convert_tflite.py

!mkdir "tflite_models"
TFLITE_MODEL = "tflite_models/flowers.tflite"
TFLITE_QUANT_MODEL = "tflite_models/flowers_quant.tflite"

# Get the concrete function from the Keras model.
run_model = tf.function(lambda x : flowers_model(x))

# Save the concrete function.
concrete_func = run_model.get_concrete_function(
    tf.TensorSpec(model.inputs[0].shape, model.inputs[0].dtype)
)

# Convert the model to standard TensorFlow Lite model
converter = tf.lite.TFLiteConverter.from_concrete_functions([concrete_func])
converted_tflite_model = converter.convert()
open(TFLITE_MODEL, "wb").write(converted_tflite_model)

# Convert the model to quantized version with post-training quantization
converter = tf.lite.TFLiteConverter.from_concrete_functions([concrete_func])
converter.optimizations = [tf.lite.Optimize.OPTIMIZE_FOR_SIZE]
tflite_quant_model = converter.convert()
open(TFLITE_QUANT_MODEL, "wb").write(tflite_quant_model)

print("TFLite models and their sizes:")
!ls "tflite_models" -lh

# >> TFLite models and their sizes:
# >> total 11M
# >> -rw-r--r-- 1 root root 2.3M May 27 19:10 flowers_quant.tflite
# >> -rw-r--r-- 1 root root 8.5M May 27 19:10 flowers.tflite