Presentation: wav2letter++: Facebook's Fast Open-Source Speech Recognition System
Share this on:
This presentation is now available to view on InfoQ.com
Watch video with transcriptAbstract
In this talk I will introduce wav2letter++, a fast open-source deep learning speech recognition framework. wav2letter++ is written entirely in C++, and uses the ArrayFire tensor library for maximum efficiency. I will explain the architecture and design of the wav2letter++ system and compare it to other major open-source speech recognition systems. In some cases wav2letter++ is more than 2x faster than other optimized frameworks for training end-to-end neural networks for speech recognition. I will also show that wav2letter++'s training times scale linearly to 64 GPUs, the highest that has been tested, for models with 100 million parameters. High-performance frameworks enable fast iteration, which is often a crucial factor in successful research and model tuning on new datasets and tasks.
2019 Tracks
-
AI Meets the Physical World
Where AI touches the physical world, think drones, ROS, NVidia, TPU and more.
-
Hands-on Codelabs & Speakers Office Hours
Codelabs are a self-guided tutorial of a product, API, or tool kit followed by an Office Hour period with the lab’s creator.
-
Papers in Production: Modern CS in the Real World
Groundbreaking papers make real-world impact.
-
Hands-on Codelabs & Speakers Office Hours
Codelabs are a self-guided tutorial of a product, API, or tool kit followed by an Office Hour period with the lab’s creator.
-
Solving Software Engineering Problems with Machine Learning
Interesting machine learning use cases changing how we develop software today, including planned topics touching on infrastructure optimization, developer experience, security, and more.
-
Predictive Architectures in the Real World
Case Study focused look at end to end predictive pipelines from places like Salesforce, Uber, Linkedin, & Netflix.
-
Predictive Data Pipelines & Architectures
Case Study focused look at end to end predictive pipelines from places like Salesforce, Uber, Linkedin, & Netflix
-
Sequential Data: Natural Language, Time Series, and Sound
Techniques, practices, and approaches around time series and sequential data. Expect topics including image recognition, NLP/NLU, preprocess, & crunching of related algorithms.
-
ML in Action
Applied track demonstrating how to train, score, and handle common machine learning use cases, including heavy concentration in the space of security and fraud