Hello TrojAI Community,
Below is an announcement we made to Performers that is relevant to non-performer teams as well. I've copied it below:
To make sure teams are using the correct PyTorch versions, we have asked the Performer teams who submitted to the Round 1 leaderboard to submit their same models (but upgraded to PyTorch 1.6) to the Round 2 Smoke Test Server (STS).
This will publicly verify that people are using the correct versions so we can avoid any infrastructure issues early on. Only performer teams that submitted on the Round 1 leaderboard are required to do this. However, if non-performer teams want to test the ir PyTorch versioning they are welcome to submit to the STS with their old round 1 models as well.
Don't worry if you get bad performance, we are only trying to confirm that the models do not error out.
Email me directly if you have questions!