无法更改在 tensorflow docker 容器上运行的 jupyter 笔记本的工作目录

Question

I have followed the steps in the official CUDA on WSL tutorial ( https://docs.nvidia.com/cuda/wsl-user-guide/index.html#ch05-sub02-jupyter ) to set up a jupyter notebook. However, I can't figure out how to change the initial working directory. I tried mounting a local directory with the -v switch as well as appending to the launch command --notebook-dir , but neither one of these solutions worked. The jupyter notebook will always start under "/tf" no matter what I do. Ideally, I would like this to be the same working directory as the one I have on Windows (C:\Users\MyUser).

The only thing I haven't tried is changing the WORKDIR in the docker image "tensorflow/tensorflow:latest-gpu-py3-jupyter" supplied by hub.docker.com as I am not even sure if it is possible to edit it (line 57).

Here is a sample command I have tried running:

docker run -it --gpus all -p 8888:8888 -v /c/Users/MyUser/MyFolder:/home/MyFolder/ tensorflow/tensorflow:latest-gpu-py3-jupyter jupyter notebook --allow-root --ip=0.0.0.0 --NotebookApp.allow_origin='https://colab.research.google.com' --notebook-dir=/c/Users/MyUser/

What is the easiest way to achieve this?

Answer 1

I was able to solve this problem by mounting the directory I want to work in under the local directory that is given in the command "Serving notebooks from local directory:/tf". In my case it's ' /tf ', but yours could be different. In addition, I changed the first '/' to '//'. Also, the container name should be the last argument (per https://stackoverflow.com/a/34503625 ). So in your case, the command looks like:

docker run -it --gpus all -p 8888:8888 -v //c/Users/MyUser/MyFolder:/tf/home/MyFolder tensorflow/tensorflow:latest-gpu-py3-jupyter