Switch to !pip install tensorflow-data-validation --user then restart kernel then run second cell to fix module not installing

[CookingWithCale]

Problem

The problem is that when you run !pip install tensorflow-data-validation on Vertex AI and Colab the module doesn't install correctly because of a mismatched version of pyarrow.

Known Files Affected

1. ~/courses/machine_learning/deepdive2/production_ml/labs/tfdv_basic_spending.ipynb

Solution

I took the lab 3 times and 3 times I was able to get it to install by adding !pip install tensorflow-data-validation --user to the first cell, then restarting the kernel, and then not running the first cell, but then run the second cell, at which time I've gotten it to work 3 out of 3 times so I'm 99.999% sure it's the fix.

I just tried it a fourth time and I did get one fail, but it worked the second time.

This official Tensorflow tutorial on TFDV says use this, but I have not tried yet, still doing the lab:

!pip download tensorflow_data_validation \
  --no-deps \
  --platform manylinux2010_x86_64 \
  --only-binary=:all: