quaterion.loss.cos_face_loss module

class CosFaceLoss(embedding_size: int, num_groups: int, margin: float | None = 0.35, scale: float | None = 64.0)

Bases: GroupLoss

Large Margin Cosine Loss as defined in https://arxiv.org/pdf/1801.09414.pdf

Parameters:

• embedding_size – Output dimension of the encoder.

• num_groups – Number of groups in the dataset.

• scale – Scaling value to make cross entropy work.

• margin – Margin value to push groups apart.

forward(embeddings: Tensor, groups: LongTensor) → Tensor

Compute loss value :param embeddings: shape: (batch_size, vector_length) - Output embeddings from the

encoder.

Parameters:

groups – shape: (batch_size,) - Group ids associated with embeddings.

Returns:

Tensor – loss value.

training: bool