Help Center> ModelArts> FAQs> Notebook> Environment Configurations> How Do I Use TensorFlow 2.X in a Notebook Instance?
View PDF

How Do I Use TensorFlow 2.X in a Notebook Instance?

Notebook instances support TensorFlow 2.1. You can select Multi-Engine 2.0 (Python3) (containing the TensorFlow-2.1.0 engine) to create an instance.

For notebook instances created using other environments, such as Multi-Engine 1.0 (Python3, Recommended) or Multi-Engine 1.0 (Python2), if TensorFlow 2.X is required, perform the following steps:

Procedure

TensorFlow 2.0 can be installed and used in notebook instances. The procedure is as follows:

  1. Create a notebook instance, open and access the notebook instance, and select and access the kernel of TensorFlow 1.13.1.
  2. Go to the Kernel page and install TensorFlow 2.0 (GPU version). Enter the following code and click Run above the code. 
    1
    		 
    !pip install tensorflow-gpu==2.0.0b1
    Figure 1 Installing TensorFlow 2.0
  3. Restart the kernel for the update to take effect.
    Figure 2 Restarting the kernel
  4. After the kernel is restarted, enter the following code to check the installation result. Then, you can use TensorFlow 2.0.
    Run the following code. If 2.0.0-beta1 is displayed, TensorFlow 2.0 is installed. 
    1
2
    		 
    import tensorflow as tf
print(tf.__version__)

    Run the following code. If True is displayed, the GPU version is installed.

    1
    		 
    print(tf.test.is_gpu_available())
    Figure 3 Verifying the installed version