Question

我正在尝试将传递给我的一些TensorFlow代码集成到Django应用程序中。

如果我做一个标准的Django python manage runserver，第一个预测（在第一次启动服务器时）起作用，而每个其他预测都会产生异常 - 如下所示。

如果我执行以下操作，我用Google搜索，python manage runserver --nothreading我不再获得例外。但是，我以前从来没有这么做过，我担心它的影响，特别是当我没有这个选项时，一旦转移到生产部署。

我对TensorFlow不太了解，但我最终想出了第一个预测之后的第二个解决方案，使用graph = tf.get_default_graph()然后使用with graph.as_default()进行每次预测...但我不知道真的知道这完全是什么，或者它是否是一个好主意。

异常

Traceback (most recent call last):
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/tensorflow/python/client/session.py", line 1075, in _run
    subfeed, allow_tensor=True, allow_operation=False)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/tensorflow/python/framework/ops.py", line 3590, in as_graph_element
    return self._as_graph_element_locked(obj, allow_tensor, allow_operation)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/tensorflow/python/framework/ops.py", line 3669, in _as_graph_element_locked
    raise ValueError("Tensor %s is not an element of this graph." % obj)
ValueError: Tensor Tensor("Placeholder:0", shape=(363, 27), dtype=float32) is not an element of this graph.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/django/core/handlers/exception.py", line 35, in inner
    response = get_response(request)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/django/core/handlers/base.py", line 128, in _get_response
    response = self.process_exception_by_middleware(e, request)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/django/core/handlers/base.py", line 126, in _get_response
    response = wrapped_callback(request, *callback_args, **callback_kwargs)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/django/views/generic/base.py", line 69, in view
    return self.dispatch(request, *args, **kwargs)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/django/utils/decorators.py", line 62, in _wrapper
    return bound_func(*args, **kwargs)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/django/contrib/auth/decorators.py", line 21, in _wrapped_view
    return view_func(request, *args, **kwargs)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/django/utils/decorators.py", line 58, in bound_func
    return func.__get__(self, type(self))(*args2, **kwargs2)
  File "/home/pembo13/virt/myproject/myproject/myproject/frontend/views.py", line 84, in dispatch
    return self.handle(request, self.client)
  File "/home/pembo13/virt/myproject/myproject/myproject/frontend/views.py", line 196, in handle
    os.path.join(settings.BASE_DIR, 'ai', 'model', 'categoryModel.pkl') # modelPath
  File "/home/pembo13/virt/myproject/myproject/ai/__init__.py", line 50, in predictCategory
    model = modeler.loadModel(modelPath)
  File "/home/pembo13/virt/myproject/myproject/ai/./modules/modeler.py", line 36, in loadModel
    model.set_weights(modelPkl['modelWeights'])
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/keras/engine/network.py", line 515, in set_weights
    K.batch_set_value(tuples)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/keras/backend/tensorflow_backend.py", line 2435, in batch_set_value
    get_session().run(assign_ops, feed_dict=feed_dict)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/tensorflow/python/client/session.py", line 900, in run
    run_metadata_ptr)
  File "/home/pembo13/virt/myproject/lib64/python3.6/site-packages/tensorflow/python/client/session.py", line 1078, in _run
    'Cannot interpret feed_dict key as Tensor: ' + e.args[0])
TypeError: Cannot interpret feed_dict key as Tensor: Tensor Tensor("Placeholder:0", shape=(363, 27), dtype=float32) is not an element of this graph.

Answer 1

在加载或构建模型后，立即保存TensorFlow图：

tf.keras.backend.clear_session()
graph = tf.get_default_graph()
global graph
with graph.as_default():
    (... do the prediction here ...)

似乎对此问题进行了长时间的讨论here

Django开发服务器中TensorFlow预测的问题（带线程）

1 个答案: