awokeknowing/aitools repository overview

⁠About awokeknowing/aitools

To run you need to have nvidia driver and nvidia-docker 2 for runtime=nvidia to work. For x-forwarding, eg to see matplatlib and gym atari games locallly, you may need to run xhost +si:localuser:root first):

docker run --runtime=nvidia -ti --rm -e DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix awokeknowing/aitools

⁠awokeknowing/aitools tagged 18.4 (latest) contain:

• Based on new nVidia CUDA 9 with OpenGL image nvidia/cudagl:9.0-devel-ubuntu16.04
• python 3.6 (not anaconda)
• OpenCV 3.4 (compiled with cuda, blas)
• Tensorflow / Keras
• Pytorch (pip, not anaconda)
• OpenAI gym (with atari, classic control, and box2d, not mujoco)
• Jupyter Lab (also Jupyter notebook)
• Various dependencies and extras related to the above (thanks floydhub)

Just type lab at the prompt when you run the image as mentioned above, with x-forwarding. Jupyter lab will start, and you can Ctrl+click the link to go to Jupyter Lab in your browser. I included a couple samples which if you run, will show OpenCV and OpenAI Gym running on your local display.

So as a final clarification, there is no 'desktop', nor browser in the image. It's just standard tools for working with AI on a cuda/ubuntu image.

NOTE: because it takes a while to build, I build locally and push the image. That's why the Dockerfile doesn't show, but you can see it in the linked repo.