tensorflow: ModuleNotFoundError: No module named 'tensorflow.compat.v2'

I am having the same error as well

@cpoptic just want to add my observation here, in my understanding tensorflow-probability is not installed along with tensorflow. Could please try with your test script after installing pip install --user --upgrade tensorflow-probability Hope this will help you

I am having the same error.

I have tensorflow 1.4.0 with Cuda 8 and cudnn 7. And I get the following error.

import tensorflow.compat.v1 as tf ImportError: No module named 'tensorflow.compat'

After installing tensorflow-probability, I still get the same error.

@MarkDaoust I fixed it by install tf-nightly-gpu.

!pip install -U "tensorflow==2.0"

I was facing similar issues but this solved the problem for me. Thanks!

One common way to get stuck in a particular tensorflow version is to have an old version of python or pip.

If you try to install a specific version and it fails then check your python and pip versions.

pip install -U "tensorflow==1.15.*"
pip install -U "tensorflow==2.2.*"

pip install tensorflow only gets you the latest version if those are up to date.

If your problem is with RASA CORE, in my case, I executed the following commands: pip install -U tensorflow==1.15 pip install -U tensorflow==2.0 pip install -U gast==0.3.3 pip install -U tf-nightly-gpu pip install -U tensorboard==1.15.0 pip install -U tensorflow==1.15

@himanshisyadav upgrade to at least TF 1.14.

Hi Everyone,

When this was published, it required tf-nightly or any of preview releases of tensorflow -2.0. That was the original source of confusion.

The tf.compat.v2 module was added in 1.14. Upgrade to 1.14 , 1.15, or 2.0 and this will work fine.

pip install -U "tensorflow==1.15" 
pip install -U "tensorflow==2.0" 

There is absolutely no way to run it with 1.13, or 1.8, sorry!

I have changed with import tensorflow.compat.v2 as tf v2 to v1 as wll but it not work ModuleNotFoundError: No module named ‘tensorflow.compat’

你的来信已收到，谢谢！

你的来信已收到，谢谢！

is this problem solved? please help I faced the same error.

Yes that works, but the issue is with the tensorflow_probability

# import tensorflow.compat.v2 as tf
import tensorflow.compat.v1 as tf
import tensorflow_probability as tfp

tf.enable_v2_behavior()

AttributeError: module ‘tensorflow.compat’ has no attribute ‘v2’

AttributeError Traceback (most recent call last) <ipython-input-10-6713d3750cfa> in <module> 17 # import tensorflow.compat.v2 as tf 18 import tensorflow.compat.v1 as tf —> 19 import tensorflow_probability as tfp 20 21 tf.enable_v2_behavior()

/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow_probability/init.py in <module> 76 77 # from tensorflow_probability.google import staging # DisableOnExport —> 78 from tensorflow_probability.python import * # pylint: disable=wildcard-import 79 from tensorflow_probability.python.version import version 80 # pylint: enable=g-import-not-at-top

/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow_probability/python/init.py in <module> 27 from tensorflow_probability.python import mcmc 28 from tensorflow_probability.python import monte_carlo —> 29 from tensorflow_probability.python import optimizer 30 from tensorflow_probability.python import positive_semidefinite_kernels 31 from tensorflow_probability.python import stats

/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow_probability/python/optimizer/init.py in <module> 28 from tensorflow_probability.python.optimizer.proximal_hessian_sparse import minimize as proximal_hessian_sparse_minimize 29 from tensorflow_probability.python.optimizer.proximal_hessian_sparse import minimize_one_step as proximal_hessian_sparse_one_step —> 30 from tensorflow_probability.python.optimizer.sgld import StochasticGradientLangevinDynamics 31 from tensorflow_probability.python.optimizer.variational_sgd import VariationalSGD 32

/anaconda3/envs/tf/lib/python3.7/site-packages/tensorflow_probability/python/optimizer/sgld.py in <module> 32 33 —> 34 class StochasticGradientLangevinDynamics(tf.compat.v2.optimizers.Optimizer): 35 “”"An optimizer module for stochastic gradient Langevin dynamics. 36

AttributeError: module ‘tensorflow.compat’ has no attribute ‘v2’