Lambda Layer issue

[lab-pc]

    "name_scope was already taken." % abs_state_scope)

RuntimeError: Exception encountered when calling layer "lambda" (type Lambda).

variable_scope module_1/ was unused but the corresponding name_scope was already taken.

Call arguments received by layer "lambda" (type Lambda):
  • inputs=tf.Tensor(shape=(None, 1), dtype=string)
  • mask=None
  • training=True

[Zaibali9999]

fn = module_objects.get(function_name)
AttributeError: 'NoneType' object has no attribute 'get

[Zaibali9999]

Issue
Everything works fine if you define the custom_layer not as a class method, but as a separate function, either within the class method or outside of it. The unexpected behaviour occurs only when this custom_layer is defined as a class method.

[Zaibali9999]

trained_model = load_model(trained_model_path, custom_objects={'make_elmo_embedding': self.make_elmo_embedding})

keras==2.2.5
h5py==2.10.0

[Zaibali9999]

probs = model.predict(test_set_x, batch_size = self.config['training_settings']['network_config']['batch_size'], verbose = self.paras.verbose)
run_metadata_ptr)
tensorflow.python.framework.errors_impl.InternalError: Unsupported object type int

[lihuiuyun]

I have the same problem，have you slove it？

[DanielLin1986]

fn = module_objects.get(function_name) AttributeError: 'NoneType' object has no attribute 'get

The bug has been fixed. Please try it again using the latest code.

[DanielLin1986]

I have the same problem，have you slove it？

The bug has been fixed. Please try again by pulling the latest version of the code.

[DanielLin1986]

probs = model.predict(test_set_x, batch_size = self.config['training_settings']['network_config']['batch_size'], verbose = self.paras.verbose) run_metadata_ptr) tensorflow.python.framework.errors_impl.InternalError: Unsupported object type int

Please run the code using Tensorflow 1.15 and Keras 2.3.1.

[aravinthk00]

I also encounter same issue but tried latest version of code but not yet it is not fixed.