Question

就存储在model.h5中的模型权重和存储在model.json中的模型体系结构而言，我有Keras模型，我的目标是将构成Keras模型的这两个文件隐匿到tensorflow Lite模型中，我尝试了几种方法，但它似乎不起作用。

当我将Tensoflow 1.15.0与以下代码一起使用时，我得到“ NameError：未定义名称'lite'”，而当我降级到Tensoflow 1.15.0时，我得到了 “ AttributeError：类型对象'TFLiteConverter'没有属性'from_keras_model'”

任何人都可以帮忙谢谢！

#from tensorflow.contrib import lite 
import tensorflow as tf

from tensorflow.contrib import lite

from keras.models import model_from_json
# Model reconstruction from JSON file
with open('drive/My Drive/Colab Notebooks/model.json', 'r') as f:
    model = model_from_json(f.read())

# Load weights into the new model
model.load_weights('drive/My Drive/Colab Notebooks/model.h5')

# Converting a tf.Keras model to a TensorFlow Lite model.
converter = lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()

Answer 1

对于此问题，我有以下解决方案：

将tensorflow更新为我目前使用的是2.1.0-rc0

然后代替

model = model_from_json(f.read())

使用

model = tf.keras.models.model_from_json(f.read())

整个代码为：

import tensorflow as tf

with open('../input/melanoma-cancer-h5-model/model.json', 'r') as f:
    model = tf.keras.models.model_from_json(f.read())

# Load weights into the new model
model.load_weights('../input/melanoma-cancer-h5-model/model.h5')

# Convert the model.
converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()
open("model.tflite","wb").write(tflite_model)

将Keras模型权重和架构转换为Tensoflow Lite模型

1 个答案: