NanToNum

Replaces NaN, positive infinity, and negative infinity values in input with valid numbers.

Abstract Signature:

NanToNum(input: Tensor, nan: float = 0.0, posinf: float, neginf: float)

PyTorchJAX (Core)NumPyKerasTensorFlowApple MLXFlax NNXPaxML / Praxis

PyTorch

API: torch.nan_to_num

Strategy: Direct Mapping

Official Docs ↗

JAX (Core)

API: jax.numpy.nan_to_num

Strategy: Direct Mapping

Official Docs ↗

NumPy

API: numpy.nan_to_num

Strategy: Direct Mapping

Official Docs ↗

Keras

API: keras.ops.nan_to_num

Strategy: Direct Mapping

Official Docs ↗

TensorFlow

API: numpy.nan_to_num

Strategy: Plugin (numpy_fallback)

Apple MLX

API: —

Strategy: Plugin (numpy_fallback)

Flax NNX

API: jax.numpy.nan_to_num

Strategy: Direct Mapping

Official Docs ↗

PaxML / Praxis

API: jax.numpy.nan_to_num

Strategy: Direct Mapping

Official Docs ↗