Micro Speech:How to run audio_preprocessor.py independently?

[ctwillson]

I have read the doc,but I don't know how to run audio_preprocessor.py like "python audio_preprocessor.py"instead of using bazel.
Any ideas for this?

[ctwillson]

@ddavis-2015 I tried to generate feature by using audio_preprocessor_float32.tflite(or audio_processer_int8.tflite) like below code:

_tflm_interpreter = runtime.Interpreter.from_file('audio_preprocessor_float.tflite')
def generate_feature_using_tflm_float(audio_frame: tf.Tensor) -> tf.Tensor:
    """

    Generate a single feature for a single audio frame.  Uses TensorFlow
    graph execution and the TensorFlow model converter to generate a
    TFLM compatible model.  This model is then used by the TFLM
    MicroInterpreter to execute a single inference operation.

    Args:
      audio_frame: tf.Tensor, a single audio frame (self.params.window_size_ms)
      with shape (1, audio_samples_count)

    Returns:
      tf.Tensor, a tensor containing a single audio feature with shape
      (self.params.filter_bank_number_of_channels,)
    """

    global _tflm_interpreter
    input_scale, input_zero_point = input_details["quantization"]
    _tflm_interpreter.set_input(audio_frame, 0)
    _tflm_interpreter.invoke()
    result = _tflm_interpreter.get_output(0)
    result = result / input_scale + input_zero_point
    result = result.astype(input_details["dtype"])
    # print(result.shape)
    return tf.convert_to_tensor(result)

however the generated feature is not equal to the feature when I trained by using audio_microfrontend which provided by tensorflow:

            micro_frontend = frontend_op.audio_microfrontend(
                int16_input,
                sample_rate=16000,
                window_size=30,
                window_step=10,
                num_channels=40,
                upper_band_limit=7600,
                lower_band_limit=20,
                out_scale=1,
                out_type=tf.float32)
            # int16_input dims: [frames, num_channels]
            output_ = tf.multiply(micro_frontend, (10.0 / 256.0))
            #print(output_)
            # print(output_,output_.shape)
            data = output_

            data = np.expand_dims(data, axis=1).astype(np.float32)
            # print('data shape:{}data shape1:{}'.format(data.shape,data.shape[1]))
            input_details = interpreter.get_input_details()[0]
            output_details = interpreter.get_output_details()[0]
            # if model_type == "Quantized":
            input_scale, input_zero_point = input_details["quantization"]
            # print(input_scale,input_zero_point,input_details["dtype"])

            data = data / input_scale + input_zero_point
            data = data.astype(input_details["dtype"])

[github-actions]

"This issue is being marked as stale due to inactivity. Remove label or comment to prevent closure in 5 days."