Hey there! Let's break down how to fix your problems with TensorFlow in Colab step by step:

1. Why tf.enable_eager_execution() throws an error? #

In TensorFlow 2.x, Eager Execution is enabled by default—you don't need to call that function anymore! That's exactly why you're hitting the AttributeError.

To confirm Eager Execution is working properly, just run a simple tensor operation directly:

import tensorflow as tf
print(tf.executing_eagerly())  # Should return True
# Test a basic tensor calculation
x = tf.constant(15)
y = tf.constant(25)
print(x + y)  # Will output tf.Tensor(40, shape=(), dtype=int32) right away

2. Fixing TensorFlow version downgrade issue #

When you run !pip install tensorflow==1.12.0 and still see the 2.0-dev version, it's because Colab's runtime hasn't loaded the newly installed package yet. Here's the fix:

• First, run this command to force a clean reinstall of the old version:

  !pip install --upgrade --force-reinstall tensorflow==1.12.0
  

• Then, go to the top menu: Runtime > Restart runtime (this is critical—Colab keeps the old TF version in memory until you restart the environment).
• After the runtime restarts, import TensorFlow and check the version again:

  import tensorflow as tf
  print(tf.__version__)  # Should now show 1.12.0
  

Once you're on TF 1.12, you can use tf.enable_eager_execution() as you originally intended.

Quick Recap #

• For TF 2.x: Skip the enable function—Eager Execution is already active by default.
• To switch to TF 1.x: Force reinstall the desired version, then restart the runtime to apply the change.

内容的提问来源于stack exchange，提问作者ginjaemocoes