Skip to content

tensorflow_wrapper

TensorFlow layer factory implementations.

This module provides factory methods for creating TensorFlow/Keras layers with consistent configuration options. Currently supports Dense and Conv2D layers.

Key classes
  • TensorflowDense: for Dense layers
  • TensorflowConv2D: for Conv2D layers
Example

dense_layer = TensorflowDense.create_layer(units=64, activation='relu') conv_layer = TensorflowConv2D.create_layer(filters=32, kernel_size=3)

TensorflowConv2D

Factory class for creating Keras Conv2D layers.

Provides a static method for creating and configuring Keras Conv2D layers with consistent parameters.

Example

Create a conv layer with 64 filters and 3x3 kernel

conv1 = TensorflowConv2D.create_layer( ... filters=64, ... kernel_size=3, ... activation='relu' ... )

Create a conv layer with 32 filters and 5x5 kernel

conv2 = TensorflowConv2D.create_layer( ... filters=32, ... kernel_size=(5, 5), ... activation='relu' ... )

Source code in scirex/core/dl/tensorflow_wrapper.py 
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
class TensorflowConv2D:
    """Factory class for creating Keras Conv2D layers.

    Provides a static method for creating and configuring Keras Conv2D layers
    with consistent parameters.

    Example:
        >>> # Create a conv layer with 64 filters and 3x3 kernel
        >>> conv1 = TensorflowConv2D.create_layer(
        ...     filters=64,
        ...     kernel_size=3,
        ...     activation='relu'
        ... )
        >>>
        >>> # Create a conv layer with 32 filters and 5x5 kernel
        >>> conv2 = TensorflowConv2D.create_layer(
        ...     filters=32,
        ...     kernel_size=(5, 5),
        ...     activation='relu'
        ... )
    """

    @staticmethod
    def create_layer(
        filters: int,
        kernel_size: Union[int, Tuple[int, int]],
        activation: Optional[Union[str, callable]] = None,
        kernel_initializer: str = "glorot_uniform",
        bias_initializer: str = "zeros",
        dtype: Optional[Union[str, tf.dtypes.DType]] = None,
    ) -> tf.keras.layers.Conv2D:
        """Create and return a Keras Conv2D layer.

        Args:
            filters: Number of output filters
            kernel_size: Size of convolution kernel
            activation: Activation function to use
            kernel_initializer: Initializer for kernel weights
            bias_initializer: Initializer for bias vector
            dtype: Data type for layer computations

        Returns:
            tf.keras.layers.Conv2D: The configured Conv2D layer
        """
        return layers.Conv2D(
            filters=filters,
            kernel_size=kernel_size,
            activation=activation,
            kernel_initializer=kernel_initializer,
            bias_initializer=bias_initializer,
            dtype=dtype,
        )

create_layer(filters, kernel_size, activation=None, kernel_initializer='glorot_uniform', bias_initializer='zeros', dtype=None) staticmethod

Create and return a Keras Conv2D layer.

Parameters:

Name Type Description Default
filters int

Number of output filters

 required
kernel_size Union[int, Tuple[int, int]]

Size of convolution kernel

 required
activation Optional[Union[str, callable]]

Activation function to use

 None
kernel_initializer str

Initializer for kernel weights

 'glorot_uniform'
bias_initializer str

Initializer for bias vector

 'zeros'
dtype Optional[Union[str, DType]]

Data type for layer computations

 None

Returns:

Type Description
Conv2D

tf.keras.layers.Conv2D: The configured Conv2D layer
Source code in scirex/core/dl/tensorflow_wrapper.py 
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
@staticmethod
def create_layer(
    filters: int,
    kernel_size: Union[int, Tuple[int, int]],
    activation: Optional[Union[str, callable]] = None,
    kernel_initializer: str = "glorot_uniform",
    bias_initializer: str = "zeros",
    dtype: Optional[Union[str, tf.dtypes.DType]] = None,
) -> tf.keras.layers.Conv2D:
    """Create and return a Keras Conv2D layer.

    Args:
        filters: Number of output filters
        kernel_size: Size of convolution kernel
        activation: Activation function to use
        kernel_initializer: Initializer for kernel weights
        bias_initializer: Initializer for bias vector
        dtype: Data type for layer computations

    Returns:
        tf.keras.layers.Conv2D: The configured Conv2D layer
    """
    return layers.Conv2D(
        filters=filters,
        kernel_size=kernel_size,
        activation=activation,
        kernel_initializer=kernel_initializer,
        bias_initializer=bias_initializer,
        dtype=dtype,
    )

TensorflowDense

Factory class for creating Keras Dense layers.

Provides a static method for creating and configuring Keras Dense layers with consistent parameters.

Example

Create a dense layer with 32 units and ReLU activation

dense1 = TensorflowDense.create_layer( ... units=32, ... activation='relu' ... )

Create an output layer with 1 unit and no activation

dense2 = TensorflowDense.create_layer( ... units=1, ... activation=None ... )

Source code in scirex/core/dl/tensorflow_wrapper.py 
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
class TensorflowDense:
    """Factory class for creating Keras Dense layers.

    Provides a static method for creating and configuring Keras Dense layers
    with consistent parameters.

    Example:
        >>> # Create a dense layer with 32 units and ReLU activation
        >>> dense1 = TensorflowDense.create_layer(
        ...     units=32,
        ...     activation='relu'
        ... )
        >>>
        >>> # Create an output layer with 1 unit and no activation
        >>> dense2 = TensorflowDense.create_layer(
        ...     units=1,
        ...     activation=None
        ... )
    """

    @staticmethod
    def create_layer(
        units: int,
        activation: Optional[Union[str, callable]] = None,
        kernel_initializer: str = "glorot_uniform",
        bias_initializer: str = "zeros",
        dtype: Optional[Union[str, tf.dtypes.DType]] = None,
    ) -> tf.keras.layers.Dense:
        """Create and return a Keras Dense layer.

        Args:
            units: Number of output units
            activation: Activation function to use
            kernel_initializer: Initializer for kernel weights
            bias_initializer: Initializer for bias vector
            dtype: Data type for layer computations

        Returns:
            tf.keras.layers.Dense: The configured Dense layer
        """
        return layers.Dense(
            units=units,
            activation=activation,
            kernel_initializer=kernel_initializer,
            bias_initializer=bias_initializer,
            dtype=dtype,
        )

create_layer(units, activation=None, kernel_initializer='glorot_uniform', bias_initializer='zeros', dtype=None) staticmethod

Create and return a Keras Dense layer.

Parameters:

Name Type Description Default
units int

Number of output units

 required
activation Optional[Union[str, callable]]

Activation function to use

 None
kernel_initializer str

Initializer for kernel weights

 'glorot_uniform'
bias_initializer str

Initializer for bias vector

 'zeros'
dtype Optional[Union[str, DType]]

Data type for layer computations

 None

Returns:

Type Description
Dense

tf.keras.layers.Dense: The configured Dense layer
Source code in scirex/core/dl/tensorflow_wrapper.py 
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
@staticmethod
def create_layer(
    units: int,
    activation: Optional[Union[str, callable]] = None,
    kernel_initializer: str = "glorot_uniform",
    bias_initializer: str = "zeros",
    dtype: Optional[Union[str, tf.dtypes.DType]] = None,
) -> tf.keras.layers.Dense:
    """Create and return a Keras Dense layer.

    Args:
        units: Number of output units
        activation: Activation function to use
        kernel_initializer: Initializer for kernel weights
        bias_initializer: Initializer for bias vector
        dtype: Data type for layer computations

    Returns:
        tf.keras.layers.Dense: The configured Dense layer
    """
    return layers.Dense(
        units=units,
        activation=activation,
        kernel_initializer=kernel_initializer,
        bias_initializer=bias_initializer,
        dtype=dtype,
    )