Feature Request: Capability to fix the seed for all keras kernel initializers

[mhaghighat]

System information

• TensorFlow version (you are using): 2.0-beta
• Are you willing to contribute it (Yes/No): No

Describe the feature and the current behavior/state.
If you create a Keras model, the only way to make your model reproducible is to repeat setting the seed for the kernel_initializer in every layer. This makes the code bulky and more prone to error. For example, a reproducible code looks like this:

np.random.seed(1)
tf.set_random_seed(1)
# FEATURE REQUEST for another parameter like the ones above that we can set
# here and avoid repeating initializer seed in each and every layer below

model = tf.keras.Sequential()
model.add(tf.keras.layers.Dense(8, activation='relu', kernel_initializer=keras.initializers.glorot_uniform(seed=1), input_shape=[1]))
model.add(tf.keras.layers.Dense(8, activation='relu', kernel_initializer=keras.initializers.glorot_uniform(seed=1)))
model.add(tf.keras.layers.Dense(1, activation='linear', kernel_initializer=keras.initializers.glorot_uniform(seed=1)))

I wonder if you can add a default seed parameter that we can set and it affects all the initializers.

Will this change the current api? How?
No

Who will benefit with this feature?
Anyone who likes clean code and would like to create reproducible models.

[jvishnuvardhan]

@mhaghighat Sorry for the late response.
Look like you are looking for the following feature.
https://www.tensorflow.org/api_docs/python/tf/config/experimental/enable_op_determinism

If you think the above is not the feature you are looking, please feel free to open a PR in keras-team/keras repo. repository.

Please note that Keras development moved to keras-team/keras repository to focus on only keras. Thanks!

[google-ml-butler]

This issue has been automatically marked as stale because it has no recent activity. It will be closed if no further activity occurs. Thank you.

[google-ml-butler]

Closing as stale. Please reopen if you'd like to work on this further.