scatter_plot Phalanx¶
Federated learning on the latest Flower, with OpenTelemetry-native observability
hub flwr 1.31 Message API | tune federated LoRA | monitoring OTel traces + metrics
What is Phalanx?¶
Phalanx is a federated-learning research testbed that rides the latest Flower
release: the flwr Message API, flwr-datasets, HuggingFace Transformers, and
PEFT/LoRA. Its distinguishing feature is OpenTelemetry-native observability —
the server emits a span and FL metrics for every round, and each client emits a
span for its local train/evaluate pass, so a federated run is visible in any OTLP
backend (Jaeger, Grafana Tempo, an OpenTelemetry Collector).
The default showcase is a federated LoRA fine-tune of a tiny BERT on IMDB sentiment, partitioned non-IID with a Dirichlet partitioner. Only the LoRA adapters and the classification head are federated; the frozen backbone stays on each client.