# eloialonso/diamond

DIAMOND (DIffusion As a Model Of eNvironment Dreams) is a reinforcement learning agent trained in a diffusion world model. NeurIPS 2024 Spotlight.

Repository: https://github.com/eloialonso/diamond
Canonical: https://ross.abutalabs.com/products/diamond
Homepage: https://diamond-wm.github.io
Language: Python
License: MIT
License Family: permissive
Topics: atari, deep-learning, diffusion-models, machine-learning, reinforcement-learning, research, world-models, artificial-intelligence
Last push: 2024-12-06T16:45:28+00:00

## Health v2 (maintenance only)
Score: 24/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 59
- inputs: {"age_days": 836, "days_push": 635, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2096, forks 166 (observed 2026-08-28T04:06:13.311335+00:00)

## What it is
DIAMOND is a Python implementation of a reinforcement learning agent trained entirely inside a diffusion-based world model, presented as a NeurIPS 2024 Spotlight paper. It includes pretrained playable world models for Atari games and a CS:GO world model that acts as an interactive neural game engine.

## Use cases
- train RL agents inside a learned diffusion world model
- play Atari games generated by a neural world model
- run an interactive neural game engine simulating CS:GO
- reproduce the DIAMOND Atari 100k benchmark results
- research diffusion models for environment dynamics modeling
- experiment with world model training pipelines in PyTorch

## When to choose
- you need sample-efficient RL training in an imagined environment
- you want to study or extend diffusion-based world models
- you want a playable neural game engine demo
- you need a strong baseline for world-model RL research

## When to avoid
- you need a production game engine or real-time commercial performance
- you lack a GPU or don't want heavy PyTorch training workloads
- you need a general-purpose RL library rather than a research codebase
- you cannot license Atari ROMs for your use

## Facets
- artifact type: library
- maturity: active
- function: reinforcement-learning, machine-learning, deep-learning, simulation, game
- domain: reinforcement-learning, artificial-intelligence, deep-learning, gaming-tools
- platform: python, cross-platform
- tags: diffusion-models, world-models, neural-game-engine, atari, csgo, research-code, neurips-2024, research, gpu

## Member repositories
- eloialonso/diamond (main) score 24

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:13.311335+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:54:35.951178+00:00, confidence not recorded.
  - readme: https://github.com/eloialonso/diamond (fetched 2026-08-28T04:06:13.311335+00:00, sha 5958ef8540c2)
  - homepage: https://diamond-wm.github.io (fetched 2026-08-29T10:34:59.038666+00:00, sha 4504603ab313)
- Data as of 2026-08-30T08:39:29.467469+00:00.