We have summarized several common issues with Docker submissions as follows: 1. Please use the folder name "TestSet" instead of "ValidationSet" during the reading or generation stage. 2. Please strictly follow the file input and output paths as specified on Synapse. 3. There are common issues with the order or size of the output data dimensions. Please ensure that the data saving format is the same as the validation stage, with the exception of cropping (do not perform any cropping on any of the four dimensions). 4. There are incorrect variable names in the output. Please use "img4ranking" as the key. 5. Please save the "img4ranking" file as an magnitude image (absolute value) instead of a complex image. 6. The Docker testing timeout is too strict. Although we do not strictly limit the time, please control it within an acceptable range.

Created by Recon CMR CMRxRecon
So far, for the majority of members, the program bottleneck does not lie in I/O. They are able to complete the inference within the designated timeframe.
We are use NFS (Network File System) with a network card speed of 1000 Mb/s .
Hi, thanks for the update. Can you give some information about the storage system? Is the output some network-attached HPC storage? If yes, can you consider giving us a local /scratch or similar that is local to the compute node executing the docker container for faster temporary storage? background: We got the feedback that our docker container is way to slow, but have difficulties recreating the issue on our hpc system. A slower output storage might be one of the differences between our system and the system you use for running the docker containers)

Important considerations for submitting a Docker container page is loading…