GitLab

* let progress goes to stderr
Signed-off-by: lizz <lizz@sensetime.com>

* Add argument file=sys.stdout
Signed-off-by: lizz <lizz@sensetime.com>

* For the sake of the stupid pytest
Signed-off-by: lizz <lizz@sensetime.com>

* line width < 120 is anti-humane
Signed-off-by: lizz <lizz@sensetime.com>

* flake8
Signed-off-by: lizz <lizz@sensetime.com>

* Lets try
Signed-off-by: lizz <lizz@sensetime.com>

* track progress of iter&enum

* reformat

* reformat with yapf

* add unitest

* add doc, and deprecate track_enum_progress

* update docs & comments

* use torch.utils.tensorboard instead of tensorboardX

* add backward compatibility

* add .idea to gitignore

* resubmit parameter shape check

* check lint

* print a table of mismatched keys

* install cython automatically as a dependency

* install numpy automatically before building

* fix linting errors

* add support for pathlib to load and dump

* minor formatting

* Update lr_updater.py

"Add min_lr to PolyLrUpdaterHook, so that it can anneal to min_lr. When min_lr=0, it behaves as usual. "

* Update lr_updater.py

* add yapf and isort to travis

* minor formatting

* remove blank lines

* skip unit tests for progressbar when python2

* update travis to ubuntu 16.04

* use a newer version ffmpeg

* add -y to add-apt-repository

* rename flowWarp to FlowWarp

* fix the compiling bug when using pip install

* update version to v0.2.10

* fix doc generation

* add docs for optical flow

* fix typo

* add BN pretrain model, add i3d pretrained on K-400

* fix flake8

Signed-off-by: lizz <363664470@qq.com>

* Make latest symlink relative. Add tests

* Skip test_save_checkpoint in the absense of torch

* Fix pep8 in test_runner

* Import runner only if torch import has succeeded

* Download one copy in distributed environment
Signed-off-by: innerlee <363664470@qq.com>

* fix
Signed-off-by: innerlee <363664470@qq.com>

* format
Signed-off-by: innerlee <363664470@qq.com>

* add iminvert

* DC now supports dimension other than 3

* fit flake8

* generalize pad_dim and beautify code

* fix bug; rename "pad_dim" to "pad_dims" for disambiguity

* fix bug

* fix flake8

* fix yapf format

* use deafault yapf

* beautify style

* flow_warp_c

* flow_warp_c

* beautify format

* beautify format

* beautify format

* beautify format

* add Cython

* modify

* fix details

* fix details

* fix type

Update to v0.2.8

* only display base lr when there are multi lrs

* change comment: base lr to lr of the first param group

* fix resnet-x101-32x4d-gn+ws hash error

* fix hash error of other models in jhu

add bgr2hls and hls2bgr

video fps should be float

Add jhu gn model urls and update to v0.2.7