marcometer / endless-memory-gym Goto Github PK

Hi,

I absolutely love this benchmark and I'm planning to use this in a paper I am writing.

I had a question regarding the Searing Spotlights environments. The other two environments, Mortar Mayhem and Mystery Path seems to be solvable by current memory based agents by providing dense reward signals or through grid like movements. However, there was no simpler configuration listed for Searing Spotlights. Is there a configuration available for Searing Spotlights available where a memory-based agent could achieve a decent level performance (which a memory-less agent wouldn't)?

My goal is to use these environments to compare the memory capabilities of different architectures. In Mystery Path and Mortar Mayhem I am able to do so with the simpler variations of these environments. The differences in performance between different memory agents are becoming apparent. But in Searing Spotlights, it seems none of the agents even get to a point where they can leverage their memory, and all agents seem to perform the same. Is there a way to make this environment easier?

Problem Description

The lights never seem to dim for the Searing Spotlights environment, causing the spotlights to be unnoticeable. Messing around with the "Light Parameters" such as "light_dim_off_duration" and "light_threshold" seem to have no effect.

Video Clip

Here is a link to a video clip showing a random policy running in the Searing Spotlights environment. The environment in the video was rendered with "rgb_array" as the render mode. The "debug_rgb_array" render mode outputs the same thing.
Google Drive Link (takes a while to load)

To Reproduce

Running on an M2 Macbook Air with macOS Sonoma 14.0.
Commands to run:

conda create -n memory-gym python=3.11 --yes
conda activate memory-gym
git clone https://github.com/MarcoMeter/drl-memory-gym.git
cd drl-memory-gym
pip install -e .
python memory_gym/searing_spotlights.py

The repository was cloned with commit 7900c0d046680fbaee1712c98daa88a2ad3be3f8.
I have also verified that I am running with PyGame 2.4.0 and Gymnasium 0.29.0.

>>> import pygame
pygame 2.4.0 (SDL 2.26.4, Python 3.11.5)
Hello from the pygame community. https://www.pygame.org/contribute.html
>>> pygame.version.ver
'2.4.0'

>>> import gymnasium
>>> gymnasium.__version__
'0.29.0'

Hi,

In order to install on Linux, this line needs to be changed to not have backwards slashes (Windows) but a forward slash.

https://github.com/MarcoMeter/drl-memory-gym/blob/26c50a052a631d3c11c41a40e75c13cd34141d10/setup.py#L11

Hi,

Is there a reason only Gymnasium v0.26.3 is supported? Is it possible to remove this requirement and support for newer versions of gymnasium by any chance?

Thanks