OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Rainbow — Kaixhin

Rainbow is a Python implementation of a deep reinforcement-learning algorithm that combines DQN, Double DQN, prioritized experience replay, a dueling network architecture, multi-step returns, distributional reinforcement learning and Noisy Nets for Atari tasks.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Rainbow — Kaixhin — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Rainbow when an Atari reinforcement-learning experiment calls for a combined DQN-based method, its documented data-efficient configuration, or evaluation from the repository’s linked pretrained models.

Who it is for

  • Researchers studying or comparing deep reinforcement-learning methods.
  • Developers training agents on Atari games.
  • Experimenters evaluating Atari agents from pretrained models.

Documented uses

  • Train an Atari agent with the combined Rainbow method.
  • Run the documented data-efficient Rainbow configuration.
  • Compare the combined approach with its individual DQN improvements.
  • Use linked pretrained models as starting points for evaluation.
Before you choose

Requirements & limits

The public GitHub repository displays an MIT license.

  • The documented environment requires atari-py, OpenCV Python, Plotly and PyTorch.
  • The repository includes a data-efficient configuration with explicit training parameters.
  • Models from release 1.3 need a first-convolution padding adjustment because of a noted architecture issue.

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

Set up Rainbow’s documented Python environment

Create and activate the project’s Anaconda environment before using the implementation.

Before you start

  • Anaconda
  • The Rainbow repository, including environment.yml

Steps

  1. From the repository directory, run conda env create -f environment.yml.

  2. Activate the environment with source activate rainbow.

Expected outcome

The README’s named rainbow environment is created and activated with the project dependencies.

Guide limitations

  • These steps cover environment setup only; they do not verify a successful training run.
  • The README lists atari-py, OpenCV Python, Plotly and PyTorch as requirements.
Documentation 1 ↗
Source-based practical guide

Run the documented data-efficient Rainbow configuration

Start the data-efficient Atari training configuration using the repository’s specified parameters.

Before you start

  • The Rainbow Python environment and listed dependencies
  • An available Atari game from the atari-py ROMs

Steps

  1. Change to the Rainbow repository directory.

  2. Run the command below with --target-update 2000 --T-max 100000 --learn-start 1600 --memory-capacity 100000 --replay-frequency 1 --multi-step 20 --architecture data-efficient --hidden-size 256 --learning-rate 0.0001 --evaluation-interval 10000.

    Documented command
    python main.py

Expected outcome

The documented data-efficient Rainbow configuration is invoked with the listed training and evaluation parameters.

Guide limitations

  • The README explains that its “unbounded” memory is implemented by setting memory capacity to the maximum number of timesteps.
  • Pretrained models from release 1.3 require changing first-convolution padding from 0 to 1.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
Kaixhin/Rainbow · Official homepage title · retained in its source language
How we review sourcesSuggest a correction