tensorflow2 h5 转 pb --- get_concrete_function 报错:TypeError: ‘NoneType‘ object is not subscriptable

最新推荐文章于 2023-02-22 23:28:24 发布

gakki-fun

最新推荐文章于 2023-02-22 23:28:24 发布

阅读量1.9k

点赞数 2

分类专栏：技术文章标签：神经网络 tensorflow 深度学习

本文链接：https://blog.csdn.net/weixin_39323526/article/details/107946958

版权

技术专栏收录该内容

26 篇文章 0 订阅

订阅专栏

import tensorflow as tf
import tensorflow_hub as hub
from tensorflow.python.framework.convert_to_constants import convert_variables_to_constants_v2
from tensorflow.keras import layers
def h5_to_pb(h5_save_path):

    model = tf.keras.models.load_model(h5_save_path, custom_objects={'KerasLayer': hub.KerasLayer, 'Dense':tf.keras.layers.Dense}, compile=False)
    # model = tf.keras.models.load_model(h5_save_path, compile=False)
    model.summary()
    full_model = tf.function(lambda Input: model(Input))
# 下面这句话出了问题
    # full_model = full_model.get_concrete_function(tf.TensorSpec(model.inputs[0].shape, model.inputs[0].dtype)) 
# 改成了这样子的,如愿生成了pb文件
 full_model = full_model.get_concrete_function(tf.TensorSpec([None, 224, 224, 3], tf.float32))

    # Get frozen ConcreteFunction
    frozen_func = convert_variables_to_constants_v2(full_model)
    frozen_func.graph.as_graph_def()

    layers = [op.name for op in frozen_func.graph.get_operations()]
    print("-" * 50)
    print("Frozen model layers: ")
    for layer in layers:
        print(layer)

    print("-" * 50)
    print("Frozen model inputs: ")
    print(frozen_func.inputs)
    print("Frozen model outputs: ")
    print(frozen_func.outputs)

    # Save frozen graph from frozen ConcreteFunction to hard drive
    tf.io.write_graph(graph_or_graph_def=frozen_func.graph,
                      logdir="./frozen_models2",
                      name="model.pb",
                      as_text=False)

h5_to_pb('model.h5')

训练完成后保存的h5文件,在转成pb格式的时候遇到了如下的问题:

明明读到了model模型,却卡在了

full_model = full_model.get_concrete_function(tf.TensorSpec(model.inputs[0].shape, model.inputs[0].dtype))  这句话

2020-08-11 23:15:41.884905: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): GeForce GTX 1080 Ti, Compute Capability 6.1
Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
keras_layer (KerasLayer)     multiple                  42626560
_________________________________________________________________
dense (Dense)                multiple                  4098
=================================================================
Total params: 42,630,658
Trainable params: 42,532,994
Non-trainable params: 97,664
_________________________________________________________________
Traceback (most recent call last):
File "/home/zhangwei/PycharmProjects/DL/CNN_From_calmisential/h5_to_pb.py", line 35, in <module>
    h5_to_pb('model.h5')
File "/home/zhangwei/PycharmProjects/DL/CNN_From_calmisential/h5_to_pb.py", line 11, in h5_to_pb
    full_model = full_model.get_concrete_function(tf.TensorSpec(model.inputs[0].shape, model.inputs[0].dtype))
TypeError: 'NoneType' object is not subscriptable

经过一个多小时的折腾,暂时找到的解决办法如下:

# 下面这句话出了问题

# full_model = full_model.get_concrete_function(tf.TensorSpec(model.inputs[0].shape, model.inputs[0].dtype))
# 改成了这样子的,如愿生成了pb文件
full_model = full_model.get_concrete_function(tf.TensorSpec([None, 224, 224, 3], tf.float32))