在 nvidia cuda docker 中运行 python 脚本

Question

I downloaded and set up the docker https://www.tensorflow.org/install/docker and the image https://github.com/NVIDIA/nvidia-docker .

Test case works:

docker run --gpus all -it --rm tensorflow / tensorflow: latest-gpu \
python -c "import tensorflow as tf; print (tf.reduce_sum (tf.random.normal ([1000, 1000])))"

What should I do to make the program written in Visual Studio Code run in docker instead of python (locally installed)?

docker run --gpus all -it --rm tensorflow / tensorflow: latest-gpu \
    python /home/something.py

It doesn't work because what I understand this is the path in the docker and not on the host machine. Should I build a new docker image with the program inside? I wanted to use the downloaded image without changing it.

Answer 1

You will have to mount your local directory [where visual code has saved your python file] on docker. Here is a working example -

docker run --gpus all -v <your_directory_where_visual_code_has_saved_python_file>:/home -it --rm tensorflow/tensorflow:latest-gpu
python /home/something.py

My working example that i tried -

docker run --gpus all -v /home/ubuntu/test:/home/test -it tensorflow/tensorflow:latest-gpu python /home/jitest.py

Answer 2

To make your program run on docker, you have many options:

1. in your Dockerfile copy the code inside the docker

2. you can mount the folder containing your python code as docker volume [recommended]

for more information check this medium articles