sayakpaul

converter = tf.lite.TFLiteConverter.from_keras_model(non_qat_flower_model)
converter.optimizations = [tf.lite.Optimize.DEFAULT]
quantized_tflite_model = converter.convert()

sayakpaul

f = open("normal_flower_model.tflite", "wb")
f.write(quantized_tflite_model)
f.close()

sayakpaul

import tensorflow_model_optimization as tfmot
qat_model = tfmot.quantization.keras.quantize_model(your_keras_model)

sayakpaul

# Copy the kernel weights and get ranked indices of the
# column-wise L2 Norms
kernel_weights = np.copy(k_weights)
ind = np.argsort(np.linalg.norm(kernel_weights, axis=0))
    
# Number of indices to be set to 0
sparsity_percentage = 0.7
cutoff = int(len(ind)*sparsity_percentage)

# The indices in the 2D kernel weight matrix to be set to 0

sayakpaul

pruning_schedule = tfmot.sparsity.keras.ConstantSparsity(
          target_sparsity=target_sparsity,
          begin_step=begin_step,
          end_step=end_step,
          frequency=frequency
)

pruned_model = tfmot.sparsity.keras.prune_low_magnitude(
    trained_model, pruning_schedule=pruning_schedule
)

sayakpaul

pruned_model.compile(loss='sparse_categorical_crossentropy',
    optimizer='adam',
    metrics=['accuracy'])

pruned_model.summary()

sayakpaul

Layer (type)                 Output Shape              Param #   
=================================================================
prune_low_magnitude_conv2d ( (None, 26, 26, 12)        230       
_________________________________________________________________
prune_low_magnitude_max_pool (None, 13, 13, 12)        1         
_________________________________________________________________
prune_low_magnitude_flatten  (None, 2028)              1         
_________________________________________________________________
prune_low_magnitude_dense (P (None, 10)                40572     
=================================================================

sayakpaul

for layer in model.layers:
	if isinstance(layer, pruning_wrapper.PruneLowMagnitude):
	    for weight in layer.layer.get_prunable_weights():
	        print(np.allclose(
	            target_sparsity, get_sparsity(tf.keras.backend.get_value(weight)), 
	            rtol=1e-6, atol=1e-6)
            )

def get_sparsity(weights):
    return 1.0 - np.count_nonzero(weights) / float(weights.size)

sayakpaul

def get_gzipped_model_size(file):
    _, zipped_file = tempfile.mkstemp('.zip')
    with zipfile.ZipFile(zipped_file, 'w', compression=zipfile.ZIP_DEFLATED) as f:
        f.write(file)

    return os.path.getsize(zipped_file)

sayakpaul

# Courtesy: TFOD API Authors
# https://github.com/tensorflow/models/blob/master/research/object_detection

model {
  ssd {
    inplace_batchnorm_update: true
    freeze_batchnorm: false
    num_classes: 37
    box_coder {
      faster_rcnn_box_coder {