whatisgithub

What is iris?

breez3young/iris — explained in plain English

Analysis updated 2026-07-18 · repo last pushed 2023-10-10

PythonAudience · researcherComplexity · 4/5DormantSetup · hard

In one sentence

IRIS is an AI agent that learns to play video games by building its own imagined model of the game, letting it practice in imaginary futures instead of playing millions of real games.

Mindmap

mindmap
  root((repo))
    What it does
      Learns game rules
      Imagines futures
      Plays Atari games
    Tech stack
      Python
      Transformer
      Tokenizer
    Use cases
      Sample-efficient AI agents
      Research reproduction
      Watch agent play
    Audience
      Researchers
      ML engineers
    Validation
      Published ICLR 2023
      Pretrained checkpoints

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Train a game-playing AI agent efficiently using imagined scenarios instead of millions of real playthroughs.

USE CASE 2

Reproduce or build on the ICLR 2023 research results using released pretrained checkpoints.

USE CASE 3

Inspect what the agent 'sees' by watching it play or controlling the imagined game world yourself.

What is it built with?

PythonTransformerPyTorch

How does it compare?

breez3young/iris0xallam/my-recipe0xhassaan/nn-from-scratch
Stars——0
LanguagePythonPythonPython
Last pushed2023-10-102022-11-22—
MaintenanceDormantDormant—
Setup difficultyhardmoderatemoderate
Complexity4/52/54/5
Audienceresearchergeneraldeveloper

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · hard Time to first run · 1h+

Requires a GPU and Python ML environment, pretrained checkpoints are provided to skip training from scratch.

Copy-paste prompts

Prompt 1
Explain how IRIS's tokenizer converts game screen images into tokens a Transformer can predict.
Prompt 2
Show me how to load one of IRIS's pretrained checkpoints and watch the agent play Breakout.
Prompt 3
Help me understand how IRIS trains a world model to imagine future game states without running the real game.
Prompt 4
Walk me through adapting IRIS's world-model approach to a different Atari game.

Frequently asked questions

What is iris?

IRIS is an AI agent that learns to play video games by building its own imagined model of the game, letting it practice in imaginary futures instead of playing millions of real games.

What language is iris written in?

Mainly Python. The stack also includes Python, Transformer, PyTorch.

Is iris actively maintained?

Dormant — no commits in 2+ years (last push 2023-10-10).

How hard is iris to set up?

Setup difficulty is rated hard, with roughly 1h+ to a first successful run.

Who is iris for?

Mainly researcher.

Open on GitHub → Ask about another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.

What are similar repos?