search

alibaba / lightweight-neural-architecture-search

This is a collection of our zero-cost NAS and efficient vision applications.
Apache License 2.0
376 stars 50 forks source link

How to search in custom search space such as HardNet? #2

Closed 1chimaruGin closed 1 year ago

1chimaruGin commented 2 years ago

Could you please provide step by step procedure?

Vericoware commented 1 year ago
  1. Build a "superxxxxx.py" like "SuperResConvK1KXK1.py" (https://github.com/alibaba/lightweight-neural-architecture-search/blob/main/nas/models/blocks/SuperResConvK1KXK1.py)
  2. Implement "entropy_forward" or "get_log_zen_score" functions in your "superxxxxx.py"
  3. Write your own "masterhardnet.py" like "masternet.py" (https://github.com/alibaba/lightweight-neural-architecture-search/blob/main/nas/models/masternet.py).
  4. Specialize your own "configxxx.py" for searching like "config_nas.py" (https://github.com/alibaba/lightweight-neural-architecture-search/blob/main/configs/config_nas.py).
  5. Design your own search space file "space_xxxx.py" like "space_K1KXK1.py" (https://github.com/alibaba/lightweight-neural-architecture-search/blob/main/nas/spaces/space_K1KXK1.py).
  6. Finally, run and debug.
1chimaruGin commented 1 year ago

Got it.

Thank you!@Vericoware