Costa Huang

Following the practice suggested by rliable

1

Hey, just wondering if there is any chance to integrate with rliable, which supports different ways to plot the curves. E.g., Interquartile Mean (IQM), Score distributions (performance profiles), etc. CC...

Rustboard Optional Integration

2

Hey, the rustboard integration looks pretty cool, but it also makes it harder to install. Any chance you could make the rust extension optional during installation? ``` (openrlbenchmark-py3.9) ➜ openrlbenchmark...

Parallel-envs-friendly ppo_continuous_action.py

3

## Description This PR modifies `ppo_continuous_action.py` to make it more parallel-envs-friendly. CC @kevinzakka. The version of `ppo_continuous_action.py` in this PR is different from that in the `master` branch in the...

Brax + PPO integration

2

## Description Test out integration with brax. It seems to work out of the box without having to implement observation normalization — https://wandb.ai/costa-huang/cleanRL/runs/2aemjwey?workspace=user-costa-huang Compilation takes ~400 seconds, and getting 6000...

Hello, I was wondering if it’s possible to detect slurm environments automatically like `jax.distributed.initialize()`. Example: The relevant jax code is here: https://github.com/google/jax/blob/main/jax/_src/clusters/slurm_cluster.py ``` #!/bin/bash #SBATCH --job-name=trl #SBATCH --partition=production-cluster #SBATCH --gpus-per-task=8...

Support wall observation and multiple improvements

2

This PR does a few things. 1. supports the wall observation. 2. fixes https://github.com/Farama-Foundation/MicroRTS-Py/pull/34#issuecomment-1012722258 * one side effect is that the `agent_sota.pt` would no longer work. Hence we deprecate the...

Memory leak issue that @xluox has

1

@xluox ### How to reproduce ``` git clone https://github.com/xluox/gym-microrts.git --recurse-submodules cd gym-microrts git checkout PCG git pull --recurse-submodules poetry install poetry run pip install torch==1.12.1+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html bash build.sh &>...

It doesn't work yet... Cache the changes here. The effort does not seem worth it because the bottleneck is in the environment. ``` ppo_gridnet.py times, step=1.3973662853240967, inference=1.2995522022247314, get_mask=0.5436160564422607, rollout=3.3615562915802, training=0.68623948097229,...

PPO's performance

1

Hi @bmazoure, Your PPO +JAX implementation caught my eyes and this is a really cool repo! Based on your [benchmark](https://wandb.ai/bmazoure/ppo_procgen_jax/reports/PPO-Procgen-JAX-version---VmlldzoxMDM4MjAx) with W&B, I compared the performance of your implementation with...

`libmem_filesys.so: cannot open shared object file`

17

Running into the issue of `libmem_filesys.so: cannot open shared object file`. I tried googling but could not find any info on this file Additionally, any chance the Preview 3 and...