neural_compressor.tensorflow.keras.layers.dense

Initialize custom dense layers for Keras quantization.

Classes

QDense

The custom quantized Dense layer.

Functions

initialize_int8_dense(fp32_layer, q_config)

Initialize int8 dense.

Module Contents

class neural_compressor.tensorflow.keras.layers.dense.QDense(name, units, activation=None, use_bias=True, kernel_initializer='glorot_uniform', bias_initializer='zeros', kernel_regularizer=None, bias_regularizer=None, activity_regularizer=None, kernel_constraint=None, bias_constraint=None, act_min_value=None, act_max_value=None, weight_min_value=None, weight_max_value=None, granularity='per_tensor', quant_status='calib', quant_mode='SCALED', quant_T='s8', quant_round_mode='HALF_AWAY_FROM_ZERO', quant_narrow_range=False, quant_axis=None, **kwargs)

The custom quantized Dense layer.

neural_compressor.tensorflow.keras.layers.dense.initialize_int8_dense(fp32_layer, q_config)

Initialize int8 dense.