* feat(detection): add new models and weights

* fix(detection): fix random features in py3.5

* feat(detection): use numerical stable focal loss and EMA normalizer in
RetinaNet

* chore(detection): rename some variables

* feat(detection): several enhancement and reformat

* feat(detection): use multi-scale training

* chore(detection): update weights and results

* fix(keypoints): fix README and hubconf

* feat(detection): add sublinear memory examples

* feat(detection): support VOC in training
TODO: evaluation

* fix(detection): fix hubconf and licences

* fix(detection): numerical stability for focal loss

feat(quant): support calibration

* feat(detection): support objects365

* feat(detection): support retinanet with SyncBN

* feat(detection): add GroupedSampler

fix(bert) fix pylint error

* add quantization codebase for resnet, shufflenetv1, and mobilenetv2
Co-authored-by: NKevin.W <xsw200831720@gmail.com>
Co-authored-by: Nwangshupeng <wangshupeng@megvii.com>
Co-authored-by: LiZhiyuan <848796515@qq.com>
Co-authored-by: Nwangjianfeng <wangjianfeng@megvii.com>

* fix(detection): fix hubconf.py import error

* fix(detection): fix hubconf.py import error

feat(detection): support Objects365 and reformat

* add pylintrc
* Create ci.yml
* fix typo in shufflenet
* add CI badge

1. add pretrained resnet152, snetv2_0.5x, 1.5x and 2.0x
2. move `import pycocotools` inside `main` to prevernt import error