Сan I record my own gameplay, create my own reward function and use this when training an agent? The main question is how to use this data during the second stage of the competition, when you need to train remotely. Can I create, maybe, an docker image with 2-3 GB of my own replays?
And am I obligated in this case to share this data with other participants?
From the rules:
Participants may only use the provided dataset; no additional datasets may be included in the source file submissions nor may be downloaded during training evaluation. During evaluation of submitted code, the individual containers will not have access to any external network to avoid any information leak. All submitted code repositories will be scrubbed to remove any files larger than 30MB to ensure participants are not checking in any model weighs pre-trained on the released training dataset. While the container running the submitted code will not have external network access, relevant exceptions are added to ensure participants can download and use the pre-trained models included in popular frameworks like PyTorch and TensorFlow. Participants can request to add network exceptions for any other publicly available pre-trained models, which will be validated by AICrowd on a case-by-case basis.
Well additionally, we use a different hold-out hidden texture during round 2 so even if you trained on your own dataset, it would damage your performance in round 2. That’s the main reason behind this rule!