problem: Run a container from a created docker image

**Issue 1: **

In the docker description it is mentioned that your test data will be mount on a folder name as '/input' and output will collect in the folder '/output'. But when you gave instuction to run the container you used '/workspace/input'.

docker run --rm --runtime=nvidia --ipc=host -e NVIDIA_VISIBLE_DEVICES=all --gpus 0 --user root -v LOCAL_PATH_INPUT:/workspace/input/:ro -v LOCAL_PATH_OUTPUT:/workspace/output/ TEAMNAME python run_inference.py

Can you verify whether the directory is ok?

Issue 2:

We are using docker in windows. It shows NVIDIA runtime error when I use '--runtime=nvidia'. If we use '--gpus all' it fix the runtime issue. Will we need to use NVIDIA -docker?

docker run --rm --runtime=nvidia --ipc=host -e NVIDIA_VISIBLE_DEVICES=all --gpus 0 --user root -v LOCAL_PATH_INPUT:/workspace/input/:ro -v LOCAL_PATH_OUTPUT:/workspace/output/ TEAMNAME python run_inference.py

Thank you very much for your valuable time.

Monibor

Hi, thank you for your information. For the issue 1, "input" and "output" folders are the folders on our server, while the "workspace/input" and "workspace/output" are the expcted folder in your docker. We use "LOCAL_PATH_INPUT:/workspace/input/:ro -v LOCAL_PATH_OUTPUT:/workspace/output/" to link them (LOCAL_PATH_INPUT is the "input" and LOCAL_PATH_OUTPUT is the "/workspace/output/" ). So please use "workspace/xx".

For issue 2, please use Nvidia-docker to fix, as there is a common issue in many pc.