lightly.loss

The lightly.loss package provides loss functions for self-supervised learning.

class lightly.loss.barlow_twins_loss.BarlowTwinsLoss(lambda_param: float = 0.005, gather_distributed: bool = False)

Implementation of the Barlow Twins Loss from Barlow Twins[0] paper. This code specifically implements the Figure Algorithm 1 from [0].

[0] Zbontar,J. et.al, 2021, Barlow Twins… https://arxiv.org/abs/2103.03230

Examples:

>>> # initialize loss function
>>> loss_fn = BarlowTwinsLoss()
>>>
>>> # generate two random transforms of images
>>> t0 = transforms(images)
>>> t1 = transforms(images)
>>>
>>> # feed through SimSiam model
>>> out0, out1 = model(t0, t1)
>>>
>>> # calculate loss
>>> loss = loss_fn(out0, out1)
forward(z_a: torch.Tensor, z_b: torch.Tensor) → torch.Tensor

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class lightly.loss.dcl_loss.DCLLoss(temperature: float = 0.1, weight_fn: Optional[Callable[[torch.Tensor, torch.Tensor], torch.Tensor]] = None, gather_distributed: bool = False)

Implementation of the Decoupled Contrastive Learning Loss from Decoupled Contrastive Learning [0].

This code implements Equation 6 in [0], including the sum over all images i and views k. The loss is reduced to a mean loss over the mini-batch. The implementation was inspired by [1].

temperature

Similarities are scaled by inverse temperature.

weight_fn

Weighting function w from the paper. Scales the loss between the positive views (views from the same image). No weighting is performed if weight_fn is None. The function must take the two input tensors passed to the forward call as input and return a weight tensor. The returned weight tensor must have the same length as the input tensors.

gather_distributed

If True then negatives from all gpus are gathered before the loss calculation.

Examples

>>> loss_fn = DCLLoss(temperature=0.07)
>>>
>>> # generate two random transforms of images
>>> t0 = transforms(images)
>>> t1 = transforms(images)
>>>
>>> # embed images using some model, for example SimCLR
>>> out0 = model(t0)
>>> out1 = model(t1)
>>>
>>> # calculate loss
>>> loss = loss_fn(out0, out1)
>>>
>>> # you can also add a custom weighting function
>>> weight_fn = lambda out0, out1: torch.sum((out0 - out1) ** 2, dim=1)
>>> loss_fn = DCLLoss(weight_fn=weight_fn)
forward(out0: torch.Tensor, out1: torch.Tensor) → torch.Tensor

Forward pass of the DCL loss.

Parameters

  • out0 – Output projections of the first set of transformed images. Shape: (batch_size, embedding_size)

  • out1 – Output projections of the second set of transformed images. Shape: (batch_size, embedding_size)

Returns

Mean loss over the mini-batch.

class lightly.loss.dcl_loss.DCLWLoss(temperature: float = 0.1, sigma: float = 0.5, gather_distributed: bool = False)

Implementation of the Weighted Decoupled Contrastive Learning Loss from Decoupled Contrastive Learning [0].

This code implements Equation 6 in [0] with a negative Mises-Fisher weighting function. The loss returns the mean over all images i and views k in the mini-batch. The implementation was inspired by [1].

temperature

Similarities are scaled by inverse temperature.

sigma

Similar to temperature but applies the inverse scaling in the weighting function.

gather_distributed

If True then negatives from all gpus are gathered before the loss calculation.

Examples

>>> loss_fn = DCLWLoss(temperature=0.07)
>>>
>>> # generate two random transforms of images
>>> t0 = transforms(images)
>>> t1 = transforms(images)
>>>
>>> # embed images using some model, for example SimCLR
>>> out0 = model(t0)
>>> out1 = model(t1)
>>>
>>> # calculate loss
>>> loss = loss_fn(out0, out1)
class lightly.loss.dino_loss.DINOLoss(output_dim: int = 65536, warmup_teacher_temp: float = 0.04, teacher_temp: float = 0.04, warmup_teacher_temp_epochs: int = 30, student_temp: float = 0.1, center_momentum: float = 0.9)

Implementation of the loss described in ‘Emerging Properties in Self-Supervised Vision Transformers’. [0]

This implementation follows the code published by the authors. [1] It supports global and local image crops. A linear warmup schedule for the teacher temperature is implemented to stabilize training at the beginning. Centering is applied to the teacher output to avoid model collapse.

output_dim

Dimension of the model output.

warmup_teacher_temp

Initial value of the teacher temperature. Should be decreased if the training loss does not decrease.

teacher_temp

Final value of the teacher temperature after linear warmup. Values above 0.07 result in unstable behavior in most cases. Can be slightly increased to improve performance during finetuning.

warmup_teacher_temp_epochs

Number of epochs for the teacher temperature warmup.

student_temp

Temperature of the student.

center_momentum

Momentum term for the center calculation.

Examples

>>> # initialize loss function
>>> loss_fn = DINOLoss(128)
>>>
>>> # generate a view of the images with a random transform
>>> view = transform(images)
>>>
>>> # embed the view with a student and teacher model
>>> teacher_out = teacher(view)
>>> student_out = student(view)
>>>
>>> # calculate loss
>>> loss = loss_fn([teacher_out], [student_out], epoch=0)
forward(teacher_out: List[torch.Tensor], student_out: List[torch.Tensor], epoch: int) → torch.Tensor

Cross-entropy between softmax outputs of the teacher and student networks.

Parameters

  • teacher_out – List of view feature tensors from the teacher model. Each tensor is assumed to contain features from one view of the batch and have length batch_size.

  • student_out – List of view feature tensors from the student model. Each tensor is assumed to contain features from one view of the batch and have length batch_size.

  • epoch – The current training epoch.

Returns

The average cross-entropy loss.

update_center(teacher_out: torch.Tensor) → None

Moving average update of the center used for the teacher output.

Parameters

teacher_out – Stacked output from the teacher model.

class lightly.loss.hypersphere_loss.HypersphereLoss(t=1.0, lam=1.0, alpha=2.0)

Implementation of the loss described in ‘Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere.’ [0]

[0] Tongzhou Wang. et.al, 2020, … https://arxiv.org/abs/2005.10242

Note

In order for this loss to function as advertized, an l1-normalization to the hypersphere is required. This loss function applies this l1-normalization internally in the loss-layer. However, it is recommended that the same normalization is also applied in your architecture, considering that this l1-loss is also intended to be applied during inference. Perhaps there may be merit in leaving it out of the inferrence pathway, but this use has not been tested.

Moreover it is recommended that the layers preceeding this loss function are either a linear layer without activation, a batch-normalization layer, or both. The directly upstream architecture can have a large influence on the ability of this loss to achieve its stated aim of promoting uniformity on the hypersphere; and if by contrast the last layer going into the embedding is a RELU or similar nonlinearity, we may see that we will never get very close to achieving the goal of uniformity on the hypersphere, but will confine ourselves to the subspace of positive activations. Similar architectural considerations are relevant to most contrastive loss functions, but we call it out here explicitly.

Examples

>>> # initialize loss function
>>> loss_fn = HypersphereLoss()
>>>
>>> # generate two random transforms of images
>>> t0 = transforms(images)
>>> t1 = transforms(images)
>>>
>>> # feed through SimSiam model
>>> out0, out1 = model(t0, t1)
>>>
>>> # calculate loss
>>> loss = loss_fn(out0, out1)
forward(z_a: torch.Tensor, z_b: torch.Tensor) → torch.Tensor
Parameters

  • x (torch.Tensor, [b, d], float)

  • y (torch.Tensor, [b, d], float)

Returns

Loss (torch.Tensor, [], float)

class lightly.loss.memory_bank.MemoryBankModule(size: int = 65536)

Memory bank implementation

This is a parent class to all loss functions implemented by the lightly Python package. This way, any loss can be used with a memory bank if desired.

size

Number of keys the memory bank can store. If set to 0, memory bank is not used.

Examples

>>> class MyLossFunction(MemoryBankModule):
>>>
>>>     def __init__(self, memory_bank_size: int = 2 ** 16):
>>>         super(MyLossFunction, self).__init__(memory_bank_size)
>>>
>>>     def forward(self, output: torch.Tensor,
>>>                 labels: torch.Tensor = None):
>>>
>>>         output, negatives = super(
>>>             MyLossFunction, self).forward(output)
>>>
>>>         if negatives is not None:
>>>             # evaluate loss with negative samples
>>>         else:
>>>             # evaluate loss without negative samples
forward(output: torch.Tensor, labels: Optional[torch.Tensor] = None, update: bool = False)

Query memory bank for additional negative samples

Parameters

  • output – The output of the model.

  • labels – Should always be None, will be ignored.

Returns

The output if the memory bank is of size 0, otherwise the output and the entries from the memory bank.

class lightly.loss.msn_loss.MSNLoss(temperature: float = 0.1, sinkhorn_iterations: int = 3, regularization_weight: float = 1.0, me_max_weight: Optional[float] = None, gather_distributed: bool = False)

Implementation of the loss function from MSN [0].

Code inspired by [1].

temperature

Similarities between anchors and targets are scaled by the inverse of the temperature. Must be in (0, inf).

sinkhorn_iterations

Number of sinkhorn normalization iterations on the targets.

regularization_weight

Weight factor lambda by which the regularization loss is scaled. Set to 0 to disable regularization.

me_max_weight

Deprecated, use regularization_weight instead. Takes precendence over regularization_weight if not None. Weight factor lambda by which the mean entropy maximization regularization loss is scaled. Set to 0 to disable mean entropy maximization reguliarization.

gather_distributed

If True, then target probabilities are gathered from all GPUs.

Examples:

>>> # initialize loss function
>>> loss_fn = MSNLoss()
>>>
>>> # generate anchors and targets of images
>>> anchors = transforms(images)
>>> targets = transforms(images)
>>>
>>> # feed through MSN model
>>> anchors_out = model(anchors)
>>> targets_out = model.target(targets)
>>>
>>> # calculate loss
>>> loss = loss_fn(anchors_out, targets_out, prototypes=model.prototypes)
forward(anchors: torch.Tensor, targets: torch.Tensor, prototypes: torch.Tensor, target_sharpen_temperature: float = 0.25) → torch.Tensor

Computes the MSN loss for a set of anchors, targets and prototypes.

Parameters

  • anchors – Tensor with shape (batch_size * anchor_views, dim).

  • targets – Tensor with shape (batch_size, dim).

  • prototypes – Tensor with shape (num_prototypes, dim).

  • target_sharpen_temperature – Temperature used to sharpen the target probabilities.

Returns

Mean loss over all anchors.

regularization_loss(mean_anchor_probs: torch.Tensor) → torch.Tensor

Calculates mean entropy regularization loss.

class lightly.loss.negative_cosine_similarity.NegativeCosineSimilarity(dim: int = 1, eps: float = 1e-08)

Implementation of the Negative Cosine Simililarity used in the SimSiam[0] paper.

[0] SimSiam, 2020, https://arxiv.org/abs/2011.10566

Examples

>>> # initialize loss function
>>> loss_fn = NegativeCosineSimilarity()
>>>
>>> # generate two representation tensors
>>> # with batch size 10 and dimension 128
>>> x0 = torch.randn(10, 128)
>>> x1 = torch.randn(10, 128)
>>>
>>> # calculate loss
>>> loss = loss_fn(x0, x1)
forward(x0: torch.Tensor, x1: torch.Tensor) → torch.Tensor

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class lightly.loss.ntx_ent_loss.NTXentLoss(temperature: float = 0.5, memory_bank_size: int = 0, gather_distributed: bool = False)

Implementation of the Contrastive Cross Entropy Loss.

This implementation follows the SimCLR[0] paper. If you enable the memory bank by setting the memory_bank_size value > 0 the loss behaves like the one described in the MoCo[1] paper.

temperature

Scale logits by the inverse of the temperature.

memory_bank_size

Number of negative samples to store in the memory bank. Use 0 for SimCLR. For MoCo we typically use numbers like 4096 or 65536.

gather_distributed

If True then negatives from all gpus are gathered before the loss calculation. This flag has no effect if memory_bank_size > 0.

Raises

ValueError – If abs(temperature) < 1e-8 to prevent divide by zero.

Examples

>>> # initialize loss function without memory bank
>>> loss_fn = NTXentLoss(memory_bank_size=0)
>>>
>>> # generate two random transforms of images
>>> t0 = transforms(images)
>>> t1 = transforms(images)
>>>
>>> # feed through SimCLR or MoCo model
>>> batch = torch.cat((t0, t1), dim=0)
>>> output = model(batch)
>>>
>>> # calculate loss
>>> loss = loss_fn(output)
forward(out0: torch.Tensor, out1: torch.Tensor)

Forward pass through Contrastive Cross-Entropy Loss.

If used with a memory bank, the samples from the memory bank are used as negative examples. Otherwise, within-batch samples are used as negative samples.

Parameters

  • out0 – Output projections of the first set of transformed images. Shape: (batch_size, embedding_size)

  • out1 – Output projections of the second set of transformed images. Shape: (batch_size, embedding_size)

Returns

Contrastive Cross Entropy Loss value.

class lightly.loss.pmsn_loss.PMSNLoss(temperature: float = 0.1, sinkhorn_iterations: int = 3, regularization_weight: float = 1, power_law_exponent: float = 0.25, gather_distributed: bool = False)

Implementation of the loss function from PMSN [0] using a power law target distribution.

temperature

Similarities between anchors and targets are scaled by the inverse of the temperature. Must be in (0, inf).

sinkhorn_iterations

Number of sinkhorn normalization iterations on the targets.

regularization_weight

Weight factor lambda by which the regularization loss is scaled. Set to 0 to disable regularization.

power_law_exponent

Exponent for power law distribution. Entry k of the distribution is proportional to (1 / k) ^ power_law_exponent, with k ranging from 1 to dim + 1.

gather_distributed

If True, then target probabilities are gathered from all GPUs.

Examples:

>>> # initialize loss function
>>> loss_fn = PMSNLoss()
>>>
>>> # generate anchors and targets of images
>>> anchors = transforms(images)
>>> targets = transforms(images)
>>>
>>> # feed through PMSN model
>>> anchors_out = model(anchors)
>>> targets_out = model.target(targets)
>>>
>>> # calculate loss
>>> loss = loss_fn(anchors_out, targets_out, prototypes=model.prototypes)
regularization_loss(mean_anchor_probs: torch.Tensor) → torch.Tensor

Calculates regularization loss with a power law target distribution.

class lightly.loss.pmsn_loss.PMSNCustomLoss(target_distribution: Callable[[torch.Tensor], torch.Tensor], temperature: float = 0.1, sinkhorn_iterations: int = 3, regularization_weight: float = 1, gather_distributed: bool = False)

Implementation of the loss function from PMSN [0] with a custom target distribution.

target_distribution

A function that takes the mean anchor probabilities tensor with shape (dim,) as input and returns a target probability distribution tensor with the same shape. The returned distribution should sum up to one. The final regularization loss is calculated as KL(mean_anchor_probs, target_dist) where KL is the Kullback-Leibler divergence.

temperature

Similarities between anchors and targets are scaled by the inverse of the temperature. Must be in (0, inf).

sinkhorn_iterations

Number of sinkhorn normalization iterations on the targets.

regularization_weight

Weight factor lambda by which the regularization loss is scaled. Set to 0 to disable regularization.

gather_distributed

If True, then target probabilities are gathered from all GPUs.

Examples:

>>> # define custom target distribution
>>> def my_uniform_distribution(mean_anchor_probabilities: Tensor) -> Tensor:
>>>     dim = mean_anchor_probabilities.shape[0]
>>>     return mean_anchor_probabilities.new_ones(dim) / dim
>>>
>>> # initialize loss function
>>> loss_fn = PMSNCustomLoss(target_distribution=my_uniform_distribution)
>>>
>>> # generate anchors and targets of images
>>> anchors = transforms(images)
>>> targets = transforms(images)
>>>
>>> # feed through PMSN model
>>> anchors_out = model(anchors)
>>> targets_out = model.target(targets)
>>>
>>> # calculate loss
>>> loss = loss_fn(anchors_out, targets_out, prototypes=model.prototypes)
regularization_loss(mean_anchor_probs: torch.Tensor) → torch.Tensor

Calculates regularization loss with a custom target distribution.

class lightly.loss.regularizer.co2.CO2Regularizer(alpha: float = 1, t_consistency: float = 0.05, memory_bank_size: int = 0)

Implementation of the CO2 regularizer [0] for self-supervised learning.

[0] CO2, 2021, https://arxiv.org/abs/2010.02217

alpha

Weight of the regularization term.

t_consistency

Temperature used during softmax calculations.

memory_bank_size

Number of negative samples to store in the memory bank. Use 0 to use the second batch for negative samples.

Examples

>>> # initialize loss function for MoCo
>>> loss_fn = NTXentLoss(memory_bank_size=4096)
>>>
>>> # initialize CO2 regularizer
>>> co2 = CO2Regularizer(alpha=1.0, memory_bank_size=4096)
>>>
>>> # generate two random trasnforms of images
>>> t0 = transforms(images)
>>> t1 = transforms(images)
>>>
>>> # feed through the MoCo model
>>> out0, out1 = model(t0, t1)
>>>
>>> # calculate loss and apply regularizer
>>> loss = loss_fn(out0, out1) + co2(out0, out1)
forward(out0: torch.Tensor, out1: torch.Tensor)

Computes the CO2 regularization term for two model outputs.

Parameters

  • out0 – Output projections of the first set of transformed images.

  • out1 – Output projections of the second set of transformed images.

Returns

The regularization term multiplied by the weight factor alpha.

class lightly.loss.swav_loss.SwaVLoss(temperature: float = 0.1, sinkhorn_iterations: int = 3, sinkhorn_epsilon: float = 0.05, sinkhorn_gather_distributed: bool = False)

Implementation of the SwaV loss.

temperature

Temperature parameter used for cross entropy calculations.

sinkhorn_iterations

Number of iterations of the sinkhorn algorithm.

sinkhorn_epsilon

Temperature parameter used in the sinkhorn algorithm.

sinkhorn_gather_distributed

If True then features from all gpus are gathered to calculate the soft codes in the sinkhorn algorithm.

forward(high_resolution_outputs: List[torch.Tensor], low_resolution_outputs: List[torch.Tensor], queue_outputs: Optional[List[torch.Tensor]] = None)

Computes the SwaV loss for a set of high and low resolution outputs.

Parameters

  • high_resolution_outputs – List of similarities of features and SwaV prototypes for the high resolution crops.

  • low_resolution_outputs – List of similarities of features and SwaV prototypes for the low resolution crops.

  • queue_outputs – List of similarities of features and SwaV prototypes for the queue of high resolution crops from previous batches.

Returns

Swapping assignments between views loss (SwaV) as described in [0].

[0]: SwaV, 2020, https://arxiv.org/abs/2006.09882

subloss(z: torch.Tensor, q: torch.Tensor)

Calculates the cross entropy for the SwaV prediction problem.

Parameters

  • z – Similarity of the features and the SwaV prototypes.

  • q – Codes obtained from Sinkhorn iterations.

Returns

Cross entropy between predictions z and codes q.

class lightly.loss.sym_neg_cos_sim_loss.SymNegCosineSimilarityLoss

Implementation of the Symmetrized Loss used in the SimSiam[0] paper.

[0] SimSiam, 2020, https://arxiv.org/abs/2011.10566

Examples

>>> # initialize loss function
>>> loss_fn = SymNegCosineSimilarityLoss()
>>>
>>> # generate two random transforms of images
>>> t0 = transforms(images)
>>> t1 = transforms(images)
>>>
>>> # feed through SimSiam model
>>> out0, out1 = model(t0, t1)
>>>
>>> # calculate loss
>>> loss = loss_fn(out0, out1)
forward(out0: torch.Tensor, out1: torch.Tensor)

Forward pass through Symmetric Loss.

Parameters

  • out0 – Output projections of the first set of transformed images. Expects the tuple to be of the form (z0, p0), where z0 is the output of the backbone and projection mlp, and p0 is the output of the prediction head.

  • out1 – Output projections of the second set of transformed images. Expects the tuple to be of the form (z1, p1), where z1 is the output of the backbone and projection mlp, and p1 is the output of the prediction head.

Returns

Contrastive Cross Entropy Loss value.

Raises

ValueError if shape of output is not multiple of batch_size. –

class lightly.loss.tico_loss.TiCoLoss(beta: float = 0.9, rho: float = 20.0, gather_distributed: bool = False)

Implementation of the Tico Loss from Tico[0] paper. This implementation takes inspiration from the code published by sayannag using Lightly. [1]

[0] Jiachen Zhu et. al, 2022, Tico… https://arxiv.org/abs/2206.10698 [1] https://github.com/sayannag/TiCo-pytorch

Args
beta:

Coefficient for the EMA update of the covariance Defaults to 0.9 [0].

rho:

Weight for the covariance term of the loss Defaults to 20.0 [0].

gather_distributed:

If True then the cross-correlation matrices from all gpus are gathered and summed before the loss calculation.

Examples

>>> # initialize loss function
>>> loss_fn = TiCoLoss()
>>>
>>> # generate two random transforms of images
>>> t0 = transforms(images)
>>> t1 = transforms(images)
>>>
>>> # feed through model
>>> out0, out1 = model(t0, t1)
>>>
>>> # calculate loss
>>> loss = loss_fn(out0, out1)
forward(z_a: torch.Tensor, z_b: torch.Tensor, update_covariance_matrix: bool = True) → torch.Tensor

Tico Loss computation. It maximize the agreement among embeddings of different distorted versions of the same image while avoiding collapse using Covariance matrix.

Parameters

  • z_a – Tensor of shape [batch_size, num_features=256]. Output of the learned backbone.

  • z_b – Tensor of shape [batch_size, num_features=256]. Output of the momentum updated backbone.

  • update_covariance_matrix – Parameter to update the covariance matrix at each iteration.

Returns

The loss.

class lightly.loss.vicreg_loss.VICRegLoss(lambda_param: float = 25.0, mu_param: float = 25.0, nu_param: float = 1.0, gather_distributed: bool = False, eps=0.0001)

Implementation of the VICReg loss [0].

This implementation is based on the code published by the authors [1].

lambda_param

Scaling coefficient for the invariance term of the loss.

mu_param

Scaling coefficient for the variance term of the loss.

nu_param

Scaling coefficient for the covariance term of the loss.

gather_distributed

If True then the cross-correlation matrices from all gpus are gathered and summed before the loss calculation.

eps

Epsilon for numerical stability.

Examples

>>> # initialize loss function
>>> loss_fn = VICRegLoss()
>>>
>>> # generate two random transforms of images
>>> t0 = transforms(images)
>>> t1 = transforms(images)
>>>
>>> # feed through model
>>> out0, out1 = model(t0, t1)
>>>
>>> # calculate loss
>>> loss = loss_fn(out0, out1)
forward(z_a: torch.Tensor, z_b: torch.Tensor) → torch.Tensor

Returns VICReg loss.

Parameters

  • z_a – Tensor with shape (batch_size, …, dim).

  • z_b – Tensor with shape (batch_size, …, dim).

class lightly.loss.vicregl_loss.VICRegLLoss(lambda_param: float = 25.0, mu_param: float = 25.0, nu_param: float = 1.0, alpha: float = 0.75, gather_distributed: bool = False, eps: float = 0.0001, num_matches: Tuple[int, int] = (20, 4))

Implementation of the VICRegL loss from VICRegL paper [0].

This implementation follows the code published by the authors [1].

lambda_param

Coefficient for the invariance term of the loss.

mu_param

Coefficient for the variance term of the loss.

nu_param

Coefficient for the covariance term of the loss.

alpha

Coefficient to weight global with local loss. The final loss is computed as (self.alpha * global_loss + (1-self.alpha) * local_loss).

gather_distributed

If True then the cross-correlation matrices from all gpus are gathered and summed before the loss calculation.

eps

Epsilon for numerical stability.

num_matches

Number of local features to match using nearest neighbors.

Examples

>>> # initialize loss function
>>> criterion = VICRegLLoss()
>>> transform = VICRegLTransform(n_global_views=2, n_local_views=4)
>>>
>>> # generate two random transforms of images
>>> views_and_grids = transform(images)
>>> views = views_and_grids[:6] # 2 global views + 4 local views
>>> grids = views_and_grids[6:]
>>>
>>> # feed through model images
>>> features = [model(view) for view in views]
>>>
>>> # calculate loss
>>> loss = criterion(
...     global_view_features=features[:2],
...     global_view_grids=grids[:2],
...     local_view_features=features[2:],
...     local_view_grids=grids[2:],
... )
forward(global_view_features: Sequence[Tuple[torch.Tensor, torch.Tensor]], global_view_grids: Sequence[torch.Tensor], local_view_features: Optional[Sequence[Tuple[torch.Tensor, torch.Tensor]]] = None, local_view_grids: Optional[Sequence[torch.Tensor]] = None) → torch.Tensor

Computes the global and local VICRegL loss from the input features.

Parameters

  • global_view_features – Sequence of (global_features, local_features) tuples from the global crop views. global_features must have size (batch_size, global_feature_dim) and local_features must have size (batch_size, grid_height, grid_width, local_feature_dim).

  • global_view_grids – Sequence of grid tensors from the global crop views. Every tensor must have shape (batch_size, grid_height, grid_width, 2).

  • local_view_features – Sequence of (global_features, local_features) tuples from the local crop views. global_features must have size (batch_size, global_feature_dim) and local_features must have size (batch_size, grid_height, grid_width, local_feature_dim). Note that grid_height and grid_width can differ between global_view_features and local_view_features.

  • local_view_grids – Sequence of grid tensors from the local crop views. Every tensor must have shape (batch_size, grid_height, grid_width, 2). Note that grid_height and grid_width can differ between global_view_features and local_view_features.

Returns

Weighted sum of the global and local loss, calculated as – self.alpha * global_loss + (1-self.alpha) * local_loss.