/EvoCraft-interactive

Primary LanguagePython

Interactive evolution for EvoCraft

Evolve intearctively Minecraft entities encoded as neural networks

Disclamer: This is an alpha version, expect further changes.

Install

  1. Set up the python API to Minecraft

  2. Clone the repo: git clone https://github.com/claireaoi/EvoCraft-interactive

  3. Install dependencies: pip install -r requirements.txt

Start interactive evolution

After launching the Minecraft server use train.py to start interactive evolution.

They generated machines will be rendered on Minecraft and you'll have to pick with the keyboard numbers which one you like most, this will provide a reward signal for the selected genotype. Press 0 to abort evolution.
Evolution stops when the maximum generation number is reached. The networks' weights (ie. the genotypes) will be saved in the weights folder. The evolution is guided by an ES algorithm.

Use python train.py --help to display all the evolution options:


train.py [--generator] [--restricted] [--dimension] [--choice_batch] [--oriented] [--position] [--lr] [--decay] [--sigma] [--generations] [--population_size] [--top_k] [--folder]

  --generator         Generator/policy type: MLP, RNN, SymMLP
  --restricted        0 or 1. Tells if want to use only a restricted list of blocks (1), else 0.
  --dimension         2 or 3. To restrict the spatial dimension we work with.
  --choice_batch      Number of structures among which to choose one.
  --oriented          0 or 1. Indicate if shall incorporate the orientations in the encoding.
  --position          Initial position for player advised, around which the structures will be evolved.
  --lr                ES learning rate.
  --decay             ES and learning rate decay.
  --sigma             ES sigma: modulates the amount of noise used to populate each new generation, the higher the more the entities will vary
  --generations       Number of generations that the ES will run.
  --population_size   Size of population (needs to be pair and be a multiple of choice_batch or will be approximated).
  --top_k             Top-k sampling, for a stochastic generation of structures. For the deterministic case, choose k=1.
  --folder            folder to store the evolved weights