Table of Contents
Shortcuts

asteroid.losses.sdr module

class asteroid.losses.sdr.PairwiseNegSDR(sdr_type, zero_mean=True, take_log=True, EPS=1e-08)[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 : \((batch, nsrc, ...)\).
  • targets: \((batch, nsrc, ...)\).
Returns:torch.Tensor – with shape \((batch, nsrc, nsrc)\). 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', EPS=1e-08)[source]

Bases: sphinx.ext.autodoc.importer._MockObject

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

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.
  • reduction (string, optional) – Specifies the reduction to apply to the output: 'none' | 'mean'. 'none': no reduction will be applied, 'mean': the sum of the output will be divided by the number of elements in the output.
Shape:
  • est_targets : \((batch, time)\).
  • targets: \((batch, time)\).
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]
class asteroid.losses.sdr.MultiSrcNegSDR(sdr_type, zero_mean=True, take_log=True, EPS=1e-08)[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 (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 : \((batch, nsrc, time)\).
  • targets: \((batch, nsrc, time)\).
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]
asteroid.losses.sdr.pairwise_neg_sisdr[source]

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 : \((batch, nsrc, ...)\).
  • targets: \((batch, nsrc, ...)\).
Returns:torch.Tensor – with shape \((batch, nsrc, nsrc)\). 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_sdsdr[source]

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 : \((batch, nsrc, ...)\).
  • targets: \((batch, nsrc, ...)\).
Returns:torch.Tensor – with shape \((batch, nsrc, nsrc)\). 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[source]

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 : \((batch, nsrc, ...)\).
  • targets: \((batch, nsrc, ...)\).
Returns:torch.Tensor – with shape \((batch, nsrc, nsrc)\). 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_sisdr[source]

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

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.
  • reduction (string, optional) – Specifies the reduction to apply to the output: 'none' | 'mean'. 'none': no reduction will be applied, 'mean': the sum of the output will be divided by the number of elements in the output.
Shape:
  • est_targets : \((batch, time)\).
  • targets: \((batch, time)\).
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_sdsdr[source]

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

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.
  • reduction (string, optional) – Specifies the reduction to apply to the output: 'none' | 'mean'. 'none': no reduction will be applied, 'mean': the sum of the output will be divided by the number of elements in the output.
Shape:
  • est_targets : \((batch, time)\).
  • targets: \((batch, time)\).
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[source]

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

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.
  • reduction (string, optional) – Specifies the reduction to apply to the output: 'none' | 'mean'. 'none': no reduction will be applied, 'mean': the sum of the output will be divided by the number of elements in the output.
Shape:
  • est_targets : \((batch, time)\).
  • targets: \((batch, time)\).
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.multisrc_neg_sisdr[source]

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

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 : \((batch, nsrc, time)\).
  • targets: \((batch, nsrc, time)\).
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_sdsdr[source]

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

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 : \((batch, nsrc, time)\).
  • targets: \((batch, nsrc, time)\).
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[source]

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

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 : \((batch, nsrc, time)\).
  • targets: \((batch, nsrc, time)\).
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.
Read the Docs v: v0.4.4
Versions
latest
stable
v0.4.4
v0.4.3
v0.4.2
v0.4.1
v0.4.0
v0.3.5_b
v0.3.4
v0.3.3
v0.3.2
v0.3.1
Downloads
On Read the Docs
Project Home
Builds
Free document hosting provided by Read the Docs.