Tf 2.6 incompatible with matplotlib

Course Q&A TensorFlow: Advanced Techniques Specialization Advanced Computer Vision with TensorFlow
1

Tf 2.6 is incompatible with numpy version required for matplotlib
Matplotlib requires numpy>=1.20; whereas tf 2.6 works with numpy 1.19.5.

When using google collab’s default version for tf, the auto grader is not able to load the model.

2

Hi @ANISHA_JAIN,

There are some incompatibilities with the new TensorFlow version in Colab with the Grader. Please check if you have these pip installs at the top of your notebook:

# Install packages for compatibility with the autograder
!pip install tensorflow==2.6.0 --quiet
!pip install keras==2.6.0 --quiet

hope it helps!

1 Like
3

Hi @Pere_Martra if this is an upgrade that is creating conflicts I suggest you raise an issue at our github account of this specilization.

1 Like
4

Hi Gent!

Thanks, I think it can be the same problem that @chris.favila reported in february, maybe Anisha is using an old notebook without the modification to install tensorflow 2.6.

If the problem persist i will open an issue in the github account of the specialization.