PyPDF separate installation in upload_docs.ipynb

[nv-edwli]

What is the rationale behind why the pyPDF package is a separate installation within the upload_docs.ipynb jupyter notebook. Why don't we enable AI Workbench to take care of the installation for the user, eg. as part of requirements.txt?

I had a user bring their own custom .pdf file to the project who skipped everything above the unzip cell because they thought they didn't need to unzip anything for their dataset. This means they mistakenly skipped the pip install pypdf cell, which broke the later upload step.