Whats the best setup/workflow for floydhub?

I am new to the technologies like jupyter and floydhub; I am trying to get some recommendations how to use them efficiently while working on the course part one.

My questions:

Would you recommend to create a new project for each lesson or one for the entire course?

Whats the best way of making the dependencies on the vm available?
-> Is there a way to save the dependencies for all jobs/lessons so I don’t have to include e.g. keras every time?

Should I only work with jupyter notebooks running on a floydhub vm or does it make sense to recreate and run the code of the lesson via normal “floyd run” job sometimes?

I would be glad about advice of any kind how to establish a efficient workflow, not limited to this questions; I just don#t want to miss possibilities/features resulting in unnecessary repetitive work or other disadvantages…


I actually like FloydHub, however in this case we need to use Python 3.6, Floyd used 3.5 so you will run into the f strings error further in the notebook. On a positive note I reached out to FloydHub and they will be upgrading to 3.6 hopefully this week.

I would create 1 environment for the whole course.

Not sure if I follow the next question, I believe Floyd uses Keras, Python 3 and tensorflow as the default so you don’t need to include Keras every time.