Python OperatorNotAllowedInGraphError仍然存在，启用了急切执行并添加了decorator

我正在用tensorflow复制一个简单的代码

我有自己定义的函数作为模型的度量

这是一个简单的三重态损耗函数（稍加修改），但如果我使用一个普通函数，问题仍然是一样的

将熊猫作为pd导入
将numpy作为np导入
导入tensorflow作为tf
从tensorflow.keras.activations导入sigmoid
从tensorflow.keras导入后端
从tensorflow.keras.models导入模型，顺序
从tensorflow.keras.layers导入输入、嵌入、展平、密集、丢失、Lambda、点、连接
@功能
def bpr_三重态_损耗（输入）：
锚定潜在、正项目潜在、负项目潜在=输入
#BPR损失
损失=1.0-backend.sigmoid(
sum（anchor\u潜伏期*positive\u item\u潜伏期，axis=-1，keepdims=True）-
sum（anchor\u潜伏期*negative\u item\u潜伏期，axis=-1，keepdims=True））
回波损耗
def getModel（n个用户，n个项目，emb\u dim=20）：
#输入层
用户输入=输入（形状=[1]，名称='user\u input'）
位置项输入=输入（形状=[1]，名称='位置项输入'）
负项输入=输入（形状=[1]，名称='负项输入'）
#嵌入层
#正负项的共享嵌入层
用户\u嵌入=嵌入（输出\u dim=emb\u dim，输入\u dim=n\u users+1，输入\u length=1，name='user\u emb'）（用户\u输入）
项目嵌入=嵌入（输出尺寸=emb尺寸，输入尺寸=n个项目+1，输入长度=1，名称='item\u emb'）
pos\u item\u嵌入=item\u嵌入（pos\u item\u输入）
负项嵌入=项嵌入（负项输入）
user\u vecs=Flatten（）（user\u嵌入）
pos\u item\u vecs=展平（）（pos\u item\u嵌入）
负项向量=展平（）（负项嵌入）
#三重态损失函数
输出=串联（[用户向量、正项向量、负项向量]）
损耗=λ（bpr\u三重态损耗，（1，）（输出）
模型=模型（输入=[锚定，正，负]，输出=损失）
model.compile（优化器='Adam'，loss='mse'，
指标=[“mae”]）
#定义和编译模型
#模型=模型（输入=[用户输入，正项输入，负项输入]，输出=输出）
#compile（optimizer='Adam'，loss=bpr\u triplet\u loss，metrics=[bpr\u triplet\u loss]）
回归模型
model=getModel（n_用户，n_项目）#67063040

当我运行这段代码时，我得到以下（开始令人沮丧）错误

---------------------------------------------------------------------------
OperatorNotAllowedInGraphError回溯（最近一次呼叫最后一次）
/opt/conda/lib/python3.7/site-packages/tensorflow\u core/python/keras/engine/base\u layer.py in\uuuuuu调用（self，input，*args，**kwargs）
841将auto_control_deps.AutomaticControlDependencies（）作为acd：
-->842输出=呼叫fn（转换输入，*args，**kwargs）
843#在`tf.identity`中的`outputs`中包裹张量以避免
/调用中的opt/conda/lib/python3.7/site-packages/tensorflow_core/python/keras/layers/core.py（self、input、mask、training）
794带变量\u范围。变量\u创建者\u范围（self.\u变量\u创建者）：
-->795返回self.函数（输入，**参数）
796
/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/def_function.py in __调用（self，*args，**kwds）
456跟踪\u计数=自。\u获取\u跟踪\u计数（）
-->457结果=自调用（*args，**kwds）
458如果跟踪计数==self.\u获取跟踪计数（）：
/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/def_function.py in_调用（self，*args，**kwds）
502初始值设定项\u map=object\u identity.ObjectIdentityDictionary（）
-->503自我初始化（参数、KWD、添加初始值设定项到=初始值设定项映射）
504最后：
/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/def_function.py in_initialize（self、args、kwds、add_initializers_to）
407 self._stateful_fn._get_concrete_function_internal_garbage_collected（#pylint:disable=protected access
-->408*args，**科威特第纳尔）
409
/opt/conda/lib/python3.7/site-packages/tensorflow\u core/python/eager/function.py in\u get\u concrete\u function\u internal\u garbage\u collected（self，*args，**kwargs）
1847 args，kwargs=None，None
->1848图形函数，u，u=self._可能定义函数（args，kwargs）
1849返回图函数
/opt/conda/lib/python3.7/site-packages/tensorflow\u core/python/eager/function.py in\u maybe\u define\u function（self、args、kwargs）
2149如果图形_函数为无：
->2150图形函数=自身。创建图形函数（args、kwargs）
2151 self.\u function\u cache.primary[cache\u key]=图形\u函数
/opt/conda/lib/python3.7/site-packages/tensorflow\u core/python/eager/function.py in\u create\u graph\u function（self、args、kwargs、override\u flat\u arg\u shapes）
2040覆盖平面形状=覆盖平面形状，
->2041按值捕获=自身。_按值捕获），
2042自我功能属性，
/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/func_graph.py in func_graph_from_py func（name，python_func，args，kwargs，signature，func_graph，autograph，autograph_options，add_control_依赖项，arg_名称，op_返回_值，集合，按值捕获，覆盖平面__arg_形状）
914
-->915 func_outputs=python_func（*func_args，**func_kwargs）
916
/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/def_function.py in wrapped_fn（*args，**kwds）
357#函数对自身进行弱引用，以避免引用循环。
-->358返回弱_-wrapped_-fn（）
359弱包裹的=weakref.ref（包裹的）
/包装器中的opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/func_graph.py（*args，**kwargs）
904如果hasattr（e，“ag\u错误\u元数据”）：
---------------------------------------------------------------------------
OperatorNotAllowedInGraphError            Traceback (most recent call last)
/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/keras/engine/base_layer.py in __call__(self, inputs, *args, **kwargs)
    841                   with auto_control_deps.AutomaticControlDependencies() as acd:
--> 842                     outputs = call_fn(cast_inputs, *args, **kwargs)
    843                     # Wrap Tensors in `outputs` in `tf.identity` to avoid

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/keras/layers/core.py in call(self, inputs, mask, training)
    794     with variable_scope.variable_creator_scope(self._variable_creator):
--> 795       return self.function(inputs, **arguments)
    796 

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/def_function.py in __call__(self, *args, **kwds)
    456     tracing_count = self._get_tracing_count()
--> 457     result = self._call(*args, **kwds)
    458     if tracing_count == self._get_tracing_count():

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/def_function.py in _call(self, *args, **kwds)
    502       initializer_map = object_identity.ObjectIdentityDictionary()
--> 503       self._initialize(args, kwds, add_initializers_to=initializer_map)
    504     finally:

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/def_function.py in _initialize(self, args, kwds, add_initializers_to)
    407         self._stateful_fn._get_concrete_function_internal_garbage_collected(  # pylint: disable=protected-access
--> 408             *args, **kwds))
    409 

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/function.py in _get_concrete_function_internal_garbage_collected(self, *args, **kwargs)
   1847       args, kwargs = None, None
-> 1848     graph_function, _, _ = self._maybe_define_function(args, kwargs)
   1849     return graph_function

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/function.py in _maybe_define_function(self, args, kwargs)
   2149         if graph_function is None:
-> 2150           graph_function = self._create_graph_function(args, kwargs)
   2151           self._function_cache.primary[cache_key] = graph_function

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/function.py in _create_graph_function(self, args, kwargs, override_flat_arg_shapes)
   2040             override_flat_arg_shapes=override_flat_arg_shapes,
-> 2041             capture_by_value=self._capture_by_value),
   2042         self._function_attributes,

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/func_graph.py in func_graph_from_py_func(name, python_func, args, kwargs, signature, func_graph, autograph, autograph_options, add_control_dependencies, arg_names, op_return_value, collections, capture_by_value, override_flat_arg_shapes)
    914 
--> 915       func_outputs = python_func(*func_args, **func_kwargs)
    916 

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/eager/def_function.py in wrapped_fn(*args, **kwds)
    357         # the function a weak reference to itself to avoid a reference cycle.
--> 358         return weak_wrapped_fn().__wrapped__(*args, **kwds)
    359     weak_wrapped_fn = weakref.ref(wrapped_fn)

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/func_graph.py in wrapper(*args, **kwargs)
    904             if hasattr(e, "ag_error_metadata"):
--> 905               raise e.ag_error_metadata.to_exception(e)
    906             else:

OperatorNotAllowedInGraphError: in converted code:

    <ipython-input-2-d2d9c7117621>:15 bpr_triplet_loss  *
        anchor_latent, positive_item_latent, negative_item_latent  = inputs
    /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/ops.py:547 __iter__
        self._disallow_iteration()
    /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/ops.py:540 _disallow_iteration
        self._disallow_when_autograph_enabled("iterating over `tf.Tensor`")
    /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/ops.py:518 _disallow_when_autograph_enabled
        " decorating it directly with @tf.function.".format(task))

    OperatorNotAllowedInGraphError: iterating over `tf.Tensor` is not allowed: AutoGraph did not convert this function. Try decorating it directly with @tf.function.


During handling of the above exception, another exception occurred:

TypeError                                 Traceback (most recent call last)
<ipython-input-8-083f67ca66a0> in <module>
----> 1 model = getModel(n_users, n_items)

<ipython-input-6-75edf8b5cf85> in getModel(n_users, n_items, emb_dim)
     20     # Triplet loss function
     21     output = concatenate([user_vecs, pos_item_vecs, neg_item_vecs])
---> 22     loss = Lambda(bpr_triplet_loss, (1,))(output)
     23 
     24     model = Model(inputs=[anchor, positive, negative], outputs=loss)

/opt/conda/lib/python3.7/site-packages/tensorflow_core/python/keras/engine/base_layer.py in __call__(self, inputs, *args, **kwargs)
    852                               'dynamic. Pass `dynamic=True` to the class '
    853                               'constructor.\nEncountered error:\n"""\n' +
--> 854                               str(e) + '\n"""')
    855           else:
    856             # We will use static shape inference to return symbolic tensors

TypeError: You are attempting to use Python control flow in a layer that was not declared to be dynamic. Pass `dynamic=True` to the class constructor.
Encountered error:
"""
in converted code:

    <ipython-input-2-d2d9c7117621>:15 bpr_triplet_loss  *
        anchor_latent, positive_item_latent, negative_item_latent  = inputs
    /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/ops.py:547 __iter__
        self._disallow_iteration()
    /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/ops.py:540 _disallow_iteration
        self._disallow_when_autograph_enabled("iterating over `tf.Tensor`")
    /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/framework/ops.py:518 _disallow_when_autograph_enabled
        " decorating it directly with @tf.function.".format(task))

    OperatorNotAllowedInGraphError: iterating over `tf.Tensor` is not allowed: AutoGraph did not convert this function. Try decorating it directly with @tf.function.

"""


    <ipython-input-2-d2d9c7117621>:15 bpr_triplet_loss  *
        anchor_latent, positive_item_latent, negative_item_latent  = inputs