DiceLoss

Computes the Dice loss value between y_true and y_pred.

Abstract Signature:

DiceLoss(y_true: Tensor, y_pred: Tensor, axis: int)

PyTorchJAX (Core)KerasTensorFlowFlax NNXPaxML / Praxis

PyTorch

API: —

Strategy: Macro '1 - (2 * ({y_true} * {y_pred}).sum(dim={axis})) / ({y_true}.sum(dim={axis}) + {y_pred}.sum(dim={axis}))'

JAX (Core)

API: —

Strategy: Macro '1 - (2 * ({y_true} * {y_pred}).sum(axis={axis})) / ({y_true}.sum(axis={axis}) + {y_pred}.sum(axis={axis}))'

Keras

API: keras.losses.dice

Strategy: Direct Mapping

Official Docs ↗

TensorFlow

API: tf.keras.losses.dice

Strategy: Direct Mapping

Official Docs ↗

Flax NNX

API: —

Strategy: Macro '1 - (2 * ({y_true} * {y_pred}).sum(axis={axis})) / ({y_true}.sum(axis={axis}) + {y_pred}.sum(axis={axis}))'

PaxML / Praxis

API: —

Strategy: Macro '1 - (2 * ({y_true} * {y_pred}).sum(axis={axis})) / ({y_true}.sum(axis={axis}) + {y_pred}.sum(axis={axis}))'