or run

npx @tessl/cli init

Version

Tile

Overview

Evals

Files

docs

audio-io.md datasets.md effects.md functional.md index.md models.md pipelines.md streaming.md transforms.md utils.md

tile.json

tessl/pypi-torchaudio

An audio package for PyTorch providing GPU-accelerated audio I/O operations, signal processing transforms, and machine learning utilities for audio data.

Workspace: tessl
Visibility: Public
Created: 3 months ago
Last updated: 3 months ago
Describes: pkg:pypi/torchaudio@2.8.x

To install, run

npx @tessl/cli install tessl/pypi-torchaudio@2.8.0

TorchAudio

A comprehensive audio processing library for PyTorch that provides GPU-accelerated audio I/O operations, signal processing transforms, and machine learning utilities specifically designed for audio data. TorchAudio supports loading and saving various audio formats, offers dataloaders for common audio datasets, implements essential audio transforms, and provides pre-trained models for speech recognition, synthesis, and source separation.

Package Information

Package Name: torchaudio
Language: Python
Installation: pip install torchaudio
PyTorch Integration: Requires PyTorch to be installed

Core Imports

import torchaudio

Common import patterns:

import torchaudio
import torchaudio.functional as F
import torchaudio.transforms as T
from torchaudio.models import Wav2Vec2Model
from torchaudio.datasets import LIBRISPEECH

Basic Usage

import torchaudio
import torch

# Load audio file
waveform, sample_rate = torchaudio.load("audio.wav")
print(f"Shape: {waveform.shape}, Sample rate: {sample_rate}")

# Apply spectrogram transform
spectrogram_transform = torchaudio.transforms.Spectrogram(
    n_fft=1024, 
    hop_length=512
)
spectrogram = spectrogram_transform(waveform)

# Apply mel spectrogram
mel_transform = torchaudio.transforms.MelSpectrogram(
    sample_rate=sample_rate,
    n_mels=80
)
mel_spectrogram = mel_transform(waveform)

# Save processed audio
torchaudio.save("processed_audio.wav", waveform, sample_rate)

Architecture

TorchAudio is built around several key architectural components:

I/O Operations: Unified interface for loading/saving audio across multiple backends (FFmpeg, SoX, SoundFile)
Functional Processing: Stateless functions for signal processing and audio transformations
Transform Objects: Stateful, PyTorch-compatible transforms that integrate with neural networks
Pre-trained Models: Ready-to-use models for speech recognition, synthesis, and source separation
Dataset Loaders: Standard interfaces for common audio datasets with consistent preprocessing
Pipeline Bundles: Pre-configured model pipelines with preprocessing and post-processing

This design ensures seamless integration with PyTorch's autograd system and enables end-to-end differentiable audio processing pipelines.

Capabilities

Audio I/O Operations

Core functionality for loading, saving, and managing audio files with support for multiple backends and formats. Includes metadata extraction and backend management.

def load(filepath: str, frame_offset: int = 0, num_frames: int = -1, 
         normalize: bool = True, channels_first: bool = True, 
         format: Optional[str] = None) -> Tuple[torch.Tensor, int]: ...

def save(filepath: str, src: torch.Tensor, sample_rate: int, 
         channels_first: bool = True, compression: Optional[float] = None) -> None: ...

def info(filepath: str, format: Optional[str] = None) -> AudioMetaData: ...

class AudioMetaData:
    sample_rate: int
    num_frames: int
    num_channels: int
    bits_per_sample: int
    encoding: str

Audio I/O Operations

Signal Processing Functions

Extensive collection of stateless audio processing functions including spectral analysis, filtering, resampling, pitch manipulation, and advanced signal processing algorithms.

def spectrogram(waveform: torch.Tensor, pad: int = 0, window: torch.Tensor = None,
                n_fft: int = 400, hop_length: Optional[int] = None, 
                win_length: Optional[int] = None, power: Optional[float] = 2.0,
                normalized: bool = False) -> torch.Tensor: ...

def melscale_fbanks(n_freqs: int, f_min: float, f_max: float, n_mels: int,
                    sample_rate: int, norm: Optional[str] = None,
                    mel_scale: str = "htk") -> torch.Tensor: ...

def resample(waveform: torch.Tensor, orig_freq: int, new_freq: int,
             resampling_method: str = "sinc_interp_kaiser") -> torch.Tensor: ...

Signal Processing Functions

Audio Transforms

PyTorch-compatible transform classes for building differentiable audio processing pipelines. Includes spectral transforms, data augmentation, and preprocessing transforms.

class Spectrogram(torch.nn.Module):
    def __init__(self, n_fft: int = 400, win_length: Optional[int] = None,
                 hop_length: Optional[int] = None, pad: int = 0,
                 window_fn: Callable = torch.hann_window, power: Optional[float] = 2.0): ...
    def forward(self, waveform: torch.Tensor) -> torch.Tensor: ...

class MelSpectrogram(torch.nn.Module):
    def __init__(self, sample_rate: int = 16000, n_fft: int = 400,
                 win_length: Optional[int] = None, hop_length: Optional[int] = None,
                 f_min: float = 0., f_max: Optional[float] = None, n_mels: int = 128): ...
    def forward(self, waveform: torch.Tensor) -> torch.Tensor: ...

class MFCC(torch.nn.Module):
    def __init__(self, sample_rate: int = 16000, n_mfcc: int = 40,
                 dct_type: int = 2, norm: str = "ortho", log_mels: bool = False): ...
    def forward(self, waveform: torch.Tensor) -> torch.Tensor: ...

Audio Transforms

Pre-trained Models

Ready-to-use neural network models for speech recognition, synthesis, and source separation. Includes Wav2Vec2, HuBERT, Tacotron2, WaveRNN, and source separation models.

class Wav2Vec2Model(torch.nn.Module):
    def __init__(self, feature_extractor: torch.nn.Module, encoder: torch.nn.Module): ...
    def forward(self, waveforms: torch.Tensor) -> torch.Tensor: ...

def wav2vec2_base(num_out: Optional[int] = None) -> Wav2Vec2Model: ...
def wav2vec2_large(num_out: Optional[int] = None) -> Wav2Vec2Model: ...

class Tacotron2(torch.nn.Module):
    def __init__(self, mask_padding: bool = False, n_mels: int = 80): ...
    def forward(self, tokens: torch.Tensor, token_lengths: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]: ...

class ConvTasNet(torch.nn.Module):
    def __init__(self, num_sources: int = 2, enc_kernel_size: int = 16): ...
    def forward(self, waveforms: torch.Tensor) -> torch.Tensor: ...

Pre-trained Models

Model Pipelines

Pre-configured model bundles with preprocessing, inference, and post-processing for production-ready audio applications. Includes ASR, TTS, and source separation pipelines.

class Wav2Vec2Bundle:
    def get_model(self) -> Wav2Vec2Model: ...
    def get_labels(self) -> List[str]: ...
    sample_rate: int

class Wav2Vec2ASRBundle(Wav2Vec2Bundle):
    def get_model(self) -> Wav2Vec2Model: ...
    def get_decoder(self) -> torch.nn.Module: ...

# Pre-trained bundle instances
WAV2VEC2_ASR_BASE_960H: Wav2Vec2ASRBundle
HUBERT_ASR_LARGE: Wav2Vec2ASRBundle
TACOTRON2_WAVERNN_CHAR_LJSPEECH: Tacotron2TTSBundle

Model Pipelines

Audio Datasets

Standard dataset loaders for common audio datasets with consistent interfaces and preprocessing. Supports speech recognition, synthesis, music analysis, and source separation datasets.

class LIBRISPEECH(torch.utils.data.Dataset):
    def __init__(self, root: str, url: str = "train-clean-100", 
                 folder_in_archive: str = "LibriSpeech", download: bool = False): ...
    def __getitem__(self, n: int) -> Tuple[torch.Tensor, int, str, int, int, int]: ...

class SPEECHCOMMANDS(torch.utils.data.Dataset):
    def __init__(self, root: str, url: str = "speech_commands_v0.02", 
                 folder_in_archive: str = "SpeechCommands", download: bool = False): ...
    def __getitem__(self, n: int) -> Tuple[torch.Tensor, int, str, str, int]: ...

class LJSPEECH(torch.utils.data.Dataset):
    def __init__(self, root: str, url: str = "https://data.keithito.com/data/speech/LJSpeech-1.1.tar.bz2",
                 folder_in_archive: str = "LJSpeech-1.1", download: bool = False): ...
    def __getitem__(self, n: int) -> Tuple[torch.Tensor, int, str, str]: ...

Audio Datasets

Streaming I/O

Advanced streaming capabilities for real-time audio processing, media encoding/decoding, and efficient handling of large audio files.

class StreamReader:
    def __init__(self, src: str, format: Optional[str] = None, option: Optional[Dict[str, str]] = None): ...
    def add_basic_audio_stream(self, frames_per_chunk: int, buffer_chunk_size: int = 3,
                               stream_index: Optional[int] = None, decoder: Optional[str] = None) -> int: ...
    def process_packet(self, timeout: Optional[float] = None, backoff: float = 10.) -> int: ...

class StreamWriter:
    def __init__(self, dst: str, format: Optional[str] = None, option: Optional[Dict[str, str]] = None): ...
    def add_audio_stream(self, sample_rate: int, num_channels: int, format: str = "fltp",
                         encoder: Optional[str] = None, codec_config: Optional[CodecConfig] = None) -> int: ...
    def write_audio_chunk(self, stream_index: int, chunk: torch.Tensor, pts: Optional[int] = None) -> None: ...

Streaming I/O

Audio Effects and Filtering

Comprehensive audio effects processing including filters, EQ, dynamic effects, and spatial audio processing capabilities.

def biquad(waveform: torch.Tensor, b0: float, b1: float, b2: float,
           a0: float, a1: float, a2: float) -> torch.Tensor: ...
def lowpass_biquad(waveform: torch.Tensor, sample_rate: int, cutoff_freq: float, Q: float = 0.707) -> torch.Tensor: ...
def highpass_biquad(waveform: torch.Tensor, sample_rate: int, cutoff_freq: float, Q: float = 0.707) -> torch.Tensor: ...
def flanger(waveform: torch.Tensor, sample_rate: int, delay: float = 0.0,
            depth: float = 2.0, regen: float = 0.0, width: float = 71.0) -> torch.Tensor: ...
def phaser(waveform: torch.Tensor, sample_rate: int, gain_in: float = 0.4,
           gain_out: float = 0.74, delay_ms: float = 3.0, decay: float = 0.4) -> torch.Tensor: ...

Audio Effects and Filtering

Utility Functions

Helper functions for audio file management, format conversion, backend configuration, and integration with other audio processing libraries.

def list_audio_backends() -> List[str]: ...
def get_audio_backend() -> Optional[str]: ...
def set_audio_backend(backend: Optional[str]) -> None: ...

def download_asset(filename: str, subfolder: str = "") -> str: ...

# SoX utilities
def init_sox_effects() -> None: ...
def shutdown_sox_effects() -> None: ...
def effect_names() -> List[str]: ...

Utility Functions

Version

Tile

Files

tessl/pypi-torchaudio

To install, run

index.md.css-3qkkll{font-size:var(--chakra-font-sizes-sm);font-weight:var(--chakra-font-weights-normal);color:var(--chakra-colors-gray-300);}docs/

TorchAudio

Package Information

Core Imports

Basic Usage

Architecture

Capabilities

Audio I/O Operations

Signal Processing Functions

Audio Transforms

Pre-trained Models

Model Pipelines

Audio Datasets

Streaming I/O

Audio Effects and Filtering

Utility Functions

index.mddocs/