GitLab

* register deconv in CONV_LAYERS

* use ConvTranspose2d implemented in MMCV

* remove repetitive register_module

* update

* add unittest for deconv

* Support to specify LR of DCN's conv_offset

* Resolve comments & add unit test

* Resolve formats

* Fix CI for DCN

* Mock DCN when cpu only

* Use mock for cpu testing

* Fix docstring and support ModulatedDCN

* set offset_lr_mult as dcn's arguments, link CU-49u01p

* fix lr bug

* fall back to set LR in constructor

* resolve comments

* add torchvision roi_align with aligned=True

* fix for lint test

* fix for lint test2

* format use yapf

* Add build_runner

* Parametrize test_runner

* Add imports to runner __init__

* Refactor max_iters and max_epochs from run to init

* Add assertion error messages

* Add test_builder

* Make change retro-compatible

* Raise ValueError if max_epochs and max_iters

* merge train/val_step and batch_processor into run_iter

* make self.train_mode of runner as an argument

* remove abstract methods of run_iter() in base_runner

* Add test case for type defined using default_args

* Refactor build_from_cfg

* Update exception of missing type

* pre-commit

* Fix default_args is None

* pre-commit

* Bring back test

* Update exception raising

* use copy instead of symlink on windows

* use platform.system() instead of sys.platform

* add 'wrap_fp16_model' to __inti__.py

* add 'wrap_fp16_model' to __init__.py

* add brightness and contrast augmentation

* remove unnecessary

* reformat

* relax the precision constrain for adjust_brightness aug

* fix percision assertion error in unit test

* remove toy

* rename alpha as factor

* use np.testing.assert_allclose in place of np.less_equal

* Mv wrappers into bricks and use wrappers in registry

* resolve import issues

* fix import issues

* set nn op forward to torch 1.6.1

* fix CI bug and add warning

* Fix CI by using patch mock

* mv warnings inside deprecated module's initialization

* update impad

* fix docstring

* add shape for impad

* fix unit test

* remove old version & fix doc

* fix linting

* fix doc

* add linear decay learning rate scheduler

* fix impad

* fix setup.cfg

* fix linting

* add yapf

* add swish

* fix lr_updater

* fix lr_updater.py

* update swish

* add swish

* fix inplace

* fix typo

* update

* add same padding

* fix docstring

* add unittest

* fix register

* change name
Co-authored-by: lixiaojie <lixiaojie@sensetime.com>

* support multiple interpolation modes for imrotate

* reformat code

* Reordered the hooks and use attributes rather than args.

Formated.

* Reordering may cause conflict, assign the value first than update such as max_iter

Rewind back the order.

* Rewind back to just use attributes, the update of max_iter and stuff will be done in new hooks.

Minor format.

* add equalize augmentation in mmcv

* delete unnecessary

* reformat

* remove clip in implementing equalize, and add uint test with case step=0

* remove clip in implementing equalize, and add uint test with case step=0

* add comments for unit test

* rename function name as imequalize

* add Color augmentation

* reformat

* reformat

* reformat docstring

* reformat docstring

* add more unit test

* add more unit test

* add clip value and uint test for image with type float

* rename function name

* add url support for VideoReader

* add a comment

* add unittest

* use a connectable url for ci

* add imtranslate

* add imtranslate

* update comments

* reformat

* Add windows instruction and fix compilation bug

* reformat codebase

* fix TypeError: Scalar value for argument borderValue is longer than 4

* add comments for imshear with case that border_value has more than 3 elements

* [Feature] Support load models from mmcls

* [Feature] Support load with mmcls://

* hard-code load mmcls

* fixed wrong commit

* add json

* remove cifar

* Add shear augmentation

* Update geometric.py

* Update geometric.py

* [feature] Use cu92 & ubuntu1604 for torch 1.3.1

* [fix]Use ubuntu-16.04 rather than latest for cu92

* [fix]: soft link cuda path

* Debug cu101 bug

* Update usr/include for cublas with cu101

* Check default nvcc version

* rm :

* add cuda root

* Try to locate cublas_v2.h

* Set cublas version

* Force cublas version

* check usr/local/cuda-10.2

* Add tree

* cp cublas files

* cp cublas files to lib64

* do not tree cuda

* fix path cp error

* cp dir

* Keep using CUDA10.1 only

* recover empty line