Arjun Parthasarathy | Searcharxiv

arXiv2025

Closing the Train-Test Gap in World Models for Gradient-Based Planning

World models paired with model predictive control (MPC) can be trained offline on large-scale datasets of expert trajectories and enable generalization to a wide range of planning tasks at inference time. Compared to traditional MPC procedures, which rely on slow search algorithms or on iteratively solving optimization problems exactly, gradient-based planning offers a computationally efficient alternative. However, the performance of gradient-based planning has thus far lagged behind that of other approaches. In this paper, we propose improved methods for training world models that enable efficient gradient-based planning. We begin with the observation that although a world model is trained on a next-state prediction objective, it is used at test-time to instead estimate a sequence of actions. The goal of our work is to close this train-test gap. To that end, we propose train-time data synthesis techniques that enable significantly improved gradient-based planning with existing world models. At test time, our approach outperforms or matches the classical gradient-free cross-entropy method (CEM) across a variety of object manipulation and navigation tasks in 10% of the time budget.

cs.LG↗

arXiv2023

Partitioning and Deployment of Deep Neural Networks on Edge Clusters

Edge inference has become more widespread, as its diverse applications range from retail to wearable technology. Clusters of networked resource-constrained edge devices are becoming common, yet no system exists to split a DNN across these clusters while maximizing the inference throughput of the system. Additionally, no production-ready orchestration system exists for deploying said models over such edge networks which adopts the robustness and scalability of the cloud. We present an algorithm which partitions DNNs and distributes them across a set of edge devices with the goal of minimizing the bottleneck latency and therefore maximizing inference throughput. The system scales well to systems of different node memory capacities and numbers of nodes, while being node fault-tolerant. We find that we can reduce the bottleneck latency by 10x over a random algorithm and 35% over a greedy joint partitioning-placement algorithm, although the joint-partitioning algorithm outperforms our algorithm in most practical use-cases. Furthermore we find empirically that for the set of representative models we tested, the algorithm produces results within 9.2% of the optimal bottleneck latency. We then developed a standalone cluster network emulator on which we tested configurations of up to 20 nodes and found a steady increase in throughput and decrease in end-to-end latency as the cluster size scales. In these tests, we observed that our system has multi-node fault-tolerance as well as network and system IO fault-tolerance. We have implemented our framework in open-source software that is publicly available to the research community at https://github.com/ANRGUSC/SEIFER.

cs.NI↗

arXiv2022

SEIFER: Scalable Edge Inference for Deep Neural Networks

Edge inference is becoming ever prevalent through its applications from retail to wearable technology. Clusters of networked resource-constrained edge devices are becoming common, yet there is no production-ready orchestration system for deploying deep learning models over such edge networks which adopts the robustness and scalability of the cloud. We present SEIFER, a framework utilizing a standalone Kubernetes cluster to partition a given DNN and place these partitions in a distributed manner across an edge network, with the goal of maximizing inference throughput. The system is node fault-tolerant and automatically updates deployments based on updates to the model's version. We provide a preliminary evaluation of a partitioning and placement algorithm that works within this framework, and show that we can improve the inference pipeline throughput by 200% by utilizing sufficient numbers of resource-constrained nodes. We have implemented SEIFER in open-source software that is publicly available to the research community.

cs.NI↗

arXiv2022

Partitioning and Placement of Deep Neural Networks on Distributed Edge Devices to Maximize Inference Throughput

Edge inference has become more widespread, as its diverse applications range from retail to wearable technology. Clusters of networked resource-constrained edge devices are becoming common, yet no system exists to split a DNN across these clusters while maximizing the inference throughput of the system. We present an algorithm which partitions DNNs and distributes them across a set of edge devices with the goal of minimizing the bottleneck latency and therefore maximizing inference throughput. The system scales well to systems of different node memory capacities and numbers of nodes. We find that we can reduce the bottleneck latency by 10x over a random algorithm and 35% over a greedy joint partitioning-placement algorithm. Furthermore we find empirically that for the set of representative models we tested, the algorithm produces results within 9.2% of the optimal bottleneck latency.

cs.NI↗

arXiv2022

DEFER: Distributed Edge Inference for Deep Neural Networks

Modern machine learning tools such as deep neural networks (DNNs) are playing a revolutionary role in many fields such as natural language processing, computer vision, and the internet of things. Once they are trained, deep learning models can be deployed on edge computers to perform classification and prediction on real-time data for these applications. Particularly for large models, the limited computational and memory resources on a single edge device can become the throughput bottleneck for an inference pipeline. To increase throughput and decrease per-device compute load, we present DEFER (Distributed Edge inFERence), a framework for distributed edge inference, which partitions deep neural networks into layers that can be spread across multiple compute nodes. The architecture consists of a single "dispatcher" node to distribute DNN partitions and inference data to respective compute nodes. The compute nodes are connected in a series pattern where each node's computed result is relayed to the subsequent node. The result is then returned to the Dispatcher. We quantify the throughput, energy consumption, network payload, and overhead for our framework under realistic network conditions using the CORE network emulator. We find that for the ResNet50 model, the inference throughput of DEFER with 8 compute nodes is 53% higher and per node energy consumption is 63% lower than single device inference. We further reduce network communication demands and energy consumption using the ZFP serialization and LZ4 compression algorithms. We have implemented DEFER in Python using the TensorFlow and Keras ML libraries, and have released DEFER as an open-source framework to benefit the research community.

cs.LG↗