mltk.core.TfliteUnidirectionalLstmLayer¶

class TfliteUnidirectionalLstmLayer[source]¶

UNIDIRECTIONAL_SEQUENCE_LSTM operation TfliteLayer

Properties

`activation`	Fused activation
`cell_clip`	Input tensor data
`cell_gate_bias_tensor`	Cell gate bias tensor of size {n_cell}
`cell_layer_norm_coeff_tensor`	(Optional) Layer norm coefficient tensor of size {n_cell}, representing a diagonal matrix
`cell_state_tensor`	The cell state tensor is defined as variable tensor, and will be modified at runtime
`cell_to_forget_weights_tensor`	(Optional) Peephole weights tensor of size {n_cell}, representing a diagonal matrix
`cell_to_input_weights_tensor`	(Optional) Peephole weights tensor of size {n_cell}, representing a diagonal matrix
`cell_to_output_weights_tensor`	(Optional) Peephole weights tensor of size {n_cell}, representing a diagonal matrix
`forget_gate_bias_tensor`	Forget gate bias tensor of size {n_cell}
`forget_layer_norm_coeff_tensor`	(Optional) Layer norm coefficient tensor of size {n_cell}, representing a diagonal matrix
`index`	Index of this layer in the model
`input_data`	Input tensor data
`input_gate_bias_tensor`	(Optional) Input gate bias tensor of size {n_cell}
`input_layer_norm_coeff_tensor`	(Optional) Layer norm coefficient tensor of size {n_cell}, representing a diagonal matrix
`input_tensor`	Input tensor of size {n_batch, n_input}
`input_to_cell_weights_tensor`	{n_cell, n_input}
`input_to_forget_weights_tensor`	{n_cell, n_input}
`input_to_input_weights_tensor`	{n_cell, n_input}
`input_to_output_weights_tensor`	{n_cell, n_input}
`inputs`	List of layer input tensor(s)
`is_time_major`	Return if this kernel uses time major or batch major
`metadata`	Additional key/value data to associated with layer
`model`	Reference to associated TfliteModel
`n_cells`	Number of LSTM cells
`n_inputs`	Return the number of inputs
`n_outputs`	Return the number of outputs
`name`	op<index>-<OpCodeStr>
`opcode`	OpCode numeric value
`opcode_str`	OpCode as a string
`options`	Layer-specific options/config
`output_data`	Output tensor data
`output_gate_bias_tensor`	Output gate bias tensor of size {n_cell}
`output_layer_norm_coeff_tensor`	(Optional) Layer norm coefficient tensor of size {n_cell}, representing a diagonal matrix
`output_state_tensor`	The output state tensor is defined as variable tensor, and will be modified at runtime
`output_tensor`	Output tensor data
`outputs`	List of layer output tensor(s)
`proj_clip`	Input tensor data
`projection_bias_tensor`	(Optional) Projection bias tensor of size {n_output}
`projection_weights_tensor`	(Optional) Projection weight tensor of size {n_output, n_cell}
`recurrent_to_cell_weights_tensor`	Recurrent weight tensor of size {n_cell, n_output}
`recurrent_to_forget_weights_tensor`	Recurrent weight tensor of size {n_cell, n_output}
`recurrent_to_input_weights_tensor`	(Optional) Recurrent weight tensor of size {n_cell, n_output}
`recurrent_to_output_weights_tensor`	Recurrent weight tensor of size {n_cell, n_output}

Methods

`__init__`
`from_flatbuffer`	Instantiate a TfliteLayer from then given TfliteModel flatbuffer operation
`get_input_data`	Get layer input tensor as np.ndarray
`get_input_tensor`	Get layer input tensor as TfliteTensor
`get_output_data`	Layer output tensor as np.ndarray
`get_output_tensor`	Layer output tensor as TfliteTensor
`get_tensor_name_value_tuples`

__init__(fb_operation, **kwargs)[source]¶

Parameters:: fb_operation (OperatorT) –

property options: TfliteUnidirectionalLstmLayerOptions¶

Layer-specific options/config

Return type:: TfliteUnidirectionalLstmLayerOptions

property activation: str¶

Fused activation

Return type:: str

property is_time_major: bool¶

Return if this kernel uses time major or batch major

Return type:: bool

property cell_clip: float¶

Input tensor data

Return type:: float

property proj_clip: float¶

Input tensor data

Return type:: float

property n_cells: int¶

Number of LSTM cells

Return type:: int

property input_tensor: TfliteTensor¶

Input tensor of size {n_batch, n_input}

Return type:: TfliteTensor

property input_data: ndarray¶

Input tensor data

Return type:: ndarray

property output_tensor: TfliteTensor¶

Output tensor data

Return type:: TfliteTensor

property output_data: ndarray¶

Output tensor data

Return type:: ndarray

property input_to_input_weights_tensor: TfliteTensor¶

{n_cell, n_input}

Type:: (Optional) Input weight tensor of size
Return type:: TfliteTensor

static from_flatbuffer(index, model, fb_operation)¶

Instantiate a TfliteLayer from then given TfliteModel flatbuffer operation

Return type:

TfliteLayer

Parameters:

index (int) –
model (TfliteModel) –
fb_operation (OperatorT) –

get_input_data(index=0)¶

Get layer input tensor as np.ndarray

Return type:: ndarray

get_input_tensor(index=0)¶

Get layer input tensor as TfliteTensor

Return type:: TfliteTensor

get_output_data(index=0)¶

Layer output tensor as np.ndarray

Return type:: ndarray

get_output_tensor(index=0)¶

Layer output tensor as TfliteTensor

Return type:: TfliteTensor

property index: int¶

Index of this layer in the model

Return type:: int

property input_to_forget_weights_tensor: TfliteTensor¶

{n_cell, n_input}

Type:: Input weight tensor of size
Return type:: TfliteTensor

property inputs: List[TfliteTensor]¶

List of layer input tensor(s)

Return type:: List[TfliteTensor]

property metadata: Dict[str, object]¶

Additional key/value data to associated with layer

NOTE: This information is generated by the framework/Python scripts: (i.e. The information does NOT come from the .tflite model)

Return type:: Dict[str, object]

property model: TfliteModel¶

Reference to associated TfliteModel

Return type:: TypeVar(TfliteModel)

property n_inputs: int¶

Return the number of inputs

Return type:: int

property n_outputs: int¶

Return the number of outputs

Return type:: int

property name: str¶

op<index>-<OpCodeStr>

Type:: Name of current layer as
Return type:: str

property opcode: BuiltinOperator¶

OpCode numeric value

Return type:: BuiltinOperator

property opcode_str: str¶

OpCode as a string

Return type:: str

property outputs: List[TfliteTensor]¶

List of layer output tensor(s)

Return type:: List[TfliteTensor]

property input_to_cell_weights_tensor: TfliteTensor¶

{n_cell, n_input}

Type:: Input weight tensor of size
Return type:: TfliteTensor

property input_to_output_weights_tensor: TfliteTensor¶

{n_cell, n_input}

Type:: Input weight tensor of size
Return type:: TfliteTensor

property recurrent_to_input_weights_tensor: TfliteTensor¶

(Optional) Recurrent weight tensor of size {n_cell, n_output}

Return type:: TfliteTensor

property recurrent_to_forget_weights_tensor: TfliteTensor¶

Recurrent weight tensor of size {n_cell, n_output}

Return type:: TfliteTensor

property recurrent_to_cell_weights_tensor: TfliteTensor¶

Recurrent weight tensor of size {n_cell, n_output}

Return type:: TfliteTensor

property recurrent_to_output_weights_tensor: TfliteTensor¶

Recurrent weight tensor of size {n_cell, n_output}

Return type:: TfliteTensor

property cell_to_input_weights_tensor: TfliteTensor¶

(Optional) Peephole weights tensor of size {n_cell}, representing a diagonal matrix

Return type:: TfliteTensor

property cell_to_forget_weights_tensor: TfliteTensor¶

(Optional) Peephole weights tensor of size {n_cell}, representing a diagonal matrix

Return type:: TfliteTensor

property cell_to_output_weights_tensor: TfliteTensor¶

(Optional) Peephole weights tensor of size {n_cell}, representing a diagonal matrix

Return type:: TfliteTensor

property input_gate_bias_tensor: TfliteTensor¶

(Optional) Input gate bias tensor of size {n_cell}

Return type:: TfliteTensor

property forget_gate_bias_tensor: TfliteTensor¶

Forget gate bias tensor of size {n_cell}

Return type:: TfliteTensor

property cell_gate_bias_tensor: TfliteTensor¶

Cell gate bias tensor of size {n_cell}

Return type:: TfliteTensor

property output_gate_bias_tensor: TfliteTensor¶

Output gate bias tensor of size {n_cell}

Return type:: TfliteTensor

property projection_weights_tensor: TfliteTensor¶

(Optional) Projection weight tensor of size {n_output, n_cell}

Return type:: TfliteTensor

property projection_bias_tensor: TfliteTensor¶

(Optional) Projection bias tensor of size {n_output}

Return type:: TfliteTensor

property output_state_tensor: TfliteTensor¶

The output state tensor is defined as variable tensor, and will be modified at runtime

Return type:: TfliteTensor

property cell_state_tensor: TfliteTensor¶

The cell state tensor is defined as variable tensor, and will be modified at runtime

Return type:: TfliteTensor

property input_layer_norm_coeff_tensor: TfliteTensor¶

(Optional) Layer norm coefficient tensor of size {n_cell}, representing a diagonal matrix

Return type:: TfliteTensor

property forget_layer_norm_coeff_tensor: TfliteTensor¶

(Optional) Layer norm coefficient tensor of size {n_cell}, representing a diagonal matrix

Return type:: TfliteTensor

property cell_layer_norm_coeff_tensor: TfliteTensor¶

(Optional) Layer norm coefficient tensor of size {n_cell}, representing a diagonal matrix

Return type:: TfliteTensor

property output_layer_norm_coeff_tensor: TfliteTensor¶

(Optional) Layer norm coefficient tensor of size {n_cell}, representing a diagonal matrix

Return type:: TfliteTensor