arXiv:2608. 06046v1 Announce Type: cross Abstract: AI training workloads are growing rapidly, making their time, energy, and infrastructure costs increasingly important.
arXiv:2402. 09589v2 Announce Type: replace-cross Abstract: We present MLCC, a novel technique to augment today's congestion control algorithms to accelerate DNN training jobs in shared GPU clusters in a fully distributed manner.
The paper introduces a dynamic framework for partitioning neural network layers across a heterogeneous edge‑cloud continuum, adapting to runtime changes in network conditions and device capabilities. It profiles models at startup, measures link quality, and periodically re‑evaluates the partitioning to optimize performance. Experiments on a Raspberry Pi, laptop, and desktop using VGG16, AlexNet, and MobileNetV2 demonstrate energy savings of 27.09–35.82% and latency reductions of 6.34–22.92% over static partitioning.
MANE is a distributed inference framework that uses a multi‑path tail architecture to allow dynamic accuracy–throughput trade‑offs during edge onloading of deep neural networks. It introduces a novel multi‑path model, a three‑stage training scheme with Joint Head Network Distillation loss, and a hysteresis‑based scheduler with an equitable device‑fallback policy. The system achieves over 80% SLO satisfaction and 6pp higher accuracy than on‑device alternatives while supporting up to 40 concurrent devices.
arXiv:2605. 01989v2 Announce Type: replace Abstract: Distributed machine learning (ML) training has become a necessity with the prevalence of billion to trillion-parameter-scale models.
arXiv:2507. 09029v5 Announce Type: replace-cross Abstract: Pre-training large neural networks at scale imposes heavy memory demands on accelerators and often requires costly communication.
arXiv:2512.04705v3 Announce Type: replace-cross Abstract: The deployment of Early-Exiting Neural Networks (EENNs) on edge accelerators requires optimizing not only the network architecture but also i...
arXiv:2606. 07565v1 Announce Type: new Abstract: Intelligent scaling of microservices in cloud platforms is crucial for mitigating escalating compute costs while avoiding service disruptions.
The paper proposes using the network itself to aid distributed training over a wide area network (WAN). It suggests employing multicast for outbound traffic and in‑line FPGAs for inbound traffic to reduce bottlenecks, extending techniques commonly used in data centers to the WAN. An optimization framework generates synchronization schedules—rotating cliques of compute islands—tailored to the network topology, and demonstrates these ideas on a nine‑city WAN model, showing how schedules adapt to network capabilities.
Ampere is a new split federated learning system that reduces both on‑device computation and device‑server communication while improving accuracy. It trains device and server blocks sequentially with local losses, eliminating gradient transfers, and uses a lightweight auxiliary network to consolidate activations into a single transfer. Experiments on CNNs and Transformers show up to 11.70 pp accuracy gains, 18.6× faster training, 911× less communication, and 14.5× less computation compared to state‑of‑the‑art SFL baselines.
The paper introduces REACT, a system that dynamically tunes communication collectives in distributed AI training to mitigate congestion without requiring network infrastructure changes. REACT operates at the application layer, detecting congestion via flow statistics and adjusting the pattern of data exchange—such as selecting different aggregation nodes in an AllReduce tree—while preserving the semantics of the communication. Evaluations on a shared academic GPU cluster show that REACT improves algorithm bandwidth by 13%–38% under congestion, with simulations indicating potential gains up to 75%.
The paper introduces ModalShare, a bandwidth allocation method for multimodal split learning that assigns each modality a keep‑ratio based on its Shapley contribution score. Unlike existing compression schemes that split the uplink budget proportionally to activation size, ModalShare explicitly optimizes the split across modalities, requiring no extra uplink traffic or client computation. Experiments on CREMA‑D and MVSA datasets show that ModalShare improves accuracy by 12.4–15.4 percentage points over equal keep‑ratios under a 5× compression budget, outperforming three compressors across multiple datasets and budgets.
arXiv:2608. 09250v1 Announce Type: new Abstract: Federated learning (FL) must serve devices with varying computational capabilities.