Use cmake to build the project. Make sure to install all the dependencies described here.

Adjust settings in cmake (build-type Release or Debug, configure CUDA, OpenBLAS or Atlas, OpenCV, OpenMP etc.)

$ cd mxnet
$ mkdir build/Release && cd build/Release
$ cmake ../../
$ ccmake . 
$ make -j `nproc`

$ cd mxnet/python
$ pip install --upgrade pip
$ pip install -e .

First run the training script and store the best performed checkpoint.

cd my
python train.py

Then modify common/data.py and train.py to finetune the model using dataset with biased labels.