singula/cuda repository overview

This image has been deprecated!!!

build tensorflow: bazel build -c opt --config=cuda --incompatible_load_argument_is_label=false //tensorflow/tools/pip_package:build_pip_package

generate whl file bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg

then the file will be there: /tmp/tensorflow_pkg/tensorflow-1.4.1-cp35-cp35m-linux_x86_64.whl

put the .whl file to your linux and install it: pip install tensorflow-1.4.1-cp35-cp35m-linux_x86_64.whl

docker-compose file:

⁠cuda for tensorflow building Powered by singula

version: '2' services: cuda: image: singula/cuda:8.0-cudnn6-devel container_name: mycuda #restart: always volumes: - /home/:/home/ - /usr/local/nvidia:/var/lib/nvidia-docker/volumes/nvidia_driver/384.90 - /dev/nvidia0:/dev/nvidia0 - /dev/nvidia-uvm:/dev/nvidia-uvm - /dev/nvidia-uvm-tools:/dev/nvidia-uvm-tools - /dev/nvidiactl:/dev/nvidiactl - /usr/lib/nvidia-384:/usr/lib/nvidia-384 - /usr/lib/x86_64-linux-gnu/libcuda.so:/usr/lib/x86_64-linux-gnu/libcuda.so - /usr/lib/x86_64-linux-gnu/libcuda.so.1:/usr/lib/x86_64-linux-gnu/libcuda.so.1 - /usr/lib/x86_64-linux-gnu/libcuda.so.384.90:/usr/lib/x86_64-linux-gnu/libcuda.so.384.90 - /usr/lib/nvidia-384/libnvidia-fatbinaryloader.so.384.90:/usr/lib/x86_64-linux-gnu/libnvidia-fatbinaryloader.so.384.90 command: sleep 999999999