Debug environment set up

The current directory debug is the working directory.

Pre-requisites:

• Anaconda 1.7
• OpenJDK 18
• Maven 3.8

Linux

1. sh prepare_test.sh

2. conda env create -f ../mdde/support/conda-p310.yml

   • To remove the environment or its old version, run conda env remove --name mdde-p310 --all

3. cd ../docker/compositions/redis/scripts
   sh start_redis_only.sh
   cd ../../../../debug
   

Leave the databases running. To shut down the Redis nodes, press Ctrl-C in the terminal window where the up command was executed or sudo docker-compose -f ../docker/compositions/redis/docker-compose.debug.yml down in another.

Open another window and start the Registry server.

1. java -jar ../registry/mdde-registry/packaged/mdde-registry-tcp.jar -p 8942 -b 8954 -c registry_config.yml

In yet another terminal window, activate conda environment.

5. conda activate mdde-p310

Setup MDDE.

6. cd ../mdde

7. pip install -e ./core

8. pip install -e ./extensions/mdde-registry-client-tcp

9. pip install -e ./extensions/integration-ray

   • You might need to additionally install TensorFlow pip install tensorflow==2.11.0'. Currently it's not installed by default to simplify experimental environments configuration (CPU or GPU).

10. Run debug or sample configuration

    • Debug, without a learner:
      1. cd test
      2. python test_environment.py
    • Sample:
      1. cd samples
      2. Ray RLlib MADDPG: python sample_ray_maddpg.py --debug
         1. [optional] argument --result-dir - Path to results dir (tensorboard)
         2. [optional] argument --temp-dir - Path to where Ray should store temp folder. Make sure it's not too long for the plasma store, otherwise ray will fail.