no code implementations • 1 Mar 2024 • Liang Luo, Buyun Zhang, Michael Tsang, Yinbin Ma, Ching-Hsiang Chu, Yuxin Chen, Shen Li, Yuchen Hao, Yanli Zhao, Guna Lakshminarayanan, Ellie Dingqiao Wen, Jongsoo Park, Dheevatsa Mudigere, Maxim Naumov
We study a mismatch between the deep learning recommendation models' flat architecture, common distributed training paradigm and hierarchical data center topology.
1 code implementation • 12 Sep 2023 • Hao-Jun Michael Shi, Tsung-Hsien Lee, Shintaro Iwasaki, Jose Gallego-Posada, Zhijing Li, Kaushik Rangadurai, Dheevatsa Mudigere, Michael Rabbat
It constructs a block-diagonal preconditioner where each block consists of a coarse Kronecker product approximation to full-matrix AdaGrad for each parameter of the neural network.
no code implementations • 19 Apr 2023 • Hiwot Tadese Kassa, Paul Johnson, Jason Akers, Mrinmoy Ghosh, Andrew Tulloch, Dheevatsa Mudigere, Jongsoo Park, Xing Liu, Ronald Dreslinski, Ehsan K. Ardestani
In Deep Learning Recommendation Models (DLRM), sparse features capturing categorical inputs through embedding tables are the major contributors to model size and require high memory bandwidth.
no code implementations • 28 Mar 2022 • Benjamin Ghaemmaghami, Mustafa Ozdal, Rakesh Komuravelli, Dmitriy Korchev, Dheevatsa Mudigere, Krishnakumar Nair, Maxim Naumov
However, this approach introduces collisions between semantically dissimilar ids that degrade model quality.
no code implementations • 11 Mar 2022 • Buyun Zhang, Liang Luo, Xi Liu, Jay Li, Zeliang Chen, Weilin Zhang, Xiaohan Wei, Yuchen Hao, Michael Tsang, Wenjun Wang, Yang Liu, Huayu Li, Yasmine Badr, Jongsoo Park, Jiyan Yang, Dheevatsa Mudigere, Ellie Wen
To overcome the challenge brought by DHEN's deeper and multi-layer structure in training, we propose a novel co-designed training system that can further improve the training efficiency of DHEN.
1 code implementation • 25 Oct 2021 • Ravi Krishna, Aravind Kalaiah, Bichen Wu, Maxim Naumov, Dheevatsa Mudigere, Misha Smelyanskiy, Kurt Keutzer
Neural architecture search (NAS) methods aim to automatically find the optimal deep neural network (DNN) architecture as measured by a given objective function, typically some combination of task accuracy and inference efficiency.
no code implementations • 21 Oct 2021 • Ehsan K. Ardestani, Changkyu Kim, Seung Jae Lee, Luoshang Pan, Valmiki Rampersad, Jens Axboe, Banit Agrawal, Fuxun Yu, Ansha Yu, Trung Le, Hector Yuen, Shishir Juluri, Akshat Nanda, Manoj Wodekar, Dheevatsa Mudigere, Krishnakumar Nair, Maxim Naumov, Chris Peterson, Mikhail Smelyanskiy, Vijay Rao
Deep Learning Recommendation Models (DLRM) are widespread, account for a considerable data center footprint, and grow by more than 1. 5x per year.
no code implementations • 12 Apr 2021 • Dheevatsa Mudigere, Yuchen Hao, Jianyu Huang, Zhihao Jia, Andrew Tulloch, Srinivas Sridharan, Xing Liu, Mustafa Ozdal, Jade Nie, Jongsoo Park, Liang Luo, Jie Amy Yang, Leon Gao, Dmytro Ivchenko, Aarti Basant, Yuxi Hu, Jiyan Yang, Ehsan K. Ardestani, Xiaodong Wang, Rakesh Komuravelli, Ching-Hsiang Chu, Serhat Yilmaz, Huayu Li, Jiyuan Qian, Zhuobo Feng, Yinbin Ma, Junjie Yang, Ellie Wen, Hong Li, Lin Yang, Chonglin Sun, Whitney Zhao, Dimitry Melts, Krishna Dhulipala, KR Kishore, Tyler Graf, Assaf Eisenman, Kiran Kumar Matam, Adi Gangidi, Guoqiang Jerry Chen, Manoj Krishnan, Avinash Nayak, Krishnakumar Nair, Bharath Muthiah, Mahmoud khorashadi, Pallab Bhattacharya, Petr Lapukhov, Maxim Naumov, Ajit Mathews, Lin Qiao, Mikhail Smelyanskiy, Bill Jia, Vijay Rao
Deep learning recommendation models (DLRMs) are used across many business-critical services at Facebook and are the single largest AI application in terms of infrastructure demand in its data-centers.
no code implementations • 17 Oct 2020 • Assaf Eisenman, Kiran Kumar Matam, Steven Ingram, Dheevatsa Mudigere, Raghuraman Krishnamoorthi, Krishnakumar Nair, Misha Smelyanskiy, Murali Annavaram
While Check-N-Run is applicable to long running ML jobs, we focus on checkpointing recommendation models which are currently the largest ML models with Terabytes of model size.
no code implementations • 20 Mar 2020 • Maxim Naumov, John Kim, Dheevatsa Mudigere, Srinivas Sridharan, Xiaodong Wang, Whitney Zhao, Serhat Yilmaz, Changkyu Kim, Hector Yuen, Mustafa Ozdal, Krishnakumar Nair, Isabel Gao, Bor-Yiing Su, Jiyan Yang, Mikhail Smelyanskiy
Large-scale training is important to ensure high performance and accuracy of machine-learning models.
Distributed, Parallel, and Cluster Computing 68T05, 68M10 H.3.3; I.2.6; C.2.1
no code implementations • 15 Jan 2020 • Rohan Saphal, Balaraman Ravindran, Dheevatsa Mudigere, Sasikanth Avancha, Bharat Kaul
However, ensemble methods are relatively less popular in reinforcement learning owing to the high sample complexity and computational expense involved in obtaining a diverse ensemble.
6 code implementations • 25 Sep 2019 • Antonio Ginart, Maxim Naumov, Dheevatsa Mudigere, Jiyan Yang, James Zou
Embedding representations power machine intelligence in many applications, including recommendation systems, but they are space intensive -- potentially occupying hundreds of gigabytes in large-scale settings.
6 code implementations • 4 Sep 2019 • Hao-Jun Michael Shi, Dheevatsa Mudigere, Maxim Naumov, Jiyan Yang
We propose a novel approach for reducing the embedding size in an end-to-end fashion by exploiting complementary partitions of the category set to produce a unique embedding vector for each category without explicit definition.
7 code implementations • 6 Jun 2019 • Udit Gupta, Carole-Jean Wu, Xiaodong Wang, Maxim Naumov, Brandon Reagen, David Brooks, Bradford Cottel, Kim Hazelwood, Bill Jia, Hsien-Hsin S. Lee, Andrey Malevich, Dheevatsa Mudigere, Mikhail Smelyanskiy, Liang Xiong, Xuan Zhang
The widespread application of deep learning has changed the landscape of computation in the data center.
18 code implementations • 31 May 2019 • Maxim Naumov, Dheevatsa Mudigere, Hao-Jun Michael Shi, Jianyu Huang, Narayanan Sundaraman, Jongsoo Park, Xiaodong Wang, Udit Gupta, Carole-Jean Wu, Alisson G. Azzolini, Dmytro Dzhulgakov, Andrey Mallevich, Ilia Cherniavskii, Yinghai Lu, Raghuraman Krishnamoorthi, Ansha Yu, Volodymyr Kondratenko, Stephanie Pereira, Xianjie Chen, Wenlin Chen, Vijay Rao, Bill Jia, Liang Xiong, Misha Smelyanskiy
With the advent of deep learning, neural network-based recommendation models have emerged as an important tool for tackling personalization and recommendation tasks.
no code implementations • 29 May 2019 • Dhiraj Kalamkar, Dheevatsa Mudigere, Naveen Mellempudi, Dipankar Das, Kunal Banerjee, Sasikanth Avancha, Dharma Teja Vooturi, Nataraj Jammalamadaka, Jianyu Huang, Hector Yuen, Jiyan Yang, Jongsoo Park, Alexander Heinecke, Evangelos Georganas, Sudarshan Srinivasan, Abhisek Kundu, Misha Smelyanskiy, Bharat Kaul, Pradeep Dubey
In this paper, we discuss the flow of tensors and various key operations in mixed precision training, and delve into details of operations, such as the rounding modes for converting FP32 tensors to BFLOAT16.
no code implementations • 26 Oct 2018 • Majid Jahani, Xi He, Chenxin Ma, Aryan Mokhtari, Dheevatsa Mudigere, Alejandro Ribeiro, Martin Takáč
In this paper, we propose a Distributed Accumulated Newton Conjugate gradiEnt (DANCE) method in which sample size is gradually increasing to quickly obtain a solution whose empirical loss is under satisfactory statistical accuracy.
no code implementations • 10 Aug 2018 • Dharma Teja Vooturi, Dheevatsa Mudigere, Sasikanth Avancha
In this work, we jointly address both accuracy and performance of sparse DNNs using our proposed class of sparse neural networks called HBsNN (Hierarchical Block sparse Neural Networks).
no code implementations • ICML 2018 • Raghu Bollapragada, Dheevatsa Mudigere, Jorge Nocedal, Hao-Jun Michael Shi, Ping Tak Peter Tang
The standard L-BFGS method relies on gradient approximations that are not dominated by noise, so that search directions are descent directions, the line search is reliable, and quasi-Newton updating yields useful quadratic models of the objective function.
no code implementations • ICLR 2018 • Dipankar Das, Naveen Mellempudi, Dheevatsa Mudigere, Dhiraj Kalamkar, Sasikanth Avancha, Kunal Banerjee, Srinivas Sridharan, Karthik Vaidyanathan, Bharat Kaul, Evangelos Georganas, Alexander Heinecke, Pradeep Dubey, Jesus Corbal, Nikita Shustrov, Roma Dubtsov, Evarist Fomenko, Vadim Pirogov
The state-of-the-art (SOTA) for mixed precision training is dominated by variants of low precision floating point operations, and in particular, FP16 accumulating into FP32 Micikevicius et al. (2017).
no code implementations • 24 Jan 2018 • Srinivas Sridharan, Karthikeyan Vaidyanathan, Dhiraj Kalamkar, Dipankar Das, Mikhail E. Smorkalov, Mikhail Shiryaev, Dheevatsa Mudigere, Naveen Mellempudi, Sasikanth Avancha, Bharat Kaul, Pradeep Dubey
The exponential growth in use of large deep neural networks has accelerated the need for training these deep neural networks in hours or even minutes.
1 code implementation • 20 Jul 2017 • Anirban Santara, Abhishek Naik, Balaraman Ravindran, Dipankar Das, Dheevatsa Mudigere, Sasikanth Avancha, Bharat Kaul
Generative Adversarial Imitation Learning (GAIL) is a state-of-the-art algorithm for learning policies when the expert's behavior is available as a fixed set of trajectories.
no code implementations • 15 Jul 2017 • Abhisek Kundu, Kunal Banerjee, Naveen Mellempudi, Dheevatsa Mudigere, Dipankar Das, Bharat Kaul, Pradeep Dubey
Aided by such an elegant trade-off between accuracy and compute, the 8-2 model (8-bit activations, ternary weights), enhanced by ternary residual edges, turns out to be sophisticated enough to achieve very high accuracy ($\sim 1\%$ drop from our FP-32 baseline), despite $\sim 1. 6\times$ reduction in model size, $\sim 26\times$ reduction in number of multiplications, and potentially $\sim 2\times$ power-performance gain comparing to 8-8 representation, on the state-of-the-art deep network ResNet-101 pre-trained on ImageNet dataset.
no code implementations • 2 May 2017 • Naveen Mellempudi, Abhisek Kundu, Dheevatsa Mudigere, Dipankar Das, Bharat Kaul, Pradeep Dubey
We address this by fine-tuning Resnet-50 with 8-bit activations and ternary weights at $N=64$, improving the Top-1 accuracy to within $4\%$ of the full precision result with $<30\%$ additional training overhead.
no code implementations • 31 Jan 2017 • Naveen Mellempudi, Abhisek Kundu, Dipankar Das, Dheevatsa Mudigere, Bharat Kaul
We propose a cluster-based quantization method to convert pre-trained full precision weights into ternary weights with minimal impact on the accuracy.
9 code implementations • 15 Sep 2016 • Nitish Shirish Keskar, Dheevatsa Mudigere, Jorge Nocedal, Mikhail Smelyanskiy, Ping Tak Peter Tang
The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks.
no code implementations • 2 Jun 2016 • Xi He, Dheevatsa Mudigere, Mikhail Smelyanskiy, Martin Takáč
Training deep neural network is a high dimensional and a highly non-convex optimization problem.
no code implementations • 22 Feb 2016 • Dipankar Das, Sasikanth Avancha, Dheevatsa Mudigere, Karthikeyan Vaidynathan, Srinivas Sridharan, Dhiraj Kalamkar, Bharat Kaul, Pradeep Dubey
We design and implement a distributed multinode synchronous SGD algorithm, without altering hyper parameters, or compressing data, or altering algorithmic behavior.
no code implementations • 12 Nov 2014 • S. N. Omkar, Dheevatsa Mudigere, J Senthilnath, M. Vijaya Kumar
Most of the traditional techniques require a model structure to be defined apriori and in case of helicopter dynamics, this is difficult due to its complexity and the interplay between various subsystems. To overcome this difficulty, non-parametric approaches are commonly adopted for helicopter system identification.