8

* reformat ops files

* update fpn files and cfgs

* update readme

* update fcn_mask_head

* update fpn_carafe

* update kernel

* update

* update

* add docstring in FPN_CARAFE

* reformat with yapf

* update

* update

* add build upsampler

* fix mask head build error

* reformat build upsample layer

* add doc string for CARAFE and PixelShuffle

* update

* update upsample_cfg_

* update

* update doc string

* rm abbr in build upsample layer

* update readme

* update model_zoo

* add link to other features in ReadMe

* fix (dpool): directly return empty if rois's length is 0

* fix (dpool): also skip empty when no_trans=True

* reformat (test/test_dpool.py): reformat to pass flake8

* fix (test_dpool): skip gpu test when GPU is not avaliable

* rm (test_dpool.py): rm test_dpool since it is not very necessary in CI

* fix workflow problem

* import sort

* update pipeline

* fix (uint8 index warning): use bool for inside flags

* fix (unint8 index): convert to bool in the final to avoid 1.1 constriant

* fix (unmap): convert inds in unmap to be bool type

* fix roi align device bug

* add device guard to more ops

* nms guard is duplicate