Signal to Distortion Ratio (SDR)

Module Interface

torchmetrics.SignalDistortionRatio

alias of torchmetrics.audio._deprecated._SignalDistortionRatio

Functional Interface

torchmetrics.functional.signal_distortion_ratio(preds, target, use_cg_iter=None, filter_length=512, zero_mean=False, load_diag=None)

Wrapper for deprecated import.

>>> import torch
>>> g = torch.manual_seed(1)
>>> preds = torch.randn(8000)
>>> target = torch.randn(8000)
>>> _signal_distortion_ratio(preds, target)
tensor(-12.0589)
>>> # use with permutation_invariant_training
>>> preds = torch.randn(4, 2, 8000)  # [batch, spk, time]
>>> target = torch.randn(4, 2, 8000)
>>> best_metric, best_perm = _permutation_invariant_training(preds, target, _signal_distortion_ratio)
>>> best_metric
tensor([-11.6375, -11.4358, -11.7148, -11.6325])
>>> best_perm
tensor([[1, 0],
        [0, 1],
        [1, 0],
        [0, 1]])

Return type

Tensor