Alert button
Picture for Oleg Rybakov

Oleg Rybakov

Alert button

USM-Lite: Quantization and Sparsity Aware Fine-tuning for Speech Recognition with Universal Speech Models

Jan 03, 2024
Shaojin Ding, David Qiu, David Rim, Yanzhang He, Oleg Rybakov, Bo Li, Rohit Prabhavalkar, Weiran Wang, Tara N. Sainath, Shivani Agrawal, Zhonglin Han, Jian Li, Amir Yazdanbakhsh

Figure 1 for USM-Lite: Quantization and Sparsity Aware Fine-tuning for Speech Recognition with Universal Speech Models
Figure 2 for USM-Lite: Quantization and Sparsity Aware Fine-tuning for Speech Recognition with Universal Speech Models
Figure 3 for USM-Lite: Quantization and Sparsity Aware Fine-tuning for Speech Recognition with Universal Speech Models
Figure 4 for USM-Lite: Quantization and Sparsity Aware Fine-tuning for Speech Recognition with Universal Speech Models
Viaarxiv icon

2-bit Conformer quantization for automatic speech recognition

May 26, 2023
Oleg Rybakov, Phoenix Meadowlark, Shaojin Ding, David Qiu, Jian Li, David Rim, Yanzhang He

Figure 1 for 2-bit Conformer quantization for automatic speech recognition
Figure 2 for 2-bit Conformer quantization for automatic speech recognition
Figure 3 for 2-bit Conformer quantization for automatic speech recognition
Figure 4 for 2-bit Conformer quantization for automatic speech recognition
Viaarxiv icon

RAND: Robustness Aware Norm Decay For Quantized Seq2seq Models

May 24, 2023
David Qiu, David Rim, Shaojin Ding, Oleg Rybakov, Yanzhang He

Figure 1 for RAND: Robustness Aware Norm Decay For Quantized Seq2seq Models
Figure 2 for RAND: Robustness Aware Norm Decay For Quantized Seq2seq Models
Figure 3 for RAND: Robustness Aware Norm Decay For Quantized Seq2seq Models
Figure 4 for RAND: Robustness Aware Norm Decay For Quantized Seq2seq Models
Viaarxiv icon

STEP: Learning N:M Structured Sparsity Masks from Scratch with Precondition

Feb 02, 2023
Yucheng Lu, Shivani Agrawal, Suvinay Subramanian, Oleg Rybakov, Christopher De Sa, Amir Yazdanbakhsh

Figure 1 for STEP: Learning N:M Structured Sparsity Masks from Scratch with Precondition
Figure 2 for STEP: Learning N:M Structured Sparsity Masks from Scratch with Precondition
Figure 3 for STEP: Learning N:M Structured Sparsity Masks from Scratch with Precondition
Figure 4 for STEP: Learning N:M Structured Sparsity Masks from Scratch with Precondition
Viaarxiv icon

Streaming Parrotron for on-device speech-to-speech conversion

Oct 25, 2022
Oleg Rybakov, Fadi Biadsy, Xia Zhang, Liyang Jiang, Phoenix Meadowlark, Shivani Agrawal

Figure 1 for Streaming Parrotron for on-device speech-to-speech conversion
Figure 2 for Streaming Parrotron for on-device speech-to-speech conversion
Figure 3 for Streaming Parrotron for on-device speech-to-speech conversion
Figure 4 for Streaming Parrotron for on-device speech-to-speech conversion
Viaarxiv icon

4-bit Conformer with Native Quantization Aware Training for Speech Recognition

Mar 29, 2022
Shaojin Ding, Phoenix Meadowlark, Yanzhang He, Lukasz Lew, Shivani Agrawal, Oleg Rybakov

Figure 1 for 4-bit Conformer with Native Quantization Aware Training for Speech Recognition
Figure 2 for 4-bit Conformer with Native Quantization Aware Training for Speech Recognition
Figure 3 for 4-bit Conformer with Native Quantization Aware Training for Speech Recognition
Figure 4 for 4-bit Conformer with Native Quantization Aware Training for Speech Recognition
Viaarxiv icon

A Scalable Model Specialization Framework for Training and Inference using Submodels and its Application to Speech Model Personalization

Mar 23, 2022
Fadi Biadsy, Youzheng Chen, Xia Zhang, Oleg Rybakov, Andrew Rosenberg, Pedro J. Moreno

Figure 1 for A Scalable Model Specialization Framework for Training and Inference using Submodels and its Application to Speech Model Personalization
Figure 2 for A Scalable Model Specialization Framework for Training and Inference using Submodels and its Application to Speech Model Personalization
Figure 3 for A Scalable Model Specialization Framework for Training and Inference using Submodels and its Application to Speech Model Personalization
Viaarxiv icon

Real time spectrogram inversion on mobile phone

Mar 10, 2022
Oleg Rybakov, Marco Tagliasacchi, Yunpeng Li, Liyang Jiang, Xia Zhang, Fadi Biadsy

Figure 1 for Real time spectrogram inversion on mobile phone
Figure 2 for Real time spectrogram inversion on mobile phone
Figure 3 for Real time spectrogram inversion on mobile phone
Figure 4 for Real time spectrogram inversion on mobile phone
Viaarxiv icon

Pareto-Optimal Quantized ResNet Is Mostly 4-bit

May 07, 2021
AmirAli Abdolrashidi, Lisa Wang, Shivani Agrawal, Jonathan Malmaud, Oleg Rybakov, Chas Leichner, Lukasz Lew

Figure 1 for Pareto-Optimal Quantized ResNet Is Mostly 4-bit
Figure 2 for Pareto-Optimal Quantized ResNet Is Mostly 4-bit
Figure 3 for Pareto-Optimal Quantized ResNet Is Mostly 4-bit
Figure 4 for Pareto-Optimal Quantized ResNet Is Mostly 4-bit
Viaarxiv icon