Map failed to load / Texture not latest version - Getting started with RL on 0 A.D

[LazyGreyMatter]

Based on the wiki: Getting Started with Reinforcement Learning, I have been unsuccessfully trying to execute the scenarios and train the model. The code keeps failing when trying to load the maps from zero_ad_rl. The maps zipped as "0ad_rl_maps.zip" give out either of the two errors:

• Unable to load textures - too old - (Temp fix by changing the scenario version from 6 to 7 (as is the current FILE_READ_VERSION in source code)).
• Unable to load map - check application logs (which basically throws .pmp File_OpenFailed error)

Since I am new to the modding part of this, can anyone please direct me as to how to fix these issues, especially the pmp file not loading error?

 

[Lion.Kanzen]

Let me help you.

@Stan`

[LazyGreyMatter]

Sure, any help is greatly appreciated!

 

[Lion.Kanzen]

8 minutes ago, LazyGreyMatter said:

Sure, any help is greatly appreciated!

 

While you wait for an official response, I want you to see the following link, to see if it is not related to your topic.

 

 

 

[LazyGreyMatter]

Appreciate you sending me the link! Unfortunately, that did not solve my problem... As you can see, even after running the script (without errors) to version 7, this is the error I am getting - 

< pmp File_OpenFailed >

[Lion.Kanzen]

@Angen @Freagarach

------

By this time, our programmers must be sleeping. (Euro zone time)

[Stan`]

0ad_rl_maps.zip

Try this mod. ping @irishninja

[Langbart]

I followed the instructions from the article wiki/GettingStartedReinforcementLearning, but I run into an issue after this command

picus@Picus zero_ad_rl-master % python3 -m zero_ad_rl.train --help

Traceback (most recent call last):
  File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.framework/Versions/3.8/lib/python3.8/runpy.py", line 193, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.framework/Versions/3.8/lib/python3.8/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/Users/picus/0ad/source/tools/rlclient/zero_ad_rl-master/zero_ad_rl/train.py", line 5, in <module>
    from ray.rllib.train import create_parser, run
  File "/Library/Python/3.8/site-packages/ray/__init__.py", line 63, in <module>
    import ray._raylet  # noqa: E402
  File "python/ray/_raylet.pyx", line 20, in init ray._raylet
ImportError: dlopen(/Library/Python/3.8/site-packages/ray/thirdparty_files/setproctitle.cpython-38-darwin.so, 2): Symbol not found: _Py_GetArgcArgv
  Referenced from: /Library/Python/3.8/site-packages/ray/thirdparty_files/setproctitle.cpython-38-darwin.so
  Expected in: flat namespace
 in /Library/Python/3.8/site-packages/ray/thirdparty_files/setproctitle.cpython-38-darwin.so
picus@Picus zero_ad_rl-master % 

Procedure macOS 10.15.7 (8/Apr/21)

Spoiler

Download & Install 0ad

• Download and install 0 AD from source (just follow wiki/BuildInstructions)

Download & Install Python 3.8

• Install Homebrew and Python@3.8
  • I had to install "python@3.8" because "python@3.9" did not work for me. Make sure you uninstall "python@3.9" completely, otherwise I had problems installing "ray". (See docs.ray.io/en/master/installation.html)

Install 'zero_ad' Python Client

• navigate to this folder in the 0 A.D. source folder.
  • ~/0ad/source/tools/rlclient/python
• run
  • sudo -H python3 -m pip install .
  • Confirmation message: Successfully installed zero-ad-0.0.1
• run
  • sudo -H python3 -m pip install -r requirements-dev.txt
  • Confirmation message: Successfully installed iniconfig-1.1.1 pluggy-0.13.1 py-1.10.0 pytest-6.2.3 toml-0.10.2
• running tests
  • 0ad/binaries/system/pyrogenesis --rl-interface=127.0.0.1:6000 --autostart-nonvisual --mod=public
  • python3 -m pytest
  • Output

    • picus@Picus python % python3 -m pytest
      ============================== test session starts ===============================

      
      platform darwin -- Python 3.8.2, pytest-6.2.3, py-1.10.0, pluggy-0.13.1
      rootdir: /Users/picus/0ad/source/tools/rlclient/python
      collected 10 items

      tests/test_actions.py ...... [ 60%]
      tests/test_evaluate.py .... [100%]

      ============================== 10 passed in 22.50s ===============================

• We can check that our installation worked by running one of the example scripts.
  • python3 ./simple-example.py
  • Output

    • picus@Picus samples % python3 ./simple-example.py
      female citizen's max health is 34.99985
      {'id': 4753, 'template': 'units/spart/support_female_citizen', 'position': [612.9468231201172, 619.9051971435547], 'angle': 0.756927490234375, 'hitpoints': 34.99985, 'invulnerability': False, 'owner': 1, 'idle': False, 'stance': 'passive', 'unitAIState': 'INDIVIDUAL.REPAIR.APPROACHING', 'unitAIOrderData': [{'target': 4776, 'autocontinue': True, 'force': True}], 'resourceCarrying': [], 'garrisonHolderID': 0}

Download & Install 'zero_ad_rl'

• Download "zero_ad_rl" from here: github.com/brollb/zero_ad_rl 
  • First install the “ray[rllib]” and then install "zero_ad_rl" via command
  • sudo -H python3 -m pip install 'ray[rllib]' (pypi.org/project/ray/)
  • sudo -H python3 -m pip install -e .
  • Confirmation message: Successfully installed zero-ad-rl
• In the "zero_ad_rl" folder you will find a zip file named '0ad_rl_maps.zip', unzip it and change the name of the folder from "0ad_rl_maps" to "rl-scenarios" and add it to your 0ad mods folder (the mod.json file names this mod "rl-scenarios", but if the mod folder is not changed as well, the maps will not be displayed).
  • if you open the game with the mod enabled you should see under Scenario some new maps e.g. "Cavalry Vs Spearmen" or "Cavalry Vs Slingers"
  • Some maps use the old version 6 in their XML files, you have to update them from version 6 to version 7 with Phab:P232 otherwise it will fail
    • But after I did that I still get 'entity' and 'CCacheLoader' errors for maps like "CavalaryVSInfantry", only "CavalryVsSlingers.xml" and "CavalryVsSpearmen.xml" worked for me.
  • UPDATE: Stan fixed this problem, by updating the maps 

Training our agent

• pyrogenesis --rl-interface=127.0.0.1:6000 --autostart-nonvisual --mod=rl-scenarios --mod=public
• python3 -m zero_ad_rl.train --help
  • error 

UPDATE: As described here (github.com/ray-project/ray/issues/10428), it does work with 3.8.5 but not with 3.8.9 (the current python@3.8 homebrew version)

Edited April 8, 2021 by Langbart
added python 3.8.5 remark

[Stan`]

you're missing the argparse module I think

[Langbart]

@LazyGreyMatter is your problem fixed? I got it to work on my computer. I briefly had a problem as you described, but the reason was that I forgot to include --mod=rl-scenarios while running the RL interface:

pyrogenesis --rl-interface=127.0.0.1:6000 --autostart-nonvisual --mod=rl-scenarios --mod=public