Query Regarding Docker Runtime Environment

Dear Organizers,

I hope this message finds you well. My name is Jaehwan Han and I am currently participating in the challenge.

I am writing to you with a query regarding the Docker runtime environment used in the competition.

In my preliminary tests, it appears that the Docker container for the challenge is designed to run on a CPU, as I could not find any GPU options within the test.sh script command. I understand that the performance of our models is not solely evaluated on their predictive accuracy, but also on memory usage and execution time as per the competition rules.

To quote the competition rules:

"If two or more final ranks obtained at point 3 are equal, compute the average of the rankings obtained at point 2 for Mem and Time to break ties."

Given that the running time of our models plays a crucial role in the final ranking, I would like to clarify if the use of GPU acceleration is allowed and, if so, how we might be able to enable this within the Docker container.

This would significantly enhance the performance of my model and ensure a fair comparison with other models that might be using GPU acceleration.

I would greatly appreciate it if you could clarify this for me.

Thank you in advance for your time and assistance.

Best regards,

Jaehwan Han

Dear Jaehwan Han,

You have the option to use a GPU for your task, and when uploading your Docker container, you can indicate whether it requires GPU support by checking the "GPU Supported" checkbox. The grand-challenge documentation state that the GPU assigned is an NVidia T4 16GB.

The Dockerfile itself is designed to be independent of the presence of a GPU. However, during the execution of 'docker run', you have the flexibility to assign a GPU using the '--gpus all' flag if you choose to do so. In order to ensure compatibility with different systems, I have not set this flag by default in the 'test.sh' script, as not all participants may have access to a GPU. When the Docker container is executed by grand-challenge, this flag would be used if the "GPU Supported" checkbox has been checked.

To provide further clarification, I will update the GitHub repository with this kind of information. This will help anyone who wishes to run your solution with a GPU to easily find the necessary instructions.

Reach out again if you have further questions. Regards, Luca