arXiv subjects

Ramachandran Ramjee

Publications and source records attributed to Ramachandran Ramjee.

24 records · Page 2Linked to original sources

arXiv2021

Varuna: Scalable, Low-cost Training of Massive Deep Learning Models

Systems for training massive deep learning models (billions of parameters) today assume and require specialized "hyper-clusters": hundreds or thousands of GPUs wired with specialized high-bandwidth interconnects such as NV-Link and Infiniband. Besides being expensive, such dependence on hyper-clusters and custom high-speed inter-connects limits the size of such clusters, creating (a) scalability limits on job parallelism; (b) resource fragmentation across hyper-clusters. In this paper, we present Varuna, a new system that enables training massive deep learning models on commodity networking. Varuna makes thrifty use of networking resources and automatically configures the user's training job to efficiently use any given set of resources. Therefore, Varuna is able to leverage "low-priority" VMs that cost about 5x cheaper than dedicated GPUs, thus significantly reducing the cost of training massive models. We demonstrate the efficacy of Varuna by training massive models, including a 200 billion parameter model, on 5x cheaper "spot VMs", while maintaining high training throughput. Varuna improves end-to-end training time by up to 18x compared to other model-parallel approaches and up to 26% compared to other pipeline parallel approaches. The code for Varuna is available at https://github.com/microsoft/varuna.

cs.DC↗

arXiv2021

Wide-minima Density Hypothesis and the Explore-Exploit Learning Rate Schedule

Several papers argue that wide minima generalize better than narrow minima. In this paper, through detailed experiments that not only corroborate the generalization properties of wide minima, we also provide empirical evidence for a new hypothesis that the density of wide minima is likely lower than the density of narrow minima. Further, motivated by this hypothesis, we design a novel explore-exploit learning rate schedule. On a variety of image and natural language datasets, compared to their original hand-tuned learning rate baselines, we show that our explore-exploit schedule can result in either up to 0.84% higher absolute accuracy using the original training budget or up to 57% reduced training time while achieving the original reported accuracy. For example, we achieve state-of-the-art (SOTA) accuracy for IWSLT'14 (DE-EN) dataset by just modifying the learning rate schedule of a high performing model.

cs.LG↗

arXiv2021

LRTuner: A Learning Rate Tuner for Deep Neural Networks

One very important hyperparameter for training deep neural networks is the learning rate schedule of the optimizer. The choice of learning rate schedule determines the computational cost of getting close to a minima, how close you actually get to the minima, and most importantly the kind of local minima (wide/narrow) attained. The kind of minima attained has a significant impact on the generalization accuracy of the network. Current systems employ hand tuned learning rate schedules, which are painstakingly tuned for each network and dataset. Given that the state space of schedules is huge, finding a satisfactory learning rate schedule can be very time consuming. In this paper, we present LRTuner, a method for tuning the learning rate as training proceeds. Our method works with any optimizer, and we demonstrate results on SGD with Momentum, and Adam optimizers. We extensively evaluate LRTuner on multiple datasets, models, and across optimizers. We compare favorably against standard learning rate schedules for the given dataset and models, including ImageNet on Resnet-50, Cifar-10 on Resnet-18, and SQuAD fine-tuning on BERT. For example on ImageNet with Resnet-50, LRTuner shows up to 0.2% absolute gains in test accuracy compared to the hand-tuned baseline schedule. Moreover, LRTuner can achieve the same accuracy as the baseline schedule in 29% less optimization steps.

cs.LG↗

arXiv2019

Privado: Practical and Secure DNN Inference with Enclaves

Cloud providers are extending support for trusted hardware primitives such as Intel SGX. Simultaneously, the field of deep learning is seeing enormous innovation as well as an increase in adoption. In this paper, we ask a timely question: "Can third-party cloud services use Intel SGX enclaves to provide practical, yet secure DNN Inference-as-a-service?" We first demonstrate that DNN models executing inside enclaves are vulnerable to access pattern based attacks. We show that by simply observing access patterns, an attacker can classify encrypted inputs with 97% and 71% attack accuracy for MNIST and CIFAR10 datasets on models trained to achieve 99% and 79% original accuracy respectively. This motivates the need for PRIVADO, a system we have designed for secure, easy-to-use, and performance efficient inference-as-a-service. PRIVADO is input-oblivious: it transforms any deep learning framework that is written in C/C++ to be free of input-dependent access patterns thus eliminating the leakage. PRIVADO is fully-automated and has a low TCB: with zero developer effort, given an ONNX description of a model, it generates compact and enclave-compatible code which can be deployed on an SGX cloud platform. PRIVADO incurs low performance overhead: we use PRIVADO with Torch framework and show its overhead to be 17.18% on average on 11 different contemporary neural networks.

cs.CR↗

arXiv2016

Perspectives on Software-Defined Networks: interviews with five leading scientists from the networking community

Software defined Networks (SDNs) have drawn much attention both from academia and industry over the last few years. Despite the fact that underlying ideas already exist through areas such as P2P applications and active networks (e.g. virtual topologies and dynamic changes of the network via software), only now has the technology evolved to a point where it is possible to scale the implementations, which justifies the high interest in SDNs nowadays. In this article, the JISA Editors invite five leading scientists from three continents (Raouf Boutaba, David Hutchison, Raj Jain, Ramachandran Ramjee, and Christian Esteve Rothenberg) to give their opinions about what is really new in SDNs. The interviews cover whether big telecom and data center companies need to consider using SDNs, if the new paradigm is changing the way computer networks are understood and taught, and what are the open issues on the topic.

cs.NI↗

arXiv2009

Analytical Models for Energy Consumption in Infrastructure WLAN STAs Carrying TCP Traffic

We develop analytical models for estimating the energy spent by stations (STAs) in infrastructure WLANs when performing TCP controlled file downloads. We focus on the energy spent in radio communication when the STAs are in the Continuously Active Mode (CAM), or in the static Power Save Mode (PSM). Our approach is to develop accurate models for obtaining the fraction of times the STA radios spend in idling, receiving and transmitting. We discuss two traffic models for each mode of operation: (i) each STA performs one large file download, and (ii) the STAs perform short file transfers. We evaluate the rate of STA energy expenditure with long file downloads, and show that static PSM is worse than just using CAM. For short file downloads we compute the number of file downloads that can be completed with given battery capacity, and show that PSM performs better than CAM for this case. We provide a validation of our analytical models using the NS-2 simulator. In contrast to earlier work on analytical modeling of PSM, our models that capture the details of the interactions between the 802.11 MAC in PSM and certain aspects of TCP.

cs.NI↗