kaira.losses.audio.MelSpectrogramLoss

Inheritance diagram of MelSpectrogramLoss — Inheritance diagram for MelSpectrogramLoss

class kaira.losses.audio.MelSpectrogramLoss(sample_rate=22050, n_fft=1024, hop_length=256, n_mels=80, f_min=0.0, f_max=8000.0, log_mel=True)[source]

Bases: BaseLoss

Mel-Spectrogram Loss Module.

This module calculates the loss between mel-spectrograms of input and target audio.

Methods

`__init__`	Initialize the MelSpectrogramLoss module.
`forward`	Forward pass through the MelSpectrogramLoss module.

__init__(sample_rate=22050, n_fft=1024, hop_length=256, n_mels=80, f_min=0.0, f_max=8000.0, log_mel=True)[source]

Initialize the MelSpectrogramLoss module.

Parameters:

sample_rate (int) – Audio sample rate. Default is 22050.
n_fft (int) – FFT size. Default is 1024.
hop_length (int) – Hop size. Default is 256.
n_mels (int) – Number of mel bands. Default is 80.
f_min (float) – Minimum frequency. Default is 0.0.
f_max (float) – Maximum frequency. Default is 8000.0.
log_mel (bool) – Whether to use log-mel spectrogram. Default is True.

forward(x: Tensor, target: Tensor) → Tensor[source]

Forward pass through the MelSpectrogramLoss module.

Parameters:

x (torch.Tensor) – The input audio tensor.
target (torch.Tensor) – The target audio tensor.

Returns:

The mel-spectrogram loss.

Return type:

torch.Tensor