TorchAudio | PyTorch Developer Day 2020
TorchAudio provides reusable, orthogonal, correct, and performant building blocks for cutting-edge experimentation in the audio domain. In this talk, machine learning scientist Vince Quenneville-Belair examines how the broad range of PyTorch capabilities — such as JIT, quantization, distributed, and mobile — enable seamless research-to-production for core end-to-end applications, such as speech recognition text-to-speech, and source separation.
