RegressionMSELoss

class lightning_pose.losses.losses.RegressionMSELoss[source]

Bases: Loss

MSE loss between ground truth and predicted coordinates.

Attributes Summary

loss_name

Methods Summary

__call__(keypoints_targ, keypoints_pred[, stage])

Call self as a function.

compute_loss(targets, predictions)

remove_nans(targets, predictions)

Attributes Documentation

loss_name = 'regression'

Methods Documentation

__call__(keypoints_targ: Tensor, {'__torchtyping__': True, 'details': ('batch', 'two_x_num_keypoints'), 'cls_name': 'TensorType'}], keypoints_pred: Tensor, {'__torchtyping__': True, 'details': ('batch', 'two_x_num_keypoints'), 'cls_name': 'TensorType'}], stage: Literal['train', 'val', 'test'] | None = None, **kwargs) → tuple[~torch.Annotated[~torch.Tensor, {'__torchtyping__': True, 'details': ((),), 'cls_name': 'TensorType'}], list[dict]][source]

Call self as a function.

compute_loss(targets: Tensor, {'__torchtyping__': True, 'details': ('batch_x_two_x_num_keypoints',), 'cls_name': 'TensorType'}], predictions: Tensor, {'__torchtyping__': True, 'details': ('batch_x_two_x_num_keypoints',), 'cls_name': 'TensorType'}]) → Tensor, {'__torchtyping__': True, 'details': ('batch_x_two_x_num_keypoints',), 'cls_name': 'TensorType'}][source]
remove_nans(targets: Tensor, {'__torchtyping__': True, 'details': ('batch', 'two_x_num_keypoints'), 'cls_name': 'TensorType'}], predictions: Tensor, {'__torchtyping__': True, 'details': ('batch', 'two_x_num_keypoints'), 'cls_name': 'TensorType'}]) → tuple[~torch.Annotated[~torch.Tensor, {'__torchtyping__': True, 'details': ('num_valid_keypoints',), 'cls_name': 'TensorType'}], ~torch.Annotated[~torch.Tensor, {'__torchtyping__': True, 'details': ('num_valid_keypoints',), 'cls_name': 'TensorType'}]][source]
__init__(data_module: BaseDataModule | UnlabeledDataModule | None = None, epsilon: float = 0.0, log_weight: float = 0.0, **kwargs) → None[source]

Initialize RegressionMSELoss.

Parameters:

  • data_module – data module providing access to datasets; passed to the parent class.

  • epsilon – loss values below this threshold are zeroed out.

  • log_weight – final weight in front of the loss term in the objective function is computed as 1.0 / (2.0 * exp(log_weight)).

__new__(**kwargs)