npu.utils.dl_utils

Classes

DLUtils()

class npu.utils.dl_utils.DLUtils
Author:

Alberto M. Esmoris Pena

Class with utils static methods to work with deep learning.

static is_using_binary_crossentropy(comp_args, default=False)

Check whether the compilation arguments dictionary uses a binary cross-entropy loss function (True) or not (False).

Parameters:

  • comp_args (dict) – The compilation arguments.

  • default (bool) – Default value to be assummed when the loss function cannot be explicitly checked.

Returns:

True if a binary cross-entropy is used, False otherwise.

Return type:

bool

static activation_layer_from_name(act_name, **kwargs)

Return the corresponding keras activation layer from the given name and arguments.

Parameters:

  • act_name (str) – The name of the activation function to be instantiated as a layer.

  • kwargs (dict) – The key-word arguments for the activation layer.

Returns:

The activation layer.