TensorFlow中sess.run执行全局变量初始化报错：fetch为None的解决方法

问题场景 #

执行代码中的sess.run(tf.compat.v1.global_variables_initializer())语句时，触发错误：

TypeError: Argument fetch = None has invalid type "NoneType". Cannot be None

相关代码 #

sess = tf.compat.v1.Session(config=config)

with open(r'C:\Users\User\Downloads\New folder (8)\Docify-master\api\data\ctpn.pb', 'rb') as f:
    graph_def = tf.compat.v1.GraphDef()
    # print("sdsds",graph_def)
    # text_format.Merge(f.read(), graph_def) 
    graph_def.ParseFromString(f.read())
    # print("sss",graph_def.ParseFromString(f.read()))
    sess.graph.as_default()
    tf.import_graph_def(graph_def, name='')
    # print("hhdsd",tf.compat.v1.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
input_img = sess.graph.get_tensor_by_name('Placeholder:0')
output_cls_prob = sess.graph.get_tensor_by_name('Reshape_2:0')
output_box_pred = sess.graph.get_tensor_by_name('rpn_bbox_pred/Reshape_1:0')

textdetector = TextDetector()

修复方法 #

核心原因 #

从预训练的pb文件导入计算图时，图中的变量已经被固化为常量，此时tf.compat.v1.global_variables_initializer()会返回None，传给sess.run()就会触发类型错误。

具体解决方案 #

1. 直接删除初始化语句
   删掉sess.run(tf.compat.v1.global_variables_initializer())这一行即可，因为预训练模型不需要重新初始化变量。修改后的代码：

   sess = tf.compat.v1.Session(config=config)
   
   with open(r'C:\Users\User\Downloads\New folder (8)\Docify-master\api\data\ctpn.pb', 'rb') as f:
       graph_def = tf.compat.v1.GraphDef()
       graph_def.ParseFromString(f.read())
       sess.graph.as_default()
       tf.import_graph_def(graph_def, name='')
   # 移除初始化语句
   input_img = sess.graph.get_tensor_by_name('Placeholder:0')
   output_cls_prob = sess.graph.get_tensor_by_name('Reshape_2:0')
   output_box_pred = sess.graph.get_tensor_by_name('rpn_bbox_pred/Reshape_1:0')
   
   textdetector = TextDetector()
   

2. 安全初始化（兼容有可初始化变量的场景）
   如果不确定图中是否存在需要初始化的变量，可以先检查全局变量列表，再选择性初始化：

   sess = tf.compat.v1.Session(config=config)
   
   with open(r'C:\Users\User\Downloads\New folder (8)\Docify-master\api\data\ctpn.pb', 'rb') as f:
       graph_def = tf.compat.v1.GraphDef()
       graph_def.ParseFromString(f.read())
       sess.graph.as_default()
       tf.import_graph_def(graph_def, name='')
   
   # 检查并初始化存在的全局变量
   global_vars = tf.compat.v1.global_variables()
   if global_vars:
       sess.run(tf.compat.v1.variables_initializer(global_vars))
   
   input_img = sess.graph.get_tensor_by_name('Placeholder:0')
   output_cls_prob = sess.graph.get_tensor_by_name('Reshape_2:0')
   output_box_pred = sess.graph.get_tensor_by_name('rpn_bbox_pred/Reshape_1:0')
   
   textdetector = TextDetector()
   

内容的提问来源于stack exchange，提问作者vishnu Vs