OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Infrastructure & Serving

Megaverse

Megaverse is an open-source C++ rendering and simulation engine with Python bindings for AI research, including reinforcement-learning experiments. Its batched renderer is described as generating up to one million observations per second on a single machine.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Megaverse — OpenViST project profile in AI Infrastructure & Serving.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Megaverse when an AI research workflow needs interactive simulated tasks and high-volume rendered observations, especially for reinforcement-learning experiments.

Who it is for

  • AI researchers generating observations from simulated environments.
  • Reinforcement-learning developers working with tasks such as TowerBuilding.
  • C++ developers extending a simulation engine or its Python bindings.

Documented uses

  • Generate large batches of observations for simulation-based research.
  • Step through an interactive environment and collect observations, rewards and completion information.
  • Train and view agents in a documented Sample Factory reinforcement-learning workflow.
Before you choose

Requirements & limits

The public repository is MIT-licensed and documents source installation. No hosted service or account-based workflow is described.

  • Megaverse is installed from source; the README documents Conda dependencies and a CMake-based C++ core with Python bindings.
  • Linux instructions include Vulkan and CUDA dependencies for the high-throughput path.
  • macOS supports only a limited OpenGL build for local debugging and small-scale experiments; Vulkan-powered batched rendering is unsupported there.
  • The documented reinforcement-learning training workflow uses Sample Factory.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Megaverse from source on macOS

Set up the documented limited OpenGL build for local debugging or small-scale experiments.

Before you start

  • macOS
  • Conda
  • Python 3.9

Steps

  1. Clone the Megaverse repository and initialize its submodules recursively.

  2. Create and activate a Conda environment with Python 3.9.

  3. Install the documented Conda dependencies: OpenCV 4.4.x, CMake 3.13 or later, and Bullet.

  4. From the repository, run the command below and the command below.

    Documented command
    python setup.py develop
    Documented command
    pip install -e .

Expected outcome

Megaverse is installed in the Conda environment for the documented limited OpenGL use.

Guide limitations

  • Vulkan-powered batched rendering is not supported on macOS.
  • The README characterizes this build as suitable for local debugging and small-scale experiments.
Documentation 1 ↗
Source-based practical guide

Train and view a TowerBuilding agent

Use the README's Sample Factory workflow to train an agent in TowerBuilding and launch its behavior viewer.

Before you start

  • Megaverse installed from source
  • Sample Factory 2.0 or later
  • A configured environment supporting the selected Megaverse rendering mode

Steps

  1. Install Sample Factory with the command below.

    Documented command
    pip install "sample-factory>=2.0"

  2. Run the README's megaverse_rl.train_megaverse command with --env=TowerBuilding and its documented APPO settings.

  3. Launch the README's megaverse_rl.enjoy command with --algo=APPO --env=TowerBuilding --experiment=TowerBuilding to observe agent behavior.

Expected outcome

The documented workflow trains an APPO agent in TowerBuilding and provides a command to view its behavior.

Guide limitations

  • The training example is configured for a very large run, including up to 2,000,000,000 environment steps; adjust only with reference to the project's training documentation.
  • The README's sample commands use Vulkan rendering, which it says is unsupported on macOS.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
alex-petrenko/megaverse · Official homepage title · retained in its source language
How we review sourcesSuggest a correction