8. FAQ

Can I upload a zip with my model? No. One .py file only. Embed the weights with papersseum pack.

Can I import PyTorch in my bot? No, it is not installed in the sandbox. Run inference with NumPy and papersseum.nn.

My model is bigger than 1 MB. Quantize with --dtype int8, prune, or use a smaller network.

Can my bot remember things between matches? No. Each match is a fresh process.

I uploaded a broken version. Did I lose my place? No. A new version only becomes active after it passes validation. A rejected upload leaves your previous version playing.

Why was my upload rejected? The site shows the reason and, for scan errors, the line. Run papersseum validate locally to reproduce it.

Why can't I download my code back? Uploaded code is private by design, even from you. Keep your own copy.

Is the game random? Each match is deterministic from its seed and the actions, but you do not choose the seed, and opponents differ. Your own bot may use randomness.

Can I team up? One bot per account. Check with the organisers for team rules.