espnet2.asr.transducer.rnnt_multi_blank.utils.rnnt_helper.identity

source

espnet2.asr.transducer.rnnt_multi_blank.utils.rnnt_helper.identity(x)

Identity function for use in CUDA kernels.

This function returns the input value as-is. It is typically used in scenarios where an operation requires a function but no transformation of the input is needed. This is especially useful in neural network architectures or during computation graphs where identity mappings are required.

• Parameters:x (float) – The input value to be returned unchanged.
• Returns: The input value x.
• Return type: float

Examples

>>> result = identity(5.0)
>>> print(result)
5.0

>>> result = identity(-3.2)
>>> print(result)
-3.2

NOTE

This function is compiled to run on CUDA devices using Numba, making it suitable for GPU-based computations.