TorchAudio
summary
Library that reads mp3 wav flac sphere files, loads datasets like LibriSpeech and LibriTTS, and exposes tensors via torchaudio.load for GPU/JIT use. It implements spectrograms, biquad filters, VAD and specaugment as nn.Modules so pipelines compile with torch.jit and run on mobile. Includes SoX and Kaldi bindings, example training pipelines: wav2letter on LibriSpeech, WaveRNN vocoder on LibriTTS, ConvTasNet on WSJ0‑2mix, and these examples are runnable end to end.
AI-written summary · about
chapters (7)
related talks
05mTorchaudio 0.3 with Kaldi Compatibility, New Transforms: A Quick Introduction
40mGathering Insights from Audio Data
36mNVIDIA NeMo: Toolkit for Conversational AI
06mTorchtext 0.4 with Supervised Learning Datasets: A Quick Introduction
14mLightning Talk: Enabling the Audio Modality for Language Models
10mPyTorch at Dolby Labs
22mVocalPy: a core Python package for acoustic comms research
15mOptimizing inference for voice models in production