Optimizer import raises an error with keras 2.7

[frgfm]

System information.

• Have I written custom code (as opposed to using a stock example script provided in Keras): no
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 20.04 LTS
• TensorFlow installed from (source or binary): binary
• TensorFlow version (use command below): 2.6.0
• Python version: 3.8.10
• Bazel version (if compiling from source): N/A
• GPU model and memory: NVIDIA GeForce RTX 2070 with Max-Q Design

Describe the problem.

I encountered a CI problem with a build job today that wasn't happening yesterday. So I checked the difference in terms of dependency and the only difference was keras. So I inspected the traceback and ended up tracking the import from keras that causes trouble. Please note this is a big issue, because the original call was from tensorflow from tensorflow.keras.utils import img_to_array and the underlying troubling import is from keras.

Describe the current behavior.

Running the standalone code throws a AlreadyExistsError

Describe the expected behavior.

Not raising an error.

• Do you want to contribute a PR? (yes/no): happy to do so, but I'm not sure how to solve this
• If yes, please read this page for instructions
• Briefly describe your candidate solution(if contributing):

Standalone code to reproduce the issue.

from keras import optimizers

Source code / logs.

---------------------------------------------------------------------------
AlreadyExistsError                        Traceback (most recent call last)
<ipython-input-1-d1acaf7a1f6f> in <module>
----> 1 from keras import optimizers

~/miniconda3/lib/python3.8/site-packages/keras/__init__.py in <module>
     23 
     24 # See b/110718070#comment18 for more details about this import.
---> 25 from keras import models
     26 
     27 from keras.engine.input_layer import Input

~/miniconda3/lib/python3.8/site-packages/keras/models.py in <module>
     18 import tensorflow.compat.v2 as tf
     19 from keras import backend
---> 20 from keras import metrics as metrics_module
     21 from keras import optimizer_v1
     22 from keras.engine import functional

~/miniconda3/lib/python3.8/site-packages/keras/metrics.py in <module>
     24 
     25 import numpy as np
---> 26 from keras import activations
     27 from keras import backend
     28 from keras.engine import base_layer

~/miniconda3/lib/python3.8/site-packages/keras/activations.py in <module>
     18 
     19 from keras import backend
---> 20 from keras.layers import advanced_activations
     21 from keras.utils.generic_utils import deserialize_keras_object
     22 from keras.utils.generic_utils import serialize_keras_object

~/miniconda3/lib/python3.8/site-packages/keras/layers/__init__.py in <module>
     21 
     22 # Generic layers.
---> 23 from keras.engine.input_layer import Input
     24 from keras.engine.input_layer import InputLayer
     25 from keras.engine.input_spec import InputSpec

~/miniconda3/lib/python3.8/site-packages/keras/engine/input_layer.py in <module>
     19 from keras import backend
     20 from keras.distribute import distributed_training_utils
---> 21 from keras.engine import base_layer
     22 from keras.engine import keras_tensor
     23 from keras.engine import node as node_module

~/miniconda3/lib/python3.8/site-packages/keras/engine/base_layer.py in <module>
     41 from keras.engine import node as node_module
     42 from keras.mixed_precision import autocast_variable
---> 43 from keras.mixed_precision import loss_scale_optimizer
     44 from keras.mixed_precision import policy
     45 from keras.saving.saved_model import layer_serialization

~/miniconda3/lib/python3.8/site-packages/keras/mixed_precision/loss_scale_optimizer.py in <module>
     16 
     17 from keras import backend
---> 18 from keras import optimizers
     19 from keras.mixed_precision import loss_scale as keras_loss_scale_module
     20 from keras.optimizer_v2 import optimizer_v2

~/miniconda3/lib/python3.8/site-packages/keras/optimizers.py in <module>
     24 from keras.optimizer_v1 import Optimizer
     25 from keras.optimizer_v1 import TFOptimizer
---> 26 from keras.optimizer_v2 import adadelta as adadelta_v2
     27 from keras.optimizer_v2 import adagrad as adagrad_v2
     28 from keras.optimizer_v2 import adam as adam_v2

~/miniconda3/lib/python3.8/site-packages/keras/optimizer_v2/adadelta.py in <module>
     20 import numpy as np
     21 from keras import backend_config
---> 22 from keras.optimizer_v2 import optimizer_v2
     23 from tensorflow.python.util.tf_export import keras_export
     24 

~/miniconda3/lib/python3.8/site-packages/keras/optimizer_v2/optimizer_v2.py in <module>
     34 
     35 
---> 36 keras_optimizers_gauge = tf.__internal__.monitoring.BoolGauge(
     37     "/tensorflow/api/keras/optimizers", "keras optimizer usage", "method")
     38 

~/miniconda3/lib/python3.8/site-packages/tensorflow/python/eager/monitoring.py in __init__(self, name, description, *labels)
    358       *labels: The label list of the new metric.
    359     """
--> 360     super(BoolGauge, self).__init__('BoolGauge', _bool_gauge_methods,
    361                                     len(labels), name, description, *labels)
    362 

~/miniconda3/lib/python3.8/site-packages/tensorflow/python/eager/monitoring.py in __init__(self, metric_name, metric_methods, label_length, *args)
    133           self._metric_name, len(self._metric_methods)))
    134 
--> 135     self._metric = self._metric_methods[self._label_length].create(*args)
    136 
    137   def __del__(self):

AlreadyExistsError: Another metric with the same name already exists.

[qlzh727]

Please see #15579 for more details. The issue should now be mitigated by using tf 2.6.2.