Order to run code, doubts on projection.py about darts-pt

[ruocwang]

Hi casarinsof,

Thank you for your interest in our work.

You are on the right track. The projection step refers to the architecture selection, so it is necessary. After the supernet is trained (train_search), you would need to run the architecture selection to pick the final architecture, which is represented as a genotype. Inserting that into the genotype.py file, give it a variable name (e.g. arch1), then run the retrain command in the README.md with "arch1".

Hope that helps!

[casarinsof]

Hello! Thank you for you reply!
Then I have 2 other questions:

• I run the code without the projection step, but juts insertiong the genotype I got after the search and directly running the evaluation, and got the same result of your paper, how so? I mean as I am missing one step, what did happen?
• The architecture selection corresponding to running projection.py, to which step of the algorithm or passage correspond in your paper? It is still not clear to me.
  Thanks a lot