NM512/dreamerv3-torch
Implementation of Dreamer v3 in pytorch.
This project offers a PyTorch implementation of the DreamerV3 algorithm, designed to train AI agents that can master diverse virtual environments. It takes in observational data (like images or states) from simulated worlds such as DeepMind Control Suite, Atari games, or Minecraft, and outputs trained agents capable of performing complex tasks within these environments. This is for AI researchers and reinforcement learning practitioners who are experimenting with advanced model-based reinforcement learning techniques.
813 stars.
Use this if you are a researcher specifically looking for an implementation of the original DreamerV3 paper's algorithm in PyTorch for controlled experimental setups.
Not ideal if you need the latest, most optimized, or fastest DreamerV3 baseline, as this implementation is outdated and a newer, faster version is available elsewhere.
Stars
813
Forks
207
Language
Python
License
MIT
Category
Last pushed
Mar 08, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/NM512/dreamerv3-torch"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Related frameworks
galilai-group/stable-worldmodel
Reliable, minimal and scalable library for evaluating and conducting world model research
danijar/dreamerv2
Mastering Atari with Discrete World Models
danijar/dreamerv3
Mastering Diverse Domains through World Models
leofan90/Awesome-World-Models
A comprehensive list of papers for the definition of World Models and using World Models for...
danijar/dreamer
Dream to Control: Learning Behaviors by Latent Imagination