William J. Dally

VS-Quant: Per-vector Scaled Quantization for Accurate Low-Precision Neural Network Inference

Feb 08, 2021
Steve Dai, Rangharajan Venkatesan, Haoxing Ren, Brian Zimmer, William J. Dally, Brucek Khailany

Figure 2 for VS-Quant: Per-vector Scaled Quantization for Accurate Low-Precision Neural Network Inference

Figure 3 for VS-Quant: Per-vector Scaled Quantization for Accurate Low-Precision Neural Network Inference

Figure 4 for VS-Quant: Per-vector Scaled Quantization for Accurate Low-Precision Neural Network Inference

Access Paper or Ask Questions

A Delay Metric for Video Object Detection: What Average Precision Fails to Tell

Aug 18, 2019
Huizi Mao, Xiaodong Yang, William J. Dally

Figure 2 for A Delay Metric for Video Object Detection: What Average Precision Fails to Tell

Figure 3 for A Delay Metric for Video Object Detection: What Average Precision Fails to Tell

Figure 4 for A Delay Metric for Video Object Detection: What Average Precision Fails to Tell

* ICCV 2019

Access Paper or Ask Questions

CaTDet: Cascaded Tracked Detector for Efficient Object Detection from Video

Sep 30, 2018
Huizi Mao, Taeyoung Kong, William J. Dally

Figure 2 for CaTDet: Cascaded Tracked Detector for Efficient Object Detection from Video

Figure 3 for CaTDet: Cascaded Tracked Detector for Efficient Object Detection from Video

Figure 4 for CaTDet: Cascaded Tracked Detector for Efficient Object Detection from Video

Access Paper or Ask Questions

Efficient Sparse-Winograd Convolutional Neural Networks

Feb 18, 2018
Xingyu Liu, Jeff Pool, Song Han, William J. Dally

Figure 2 for Efficient Sparse-Winograd Convolutional Neural Networks

Figure 3 for Efficient Sparse-Winograd Convolutional Neural Networks

Figure 4 for Efficient Sparse-Winograd Convolutional Neural Networks

* Published as a conference paper at ICLR 2018

Access Paper or Ask Questions

Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training

Feb 05, 2018
Yujun Lin, Song Han, Huizi Mao, Yu Wang, William J. Dally

Figure 2 for Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training

Figure 3 for Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training

Figure 4 for Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training

* ICLR 2018
* we find 99.9% of the gradient exchange in distributed SGD is redundant; we reduce the communication bandwidth by two orders of magnitude without losing accuracy

Access Paper or Ask Questions

Exploring the Regularity of Sparse Structure in Convolutional Neural Networks

Jun 05, 2017
Huizi Mao, Song Han, Jeff Pool, Wenshuo Li, Xingyu Liu, Yu Wang, William J. Dally

Figure 2 for Exploring the Regularity of Sparse Structure in Convolutional Neural Networks

Figure 3 for Exploring the Regularity of Sparse Structure in Convolutional Neural Networks

Figure 4 for Exploring the Regularity of Sparse Structure in Convolutional Neural Networks

* submitted to NIPS 2017

Access Paper or Ask Questions

SCNN: An Accelerator for Compressed-sparse Convolutional Neural Networks

May 23, 2017
Angshuman Parashar, Minsoo Rhu, Anurag Mukkara, Antonio Puglielli, Rangharajan Venkatesan, Brucek Khailany, Joel Emer, Stephen W. Keckler, William J. Dally

Figure 2 for SCNN: An Accelerator for Compressed-sparse Convolutional Neural Networks

Figure 3 for SCNN: An Accelerator for Compressed-sparse Convolutional Neural Networks

Figure 4 for SCNN: An Accelerator for Compressed-sparse Convolutional Neural Networks

Access Paper or Ask Questions

Trained Ternary Quantization

Feb 23, 2017
Chenzhuo Zhu, Song Han, Huizi Mao, William J. Dally

Figure 2 for Trained Ternary Quantization

Figure 3 for Trained Ternary Quantization

Figure 4 for Trained Ternary Quantization

* Accepted for Poster Presentation on ICLR 2017

Access Paper or Ask Questions

DSD: Dense-Sparse-Dense Training for Deep Neural Networks

Feb 21, 2017
Song Han, Jeff Pool, Sharan Narang, Huizi Mao, Enhao Gong, Shijian Tang, Erich Elsen, Peter Vajda, Manohar Paluri, John Tran, Bryan Catanzaro, William J. Dally

Figure 2 for DSD: Dense-Sparse-Dense Training for Deep Neural Networks

Figure 3 for DSD: Dense-Sparse-Dense Training for Deep Neural Networks

Figure 4 for DSD: Dense-Sparse-Dense Training for Deep Neural Networks

* Published as a conference paper at ICLR 2017

Access Paper or Ask Questions

ESE: Efficient Speech Recognition Engine with Sparse LSTM on FPGA

Feb 20, 2017
Song Han, Junlong Kang, Huizi Mao, Yiming Hu, Xin Li, Yubin Li, Dongliang Xie, Hong Luo, Song Yao, Yu Wang, Huazhong Yang, William J. Dally

Figure 2 for ESE: Efficient Speech Recognition Engine with Sparse LSTM on FPGA

Figure 3 for ESE: Efficient Speech Recognition Engine with Sparse LSTM on FPGA

Figure 4 for ESE: Efficient Speech Recognition Engine with Sparse LSTM on FPGA

* Accepted as full paper in FPGA'17, Monterey, CA; Also appeared at 1st International Workshop on Efficient Methods for Deep Neural Networks at NIPS 2016, Barcelona, Spain

Access Paper or Ask Questions

SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size

Nov 04, 2016
Forrest N. Iandola, Song Han, Matthew W. Moskewicz, Khalid Ashraf, William J. Dally, Kurt Keutzer

Figure 2 for SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size

Figure 3 for SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size

Figure 4 for SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size

* In ICLR Format

Access Paper or Ask Questions

EIE: Efficient Inference Engine on Compressed Deep Neural Network

May 03, 2016
Song Han, Xingyu Liu, Huizi Mao, Jing Pu, Ardavan Pedram, Mark A. Horowitz, William J. Dally

Figure 2 for EIE: Efficient Inference Engine on Compressed Deep Neural Network

Figure 3 for EIE: Efficient Inference Engine on Compressed Deep Neural Network

Figure 4 for EIE: Efficient Inference Engine on Compressed Deep Neural Network

* External Links: TheNextPlatform: http://goo.gl/f7qX0L ; O'Reilly: https://goo.gl/Id1HNT ; Hacker News: https://goo.gl/KM72SV ; Embedded-vision: http://goo.gl/joQNg8 ; Talk at NVIDIA GTC'16: http://goo.gl/6wJYvn ; Talk at Embedded Vision Summit: https://goo.gl/7abFNe ; Talk at Stanford University: https://goo.gl/6lwuer. Published as a conference paper in ISCA 2016

Access Paper or Ask Questions

Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding

Feb 15, 2016
Song Han, Huizi Mao, William J. Dally

Figure 2 for Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding

Figure 3 for Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding

Figure 4 for Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding

* Published as a conference paper at ICLR 2016 (oral)

Access Paper or Ask Questions

Learning both Weights and Connections for Efficient Neural Networks

Oct 30, 2015
Song Han, Jeff Pool, John Tran, William J. Dally

Figure 2 for Learning both Weights and Connections for Efficient Neural Networks

Figure 3 for Learning both Weights and Connections for Efficient Neural Networks

Figure 4 for Learning both Weights and Connections for Efficient Neural Networks

* Published as a conference paper at NIPS 2015

Access Paper or Ask Questions