LogCumSumExp

Logarithm of the cumulative summation of exponentiation.

Abstract Signature:

LogCumSumExp(input: Tensor, dim: int)

PyTorch

API: torch.logcumsumexp
Strategy: Direct Mapping
Official Docs ↗

JAX (Core)

API: jax.numpy.logaddexp.accumulate
Strategy: Macro 'jax.numpy.log(jax.numpy.cumsum(jax.numpy.exp({input}), axis={dim}))'

NumPy

API: numpy.logaddexp.accumulate
Strategy: Macro 'numpy.log(numpy.cumsum(numpy.exp({input}), axis={dim}))'

Keras

API: keras.ops.logcumsumexp
Strategy: Direct Mapping
Official Docs ↗

TensorFlow

API: tf.math.cumulative_logsumexp
Strategy: Direct Mapping
Official Docs ↗

Apple MLX

API: mlx.core.logcumsumexp
Strategy: Macro 'mlx.core.log(mlx.core.cumsum(mlx.core.exp({input}), axis={dim}))'

Flax NNX

API: jax.numpy.logaddexp.accumulate
Strategy: Macro 'jax.numpy.log(jax.numpy.cumsum(jax.numpy.exp({input}), axis={dim}))'

PaxML / Praxis

API: jax.numpy.logaddexp.accumulate
Strategy: Direct Mapping
Official Docs ↗