Thanks a lot for your feedback Stas. I will work on the two usability comments.

Regarding the out of memory error which is not allowing you to complete lesson 1, do you see this from original course notebook? Or it shows up with your changes for heavy utilisation? Can you let me know the cell block where you see this error?
We have hundreds of fast.ai students using Clouderizer, I am surprised no one has brought this up so far. But based on the error you described above, obviously something is missing.

My opinion of getting full Tesla K80 GPU from Colab was based on output from nvidia-smi tool.

I might be wrong in assuming above means full GPU RAM is available to us.