Deriva — Derive the Algorithm

← Projectsmini-transformer / L2Embeddings + Positions~20 min

one move: preserve an invariant

Make multi-head attention's shape contract explicit by splitting one representation into equal-width head vectors. In the Mini Transformer system, implement this level as a deterministic contract before composing it with the next service.

Required API

def split_heads(vector, head_count):
    return [head, ...]

Behavior

divide the model width evenly
preserve feature order
return one list per head

Examples — tap to reveal

Constraints

Use only the Python standard library.
Return deterministic values so every run can be compared with the contract.
Do not mutate caller-owned inputs.

Opening…