Step to Training git clone https://github.com/liruihui/PU-GAN.git cd PU-GAN/Docker docker build -t tensorflow/pu-gan -f Dockerflie . Replace shell scripts under this directory to original ones cp tf_approxmatch_compile.sh ../tf_ops/approxmatch/tf_approxmatch_compile.sh cp tf_grouping_compile.sh ../tf_ops/grouping/tf_grouping_compile.sh cp tf_interpolate_compile.sh ../tf_ops/interpolation/tf_interpolate_compile.sh cp tf_nndistance_compile.sh ../tf_ops/nn_distance/tf_nndistance_compile.sh cp tf_sampling_compile.sh ../tf_ops/sampling/tf_sampling_compile.sh cp compile.sh ../tf_ops/compile.sh Start training cd ../ docker run -it --rm \ -e DISPLAY=unix$DISPLAY \ -v $(pwd):/workspace/ \ -v /tmp/.X11-unix:/tmp/.X11-unix \ -w /workspace \ --name pu-gan-runtime \ --gpus all \ --shm-size 8G \ tensorflow/pu-gan cd tf_ops bash compile.sh cd ../ python pu_gan.py --phase train