W5500 evb pico 中的 TinyML 开始

使用 TensorFlow Lite 和 W5500-EVB-Pico 进行正弦波预测和 LED 控制

转发: TinyML in W5500 evb pico Beginning

项目介绍

生成数据

该项目的重点是预测正弦波并实时控制 LED。 主要目标是开发一个使用 TensorFlow 预测正弦波的模型，并将该模型量化为 TensorFlow Lite 格式，以减少内存使用和计算成本。 此过程最大限度地提高了模型的效率，并使其能够在真实硬件上实时运行。 该项目的关键部分是使用 W5500-EVB-Pico 板根据预测的正弦波数据控制 LED。 该板运行 TensorFlow Lite 模型，并提供实时处理数据和以正弦波点亮 LED 所需的计算能力。 这为需要复杂数据处理和实时响应的应用提供了适合的解决方案。

数据生成

# We'll generate this many sample datapoints
SAMPLES = 1000

# Set a "seed" value, so we get the same random numbers each time we run this
# notebook
np.random.seed(1337)

# Generate a uniformly distributed set of random numbers in the range from
# 0 to 2π, which covers a complete sine wave oscillation
x_values = np.random.uniform(low=0, high=2*math.pi, size=SAMPLES)

# Shuffle the values to guarantee they're not in order
np.random.shuffle(x_values)

# Calculate the corresponding sine values
y_values = np.sin(x_values)
# Add a small random number to each y value
y_values += 0.1 * np.random.randn(*y_values.shape)

# Plot our data
plt.plot(x_values, y_values, 'b.')
plt.show()

生成正弦波数据并通过给出 +-0.1 的误差值来配置用于机器学习的数据集。

创建神经网络模型

model_2 = tf.keras.Sequential()

# First layer takes a scalar input and feeds it through 16 "neurons". The
# neurons decide whether to activate based on the 'relu' activation function.
model_2.add(layers.Dense(16, activation='relu', input_shape=(1,)))

# The new second layer may help the network learn more complex representations
model_2.add(layers.Dense(16, activation='relu'))

# Final layer is a single neuron, since we want to output a single value
model_2.add(layers.Dense(1))

# Compile the model using a standard optimizer and loss function for regression
model_2.compile(optimizer='rmsprop', loss='mse', metrics=['mae'])

history_2 = model_2.fit(x_train, y_train, epochs=600, batch_size=16,
                    validation_data=(x_validate, y_validate))

预测结果

# Calculate and print the loss on our test dataset
loss = model_2.evaluate(x_test, y_test)

# Make predictions based on our test dataset
predictions = model_2.predict(x_test)

# Graph the predictions against the actual values
plt.clf()
plt.title('Comparison of predictions and actual values')
plt.plot(x_test, y_test, 'b.', label='Actual')
plt.plot(x_test, predictions, 'r.', label='Predicted')
plt.legend()
plt.show()

这是描绘评估数据和预测数据的图表。 如果您查看预测图，它似乎具有相当的正弦形状。

转换为量化模型

# Convert the model to the TensorFlow Lite format with quantization
converter = tf.lite.TFLiteConverter.from_keras_model(model_2)
converter.optimizations = [tf.lite.Optimize.OPTIMIZE_FOR_SIZE]
tflite_model = converter.convert()

# Save the model to disk
open("sine_model_quantized.tflite", "wb").write(tflite_model)

# Linux
# Install xxd if it is not available
!apt-get -qq install xxd
# Save the file as a C source file
!xxd -i sine_model_quantized.tflite > sine_model_quantized.cc
# Print the source file
!cat sine_model_quantized.cc

现有的tensorflow模型将归一化数据表示为32位浮点数，但tensorflow lite模型将归一化数据表示为8位整数。 以降低数据归一化精度为代价，理论性能提升高达4倍。

量化结果

unsigned char sine_model_quantized_tflite[] = {
  0x18, 0x00, 0x00, 0x00, 0x54, 0x46, 0x4c, 0x33, 0x00, 0x00, 0x0e, 0x00,
  0x18, 0x00, 0x04, 0x00, 0x08, 0x00, 0x0c, 0x00, 0x10, 0x00, 0x14, 0x00,
  0x0e, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x10, 0x0a, 0x00, 0x00,
  0xb8, 0x05, 0x00, 0x00, 0xa0, 0x05, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
  0x0b, 0x00, 0x00, 0x00, 0x90, 0x05, 0x00, 0x00, 0x7c, 0x05, 0x00, 0x00,
  0x24, 0x05, 0x00, 0x00, 0xd4, 0x04, 0x00, 0x00, 0xc4, 0x00, 0x00, 0x00,
  0x74, 0x00, 0x00, 0x00, 0x24, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00,
  0x14, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
  0x54, 0xf6, 0xff, 0xff, 0x58, 0xf6, 0xff, 0xff, 0x5c, 0xf6, 0xff, 0xff,
  0x60, 0xf6, 0xff, 0xff, 0xc2, 0xfa, 0xff, 0xff, 0x04, 0x00, 0x00, 0x00,

...
...
  0x00, 0x00, 0x0a, 0x00, 0x0c, 0x00, 0x07, 0x00, 0x00, 0x00, 0x08, 0x00,
  0x0a, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x09, 0x03, 0x00, 0x00, 0x00
};
unsigned int sine_model_quantized_tflite_len = 2640;

它是量化的数据。 数据作为 8 位整数进入 sine_model_quantized_tflite 数组。

主要C函数总结

void setup() {
  // Set up logging. Google style is to avoid globals or statics because of
  // lifetime uncertainty, but since this has a trivial destructor it's okay.
  // NOLINTNEXTLINE(runtime-global-variables)
  static tflite::MicroErrorReporter micro_error_reporter;
  error_reporter = &micro_error_reporter;

  // Map the model into a usable data structure. This doesn't involve any
  // copying or parsing, it's a very lightweight operation.
  model = tflite::GetModel(g_sine_model_data);
  if (model->version() != TFLITE_SCHEMA_VERSION) {
    TF_LITE_REPORT_ERROR(error_reporter,
                         "Model provided is schema version %d not equal "
                         "to supported version %d.",
                         model->version(), TFLITE_SCHEMA_VERSION);
    return;
  }

  // This pulls in all the operation implementations we need.
  // NOLINTNEXTLINE(runtime-global-variables)
  static tflite::ops::micro::AllOpsResolver resolver;

  // Build an interpreter to run the model with.
  static tflite::MicroInterpreter static_interpreter(
      model, resolver, tensor_arena, kTensorArenaSize, error_reporter);
  interpreter = &static_interpreter;

  // Allocate memory from the tensor_arena for the model's tensors.
  TfLiteStatus allocate_status = interpreter->AllocateTensors();
  if (allocate_status != kTfLiteOk) {
    TF_LITE_REPORT_ERROR(error_reporter, "AllocateTensors() failed");
    return;
  }

void loop() {
  // Calculate an x value to feed into the model. We compare the current
  // inference_count to the number of inferences per cycle to determine
  // our position within the range of possible x values the model was
  // trained on, and use this to calculate a value.
  float position = static_cast<float>(inference_count) /
                   static_cast<float>(kInferencesPerCycle);
  float x_val = position * kXrange;

  // Place our calculated x value in the model's input tensor
  input->data.f[0] = x_val;

  // Run inference, and report any error
  TfLiteStatus invoke_status = interpreter->Invoke();
  if (invoke_status != kTfLiteOk) {
    TF_LITE_REPORT_ERROR(error_reporter, "Invoke failed on x_val: %f\n",
                         static_cast<double>(x_val));
    return;
  }

在设置和循环功能中设置 W5500-EVB-PICO。 设置日志记录、加载模型、设置解释器并分配内存。 之后，代码会在循环函数中重复执行。

创建生成文件

make -f {tensorflow git directory} Makefile sinWaves

二进制输出结果

当你通过make构建时，你可以看到x和y值正在输出。

在 Arduino 上构建

这是在Arduino IDE上运行的结果。

结果

结果很好，类似于正弦波。