src/pqc_federated_learning/aggregators/fedsum.py
| 1 | """FedSum: unweighted elementwise sum of client tensors.""" |
| 2 | |
| 3 | from __future__ import annotations |
| 4 | |
| 5 | from pqc_federated_learning.aggregators.base import Aggregator |
| 6 | from pqc_federated_learning.errors import InsufficientUpdatesError, ShapeMismatchError |
| 7 | from pqc_federated_learning.update import ClientUpdate, GradientTensor |
| 8 | |
| 9 | |
| 10 | class FedSumAggregator(Aggregator): |
| 11 | name = "fedsum" |
| 12 | |
| 13 | def aggregate(self, updates: list[ClientUpdate]) -> list[GradientTensor]: |
| 14 | if not updates: |
| 15 | raise InsufficientUpdatesError("FedSum requires at least one update") |
| 16 | |
| 17 | names = [t.name for t in updates[0].tensors] |
| 18 | shapes = {t.name: t.shape for t in updates[0].tensors} |
| 19 | for u in updates[1:]: |
| 20 | if [t.name for t in u.tensors] != names: |
| 21 | raise ShapeMismatchError("tensor name mismatch across updates") |
| 22 | for t in u.tensors: |
| 23 | if t.shape != shapes[t.name]: |
| 24 | raise ShapeMismatchError( |
| 25 | f"{t.name} shape mismatch: {t.shape} vs {shapes[t.name]}" |
| 26 | ) |
| 27 | |
| 28 | out: list[GradientTensor] = [] |
| 29 | for name in names: |
| 30 | shape = shapes[name] |
| 31 | length = 1 |
| 32 | for d in shape: |
| 33 | length *= d |
| 34 | agg = [0.0] * length |
| 35 | for u in updates: |
| 36 | tensor = next(t for t in u.tensors if t.name == name) |
| 37 | for i, v in enumerate(tensor.values): |
| 38 | agg[i] += v |
| 39 | out.append(GradientTensor(name=name, shape=shape, values=tuple(agg))) |
| 40 | return out |
| 41 | |