Ls Imagine: Explore visual reinforcement learning

Understand the project

Practical overview

Use it when you want research code and task-specific checkpoints for studying visual reinforcement learning with learned world models in MineDojo.

Researchers reproducing visual reinforcement-learning experiments in MineDojo.
Developers exploring world models and imagination-based training for open-world agents.

Study how longer imagined experience can support exploration and long-horizon decisions.
Set up the research environment for LS-Imagine experiments.
Evaluate a supplied pretrained checkpoint on a documented task.

Before you choose

The public GitHub repository is licensed under Apache-2.0.

The setup depends on MineDojo and MineCLIP as well as the project’s Python requirements.
The README specifies Ubuntu 20.04, Python 3.9 and JDK 1.8.0_171 for its documented setup.
Pretrained weights are listed for tasks including harvesting, mining and shearing.
TensorBoard is available for recording training data; Weights & Biases is an optional alternative that requires an API key.

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 the documented environment and dependencies needed to proceed with LS-Imagine setup.

Create and activate a Conda environment with Python 3.9.
Install JDK 1.8.0_171, then install MineDojo and MineCLIP using their official documentation.
Install the project dependencies with pip install -r requirements.txt.
Download the MineCLIP weight and place it at ./weights/mineclip_attn.pth.

The documented environment and MineCLIP weight are in place for the subsequent training or evaluation instructions.

The README notes that MineDojo installation may encounter errors and refers readers to the project installation guide.
This guide covers setup only, not a training or evaluation run.

Evidence and transparency

Source record details

Source availability: Active
Sources checked: 15 October 2026
Original website title: qiwang067/LS-Imagine · Official homepage title · retained in its source language