Crafter is a thrilling open-world survival game designed specifically for evaluating various agent abilities within a single environment. In this guide, we’ll walk you through the steps to set up and play Crafter, along with some troubleshooting tips for smooth gameplay.

Overview of Crafter

Crafter presents randomly generated 2D worlds where players must forage for resources such as food and water, find shelter, defend against monsters, gather materials, and create tools. The game serves as a benchmark for reinforcement learning, offering insight into agent capabilities through:

• Research Challenges: It tests methods for generalization, representation learning, and long-term reasoning.
• Meaningful Evaluation: Agents are assessed based on significant achievements unlocked during gameplay.
• Iteration Speed: The game requires less computational power than other benchmarks, evaluating multiple agent abilities in one environment.

For a deeper understanding, refer to the research paper: Benchmarking the Spectrum of Agent Capabilities.

Getting Started

Follow these steps to get Crafter up and running:

• Step 1: Install Crafter by executing the following command in your terminal:

python3 -m pip install crafter

• Step 2: Install pygame, which is necessary for the human interface:

python3 -m pip install pygame

• Step 3: Start the game with this command:

python3 -m crafter.run_gui

Keyboard Mapping

The key controls in Crafter are as follows:

Key      Action
WASD     Move around
SPACE    Collect materials, drink from lake, hit creature
TAB      Sleep
T        Place a table
R        Place a rock
F        Place a furnace
P        Place a plant
1        Craft a wood pickaxe
2        Craft a stone pickaxe
3        Craft an iron pickaxe
4        Craft a wood sword
5        Craft a stone sword
6        Craft an iron sword

Understanding the Code

Here’s a sample code snippet for creating a Crafter environment:

import gym
import crafter

# Create a Crafter environment
env = gym.make("CrafterReward-v1")  # Or use "CrafterNoReward-v1"

# Recorder setup (optional)
env = crafter.Recorder(env, path_to_log_dir, save_stats=True, save_video=False, save_episode=False)

# Reset the environment
obs = env.reset()
done = False

# Main loop
while not done:
    action = env.action_space.sample()  # Choose a random action
    obs, reward, done, info = env.step(action)  # Perform the action

You can think of creating a Crafter environment like setting up a grocery store where the aisles represent different challenges (food, water, shelter). Each time you send a shopping cart (the agent) down an aisle, it randomly picks items (actions) to gather until it completes its shopping trip (the episode).

Performance Evaluation

Crafter evaluates agents based on a budget of 1 million environment steps, focusing on a range of achievements. Key evaluation criteria include:

• Success Rates: Percentage of achievements unlocked during training episodes, providing insight into the agent’s abilities.
• Score: Geometric mean of success rates, accentuating improvements on difficult achievements.

Troubleshooting

If you encounter issues while setting up or playing Crafter, consider the following troubleshooting tips:

• Ensure that your Python and pip installations are updated. Old versions might lead to compatibility issues.
• Check for any missing dependencies. If pygame fails to install, try reinstalling it.
• Run your commands with administrative privileges if you face permission errors.
• For custom installations or paths, double-check that you have the correct directory structure set up.
• For more insights, updates, or to collaborate on AI development projects, stay connected with fxis.ai.

At fxis.ai, we believe that such advancements are crucial for the future of AI, as they enable more comprehensive and effective solutions. Our team is continually exploring new methodologies to push the envelope in artificial intelligence, ensuring that our clients benefit from the latest technological innovations.