Dear LISA Organizers, for some reason when submitting my model to the testing phase of 2b the error log ({job}_log.zip) gives in the middle mistakes like this: ``` FileNotFoundError: No .nii or .nii.gz predictions found in {output_dir} ``` While the {job}_log.tex doesn't visualize the full logging of my model and only the second part of it When I run the model on my local system like this (as provided in your container example with additional "shm-size=8g" and "gpus all"), then everything works fine and the predictions are written from the /output to $PWD/output directory on my system as expected, for example: ``` docker run --rm --network none --volume $PWD/input:/input:ro --volume $PWD/output:/output:rw --gpus all --shm-size=8g docker.synapse.org/syn68368708/2b:v1.1 ``` In inability to find any other clues, what could be the reason of the mistakes I hope you could help me understand the issue of the problem. Additional info: - It can take for my model up to 1-2 minutes to do the prediction for each subject, depending on the GPUs, that are used. - There are additional logs like this in the logfile: ``` INFO:cwltool:[job run_docker.cwl] Max memory used: 54MiB ``` - my image uses pytorch-cuda with cuda version cu128 UPDATE: The issue is resolved, we were able to submit and to get around the problem, however (just a sidenote) we believe that the error with limited shared memory still exists when evaluating dockers.