Conflicting Python versions in SageMaker Studio notebook with Python 3.8 kernel

Question

I'm trying to run a SageMaker kernel with Python 3.8 in SageMaker Studio, and the notebook appears to use a separate distribution of Python 3.7. The running app is indicated as tensorflow-2.6-cpu-py38-ubuntu20.04-v1. When I run !python3 -V I get Python 3.8.2. However, the Python instance inside the notebook is different:

import sys
sys.version

gives '3.7.12 | packaged by conda-forge | (default, Oct 26 2021, 06:08:21) \n[GCC 9.4.0]'

Similarly, running %pip -V and %conda info indicates Python 3.7.

Also, import tensorflow fails, as it isn't preinstalled in the Python environment that the notebook invokes.

I'm running in the eu-west-2 region. Is there anything I can do to address this short of opening a support ticket?

Answer 1

are you still facing this issue?

I am in eu-west-2 using a SageMaker Studio notebook and the TensorFlow 2.6 Python 3.8 CPU Optimized image (running app is tensorflow-2.6-cpu-py38-ubuntu20.04-v1).

When I run the below commands, I get the right outputs.

!python3 -V

returns Python 3.8.2

import sys
sys.version 

returns 3.8.2 (default, Dec 9 2021, 06:26:16) \n[GCC 9.3.0]'

import tensorflow as tf
print(tf.__version__)

returns 2.6.2

It seems this has now been fixed