run_npu_inference.py

#
# Sine model inference accelerated with vx_delegate
#

import numpy as np
import tflite_runtime.interpreter as tflite
import math

# Set path to the  VX_delegate Libraries
#
# Delegate path:
DELEGATE_PATH = "./libvx_delegate.so"

# Set path to the TFLite model
#
# Model path:
MODEL_PATH = "./sine_model.tflite"

vx_delegate = tflite.load_delegate(
    library = DELEGATE_PATH,
    options={"logging-severity":"debug"}
)

interpreter = tflite.Interpreter(
    model_path = MODEL_PATH,
    experimental_delegates = [vx_delegate]
)

interpreter.allocate_tensors()

input_details = interpreter.get_input_details()
output_details = interpreter.get_output_details()

# print("Input details\n", input_details)
# print("Output details\n", output_details)

input_type = input_details[0]['dtype']
output_type = output_details[0]['dtype']
input_value = math.pi/2

np_features = np.array((input_value,))
np_features = np_features.astype(input_type)
np_features = np.expand_dims(np_features, axis=0)

interpreter.set_tensor(input_details[0]['index'], np_features)
interpreter.invoke()
output = interpreter.get_tensor(output_details[0]['index'])

prediction = output.astype(output_type)[0][0]
actual = math.sin(input_value)
print('Actual {}, Predicted {}'.format(actual, prediction))