构建神经网络及训练时遇TypeError问题求助

训练模型时遇到TypeError停滞，相关代码如下：

link = "https://tfhub.dev/google/imagenet/efficientnet_v2_imagenet1k_b0/feature_vector/2"
input_layer = tf.keras.layers.Input(shape=(224, 224, 3), dtype=tf.float32, name="input", trainable=False)
feature_extractor = hub.KerasLayer(link, trainable=False)

model = tf.keras.Sequential([
    input_layer,
    feature_extractor,
    tf.keras.layers.Dense(data_info.features['label'].num_classes, activation="softmax")
])

运行时触发如下报错：

TypeError                                 Traceback (most recent call last)
Cell In[64], line 7
      4 input_layer = tf.keras.layers.Input(shape=(224, 224, 3), dtype=tf.float32, name="input")
      5 feature_extractor = hub.KerasLayer(link, trainable=False)
----> 7 model = tf.keras.Sequential([
      8     input_layer,
      9     feature_extractor,
     10     tf.keras.layers.Dense(data_info.features['label'].num_classes, activation="softmax")
     11 ])
     13 model.compile(optimizer="adam", loss="sparse_categorical_crossentropy", metrics=["accuracy"])
     14 model.summary()

File c:\Users\steel\AppData\Local\Programs\Python\Python310\lib\site-packages\tensorflow\python\trackable\base.py:204, in no_automatic_dependency_tracking.<locals>._method_wrapper(self, *args, **kwargs)
    202 self._self_setattr_tracking = False  # pylint: disable=protected-access
    203 try:
---> 204   result = method(self, *args, **kwargs)
    205 finally:
    206   self._self_setattr_tracking = previous_value  # pylint: disable=protected-access

File c:\Users\steel\AppData\Local\Programs\Python\Python310\lib\site-packages\keras\src\utils\traceback_utils.py:70, in filter_traceback.<locals>.error_handler(*args, **kwargs)
     67     filtered_tb = _process_traceback_frames(e.__traceback__)
     68     # To get the full stack trace, call:
     69     # `tf.debugging.disable_traceback_filtering()`
---&gt; 70     raise e.with_traceback(filtered_tb) from None
...


Call arguments received by layer "keras_layer_28" (type KerasLayer):
  • inputs=tf.Tensor(shape=(None, 224, 224, 3), dtype=float32)
  • training=None

已尝试添加/移除输入层、修改层参数，均未解决问题，怀疑是TF Hub提取的网络存在问题，求验证方法及解决建议。

解决建议及验证方法 #

• 调整模型构建方式，移除独立Input层
  Sequential模型中直接添加带input_shape参数的KerasLayer，避免Input层与Hub层的兼容性问题：

  link = "https://tfhub.dev/google/imagenet/efficientnet_v2_imagenet1k_b0/feature_vector/2"
  # 直接在KerasLayer中指定输入形状
  feature_extractor = hub.KerasLayer(link, trainable=False, input_shape=(224,224,3))
  
  model = tf.keras.Sequential([
      feature_extractor,
      tf.keras.layers.Dense(data_info.features['label'].num_classes, activation="softmax")
  ])
  

• 单独验证TF Hub模块可用性
  编写测试代码，直接调用Hub层，确认模块本身是否能正常工作：

  import tensorflow as tf
  import tensorflow_hub as hub
  
  link = "https://tfhub.dev/google/imagenet/efficientnet_v2_imagenet1k_b0/feature_vector/2"
  feature_extractor = hub.KerasLayer(link, trainable=False)
  
  # 生成测试输入张量
  test_input = tf.random.normal((1, 224, 224, 3))
  # 显式指定training参数为False
  output = feature_extractor(test_input, training=False)
  print("模块输出形状:", output.shape)  # 正常输出应为(1, 1280)
  

  如果此测试报错，说明模块加载或版本兼容性存在问题。

• 检查TensorFlow与TF Hub版本兼容性
  该EfficientNet V2特征向量模块要求TensorFlow 2.6及以上版本，建议使用匹配的稳定版本组合（如TensorFlow 2.15 + TF Hub 0.15.0），避免版本不兼容导致的底层错误。

• 显式指定training参数
  在构建模型时，确保Hub层的training参数明确设置为False，避免默认的None值引发的内部逻辑冲突：

  # 在模型中调用Hub层时显式指定training
  model = tf.keras.Sequential([
      input_layer,
      tf.keras.layers.Lambda(lambda x: feature_extractor(x, training=False)),
      tf.keras.layers.Dense(data_info.features['label'].num_classes, activation="softmax")
  ])
  

内容的提问来源于stack exchange，提问作者Jaheen Ahsan