I am new to synapse so my question can be naive. Do we include the model weights trained locally to the docker image or will the model be trained by the organisers after the submission? I assumed it's the latter to make sure everyone uses only the provided data in training. If this is correct, what are the constraints on training time, memory (both RAM and GPU), and are the training images accessible in the `/mnt/data`?

Created by Kaisar Kushibar pianoza
We perform only inference and no training on our cluster. Please include weights etc in the docker, so it can run as a standalone evaluation script. Cheers (and sry for the late reply), David

Submission preparation and docker image contents page is loading…