asteroid.losses.sdr module¶

class asteroid.losses.sdr.MultiSrcNegSDR(sdr_type, zero_mean=True, take_log=True)[source]¶

Bases: sphinx.ext.autodoc.importer._MockObject

Base class for computing negative SI-SDR, SD-SDR and SNR for a given permutation of source and their estimates.

Parameters:	sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(MultiSrcNegSDR("sisdr"),
>>>                            pit_from='perm_avg')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

forward(est_targets, targets)[source]¶

class asteroid.losses.sdr.NoSrcSDR(*args, **kwargs)[source]¶: Bases: asteroid.losses.sdr.SingleSrcNegSDR, asteroid.utils.deprecation_utils.DeprecationMixin

class asteroid.losses.sdr.NonPitSDR(*args, **kwargs)[source]¶: Bases: asteroid.losses.sdr.MultiSrcNegSDR, asteroid.utils.deprecation_utils.DeprecationMixin

class asteroid.losses.sdr.PairwiseNegSDR(sdr_type, zero_mean=True, take_log=True)[source]¶

Bases: sphinx.ext.autodoc.importer._MockObject

Base class for pairwise negative SI-SDR, SD-SDR and SNR on a batch.

Parameters:	sdr_type (str) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape: [batch, n_src, time]. Batch of target estimates.
targets (torch.Tensor): Expected shape: [batch, n_src, time]. Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch, n_src, n_src]. Pairwise losses.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(PairwiseNegSDR("sisdr"),
>>>                            pit_from='pairwise')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

forward(est_targets, targets)[source]¶

class asteroid.losses.sdr.SingleSrcNegSDR(sdr_type, zero_mean=True, take_log=True, reduction='none')[source]¶

Bases: sphinx.ext.autodoc.importer._MockObject

Base class for single-source negative SI-SDR, SD-SDR and SNR.

Parameters:

sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1].
zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss.
take_log (bool, optional) – by default the log10 of sdr is returned.
reduction (string, optional) – Specifies the reduction to apply to the output:
| 'mean'. 'none' ('none') – no reduction will be applied,
'mean' – the sum of the output will be divided by the number of
in the output. (elements) –

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(SingleSrcNegSDR("sisdr"),
>>>                            pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

forward(est_target, target)[source]¶

asteroid.losses.sdr.multisrc_neg_sdsdr¶

Base class for computing negative SI-SDR, SD-SDR and SNR for a given permutation of source and their estimates.

Parameters:	sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(MultiSrcNegSDR("sisdr"),
>>>                            pit_from='perm_avg')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.multisrc_neg_sisdr¶

Base class for computing negative SI-SDR, SD-SDR and SNR for a given permutation of source and their estimates.

Parameters:	sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(MultiSrcNegSDR("sisdr"),
>>>                            pit_from='perm_avg')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.multisrc_neg_snr¶

Base class for computing negative SI-SDR, SD-SDR and SNR for a given permutation of source and their estimates.

Parameters:	sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(MultiSrcNegSDR("sisdr"),
>>>                            pit_from='perm_avg')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.nonpit_neg_sdsdr¶

Base class for computing negative SI-SDR, SD-SDR and SNR for a given permutation of source and their estimates.

Parameters:	sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(MultiSrcNegSDR("sisdr"),
>>>                            pit_from='perm_avg')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.nonpit_neg_sisdr¶

Base class for computing negative SI-SDR, SD-SDR and SNR for a given permutation of source and their estimates.

Parameters:	sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(MultiSrcNegSDR("sisdr"),
>>>                            pit_from='perm_avg')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.nonpit_neg_snr¶

Base class for computing negative SI-SDR, SD-SDR and SNR for a given permutation of source and their estimates.

Parameters:	sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(MultiSrcNegSDR("sisdr"),
>>>                            pit_from='perm_avg')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.nosrc_neg_sdsdr¶

Base class for single-source negative SI-SDR, SD-SDR and SNR.

Parameters:

sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1].
zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss.
take_log (bool, optional) – by default the log10 of sdr is returned.
reduction (string, optional) – Specifies the reduction to apply to the output:
| 'mean'. 'none' ('none') – no reduction will be applied,
'mean' – the sum of the output will be divided by the number of
in the output. (elements) –

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(SingleSrcNegSDR("sisdr"),
>>>                            pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.nosrc_neg_sisdr¶

Base class for single-source negative SI-SDR, SD-SDR and SNR.

Parameters:

sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1].
zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss.
take_log (bool, optional) – by default the log10 of sdr is returned.
reduction (string, optional) – Specifies the reduction to apply to the output:
| 'mean'. 'none' ('none') – no reduction will be applied,
'mean' – the sum of the output will be divided by the number of
in the output. (elements) –

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(SingleSrcNegSDR("sisdr"),
>>>                            pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.nosrc_neg_snr¶

Base class for single-source negative SI-SDR, SD-SDR and SNR.

Parameters:

sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1].
zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss.
take_log (bool, optional) – by default the log10 of sdr is returned.
reduction (string, optional) – Specifies the reduction to apply to the output:
| 'mean'. 'none' ('none') – no reduction will be applied,
'mean' – the sum of the output will be divided by the number of
in the output. (elements) –

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(SingleSrcNegSDR("sisdr"),
>>>                            pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.pairwise_neg_sdsdr¶

Base class for pairwise negative SI-SDR, SD-SDR and SNR on a batch.

Parameters:	sdr_type (str) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape: [batch, n_src, time]. Batch of target estimates.
targets (torch.Tensor): Expected shape: [batch, n_src, time]. Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch, n_src, n_src]. Pairwise losses.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(PairwiseNegSDR("sisdr"),
>>>                            pit_from='pairwise')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.pairwise_neg_sisdr¶

Base class for pairwise negative SI-SDR, SD-SDR and SNR on a batch.

Parameters:	sdr_type (str) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape: [batch, n_src, time]. Batch of target estimates.
targets (torch.Tensor): Expected shape: [batch, n_src, time]. Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch, n_src, n_src]. Pairwise losses.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(PairwiseNegSDR("sisdr"),
>>>                            pit_from='pairwise')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.pairwise_neg_snr¶

Base class for pairwise negative SI-SDR, SD-SDR and SNR on a batch.

Parameters:	sdr_type (str) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1]. zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss. take_log (bool, optional) – by default the log10 of sdr is returned.

Shape:

est_targets (torch.Tensor): Expected shape: [batch, n_src, time]. Batch of target estimates.
targets (torch.Tensor): Expected shape: [batch, n_src, time]. Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch, n_src, n_src]. Pairwise losses.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(PairwiseNegSDR("sisdr"),
>>>                            pit_from='pairwise')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.singlesrc_neg_sdsdr¶

Base class for single-source negative SI-SDR, SD-SDR and SNR.

Parameters:

sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1].
zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss.
take_log (bool, optional) – by default the log10 of sdr is returned.
reduction (string, optional) – Specifies the reduction to apply to the output:
| 'mean'. 'none' ('none') – no reduction will be applied,
'mean' – the sum of the output will be divided by the number of
in the output. (elements) –

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(SingleSrcNegSDR("sisdr"),
>>>                            pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.singlesrc_neg_sisdr¶

Base class for single-source negative SI-SDR, SD-SDR and SNR.

Parameters:

sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1].
zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss.
take_log (bool, optional) – by default the log10 of sdr is returned.
reduction (string, optional) – Specifies the reduction to apply to the output:
| 'mean'. 'none' ('none') – no reduction will be applied,
'mean' – the sum of the output will be divided by the number of
in the output. (elements) –

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(SingleSrcNegSDR("sisdr"),
>>>                            pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.

asteroid.losses.sdr.singlesrc_neg_snr¶

Base class for single-source negative SI-SDR, SD-SDR and SNR.

Parameters:

sdr_type (string) – choose between “snr” for plain SNR, “sisdr” for SI-SDR and “sdsdr” for SD-SDR [1].
zero_mean (bool, optional) – by default it zero mean the target and estimate before computing the loss.
take_log (bool, optional) – by default the log10 of sdr is returned.
reduction (string, optional) – Specifies the reduction to apply to the output:
| 'mean'. 'none' ('none') – no reduction will be applied,
'mean' – the sum of the output will be divided by the number of
in the output. (elements) –

Shape:

est_targets (torch.Tensor): Expected shape [batch, time].: Batch of target estimates.
targets (torch.Tensor): Expected shape [batch, time].: Batch of training targets.

Returns:	`torch.Tensor` – with shape [batch] if reduction=’none’ else [] scalar if reduction=’mean’.

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(SingleSrcNegSDR("sisdr"),
>>>                            pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

References

[1] Le Roux, Jonathan, et al. “SDR half-baked or well done.” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2019.