Half

Casts the input tensor to 16-bit floating point.

Abstract Signature:

Half(input: Tensor)

PyTorchJAX (Core)NumPyKerasTensorFlowApple MLX

PyTorch

API: torch.half

Strategy: Plugin (type_methods)

JAX (Core)

API: jax.numpy.float16

Strategy: Macro '{input}.astype(jax.numpy.float16)'

NumPy

API: numpy.astype

Strategy: Macro '{input}.astype(numpy.float16)'

Keras

API: keras.ops.cast

Strategy: Macro 'keras.ops.cast({input}, 'float16')'

TensorFlow

API: tf.cast

Strategy: Macro 'tf.cast({input}, tf.float16)'

Apple MLX

API: mlx.core.astype

Strategy: Macro '{input}.astype(mlx.core.float16)'