[HANDS-ON BUG]huggy.zip

[jichuan68]

Hi everyone,

I'm currently working through Bonus Unit 1 (Train Huggy) of the Deep RL course and trying to set up the environment on my local Windows machine.

I'm having trouble finding a working download link for the pre-compiled Windows version of the Huggy.zip environment file (the one containing Huggy.exe).

The link currently in the Colab notebook (https://github.com/huggingface/Huggy/raw/main/Huggy.zip) seems to provide the Linux version (.x86_64 file). I've also tried some other links found online that pointed to a huggingface-projects/Unity-ML-Agents-Environments repository, but those links appear to be broken ("Repository not found").

Could someone please point me to the correct and currently active download link for the Windows version of the Huggy environment needed for this bonus unit?

Thanks so much for your help!

[Vzaldat]

Hi,
The environment file used for training has to be created by you, which is couple of commands. After you have installed the huggy.zip file, you can verify by going to the directory where the Huggy.zip is located and checking if the zip file is extracted.

To create the environment file for Huggy, you should go to ml-agents -> config -> ppo and add a new file, Huggy.yaml and paste the code contents given in the bonus unit. You can use these commands to create the environment file in Windows -

1. type nul > Huggy.yaml (in command prompt)
2. New-Item -Path "Huggy.yaml" -ItemType File. (in Powershell)
3. Add file if you are using a code editor like VSCode and name it Huggy.yaml

Make sure you add this environment file to the ml-agents/config/ppo directory for training commands to work.

Once the file is created, you will have to paste the contents in bonus unit or you can find them here -

behaviors:
  Huggy:
    trainer_type: ppo
    hyperparameters:
      batch_size: 2048
      buffer_size: 20480
      learning_rate: 0.0003
      beta: 0.005
      epsilon: 0.2
      lambd: 0.95
      num_epoch: 3
      learning_rate_schedule: linear
    network_settings:
      normalize: true
      hidden_units: 512
      num_layers: 3
      vis_encode_type: simple
    reward_signals:
      extrinsic:
        gamma: 0.995
        strength: 1.0
    checkpoint_interval: 200000
    keep_checkpoints: 15
    max_steps: 2e6
    time_horizon: 1000
    summary_freq: 50000