Meta-DO

Implementation of Detect and Act: Automated Dynamic Optimizer through Meta-Black-Box Optimization (Meta-DO).

Dependencies

python==3.11.5
numpy==1.26.4
pygame>=2.6.1
torch==2.6.0
torchvision==0.21.0
torchaudio==2.6.0
metaevobox  # see https://github.com/MetaEvo/MetaBox for more details

Train

The train process can be easily activated via the command below:

python model_trainer.py

Recording results: Log files will be saved to ./output/train_log/ . The saved checkpoints will be saved to ./agent_model/train/. The file structure is as follow:

|--agent_model
   |--train
      |--GLEET
         |--run_Name
            |--checkpoint0.pkl
            |--checkpoint1.pkl
            |--...

|--output
   |--tensorboard
   |--test
   |--train_log

Test

The test process can be easily activated via the command below:

python model_tester.py

The test results will be saved to ./output/test/.