Execution Failed : The algorithm failed on one or more cases.

Dear Challenge Organizers,

I am attempting to submit the test algorithm, but it continues to fail despite my model’s resources (GPU 16GB, RAM 32GB) meeting the challenge requirements.

I have followed the provided submission repository with a few modifications, and it works perfectly on my local setup. The changes I made are detailed below

Docker file: FROM --platform=linux/amd64 pytorch/pytorch => FROM --platform=linux/amd64 pytorch/pytorch:2.0.0-cuda11.7-cudnn8-runtime

Inference file: Reduced model parameters and input size to lower GPU resource usage

Could you please check my submission logs?

Additionally, according to the Task & Evaluation tab, only five submission attempts are allowed during the Algorithm Refinement Period. I would like to know if failed submissions are counted towards this limit.

Thank you for your assistance.

Best regards,

Please check your inbox on the Grand-Challenge website for additional details.

Additionally, when using sliding window inference, transfer the data to the CPU instead of the GPU. You can do this as shown below:

sliding_window_inference(image, (patch_size, patch_size, patch_size), num_samples, model, device='cpu')

Also, failed submission do not count towards the limit.