perdonliu / cse-autoloss Goto Github PK

Hi! I like your work very much, and I want to follow it. But from the paper, some details are not described very clearly. I hope you can help me understand the question：

The loss function can be represented as a directed acyclic graph(DAG), but how can I design the cross-over and mutation for the DAG. Some details of the cross-over and mutation are not found in the paper.

I want to follow your code for my own task, but I did not find the code for searching loss. Could you public loss searching code?
Thank you very much.

Hi colleagues,

First of all, thank you for sharing this work. I am wondering if it is possible to share the 12 epoch trained model with ResNet-50 backbone?

Thank you!
Baris.