https://github.com/sash-a/scalablehrles.jl

OpenAI's ES used in a feudal hrl style
https://github.com/sash-a/scalablehrles.jl

deep-neuroevolution evolution-strategies hierarchical-reinforcement-learning reinforcement-learning

Last synced: 14 days ago
JSON representation

OpenAI's ES used in a feudal hrl style

• Host: GitHub
• URL: https://github.com/sash-a/scalablehrles.jl
• Owner: sash-a
• Created: 2021-09-09T08:11:47.000Z (almost 4 years ago)
• Default Branch: master
• Last Pushed: 2022-01-24T08:49:57.000Z (over 3 years ago)
• Last Synced: 2025-06-04T04:08:48.635Z (24 days ago)
• Topics: deep-neuroevolution, evolution-strategies, hierarchical-reinforcement-learning, reinforcement-learning
• Language: Julia
• Homepage:
• Size: 1.51 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Scalable hierarchical evolution strategies

This is mostly Julia's wonderful multiple dispatch on top of my [ScalableES](https://github.com/sash-a/ScalableES.jl/) codebase. Most of the structs in here simply hold two of the similar type from ScalableES, one for the controller (c) and another for the primitive (p) for example:

```

mutable struct HrlAdam <: ScalableES.AbstractOptim

    copt::ScalableES.Adam

    popt::ScalableES.Adam

    HrlAdam(cdim::Int, pdim::Int, lr::Real) = new(ScalableES.Adam(cdim, lr), ScalableES.Adam(pdim, lr))

end

```

## How to run

For install instructions see [ScalableES](https://github.com/sash-a/ScalableES.jl/)

```

julia --project -t 8 scripts/runner.jl config/cfg.yml

```

## Examples

Envs that have been trained using this method:

### Ant Gather

![](vids/antgather.gif)

### Ant Maze

![](vids/antmaze.gif)

### Ant Push

![](vids/antpush.gif)