SquaredHingeLoss

Computes the squared hinge loss between y_true & y_pred.

Abstract Signature:

SquaredHingeLoss(y_true: Tensor, y_pred: Tensor)

PyTorchJAX (Core)KerasTensorFlowFlax NNXPaxML / Praxis

PyTorch

API: —

Strategy: Macro 'torch.mean(torch.clamp(1 - {y_true} * {y_pred}, min=0) ** 2, dim=-1)'

JAX (Core)

API: —

Strategy: Macro 'jax.numpy.mean(jax.numpy.square(jax.numpy.maximum(1 - {y_true} * {y_pred}, 0)), axis=-1)'

Keras

API: keras.losses.squared_hinge

Strategy: Direct Mapping

Official Docs ↗

TensorFlow

API: tf.keras.losses.squared_hinge

Strategy: Direct Mapping

Official Docs ↗

Flax NNX

API: —

Strategy: Macro 'jax.numpy.mean(jax.numpy.square(jax.numpy.maximum(1 - {y_true} * {y_pred}, 0)), axis=-1)'

PaxML / Praxis

API: —

Strategy: Macro 'jax.numpy.mean(jax.numpy.square(jax.numpy.maximum(1 - {y_true} * {y_pred}, 0)), axis=-1)'