DEPRECATION WARNING: How to remove tf.keras warning "calling VarianceScaling.__init__ with dtype is deprecated..."

Question

The following code generates the warning in tensorflow r1.12 python API:

#!/usr/bin/python3
import tensorflow as tf

M = tf.keras.models.Sequential();
M.add(tf.keras.layers.Dense(2)); 

The complete warning text is this:

WARNING: Logging before flag parsing goes to stderr.
W0213 15:50:07.239809 140701996246848 deprecation.py:506] From /home/matias/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/init_ops.py:1253: calling VarianceScaling.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.
Instructions for updating:
Call initializer instance with the dtype argument instead of passing it to the constructor

I have tried different approaches like initializing and calling a kernel initializer before adding Dense layer and passing it to Dense constructor, but it seems to not change anything. Is this warning inevitable? A 'yes' as an answer would be enough for me.

Answer 1

The warning may be caused upstream by abseil-py, a dependency of tensorflow. See the details here. An easy fix may be to update abseil-py by running:

pip install --upgrade absl-py

(In my case, the conflicting version was 0.7.1 and the problem was fixed in the updated version, 0.8.1)

Answer 2

You are running tensor flow 2.0 and it looks like VarianceScaling.init is deprecated. It might mean that Sequential will need to be more explicitly initialized in the future. for example:

model = tf.keras.Sequential([
# Adds a densely-connected layer with 64 units to the model:
layers.Dense(64, activation='relu', input_shape=(32,)),
# Add another:
layers.Dense(64, activation='relu'),
# Add a softmax layer with 10 output units:
layers.Dense(10, activation='softmax')])