General Rules
The rules for each dataset will vary depending on the nature of the problem. However, the following is a list of general rules that should apply to all challenges.
Rules
- You can not train the model on the testing dataset
- All results must be reproducible (code should be published, RNG seed should be disclosed, etc)
- Training/compression process should be automated. Any manual processes, if required, should be documented enough to be reproducible
- Optimiser weights aren't counted towards the total number of trainable parameters