谷歌语音识别官方speech_commands(audio_recognition)的使用指南（三）

有了pb文件，那我们必须把它转化为tf文件啊

import tensorflow as tf

graph_def_file = "E:/mian/tflllll/my_frozen_graph.pb"
input_arrays = ["Reshape_1"]
output_arrays = ["add_2"]

converter = tf.lite.TFLiteConverter.from_frozen_graph(
  graph_def_file, input_arrays, output_arrays)
tflite_model = converter.convert()
open("converted_model.tflite", "wb").write(tflite_model)

input_arrays选择前几个输入节点会报错，大概是形状未知？