To Do List

• Official release
  • Marketing on social media
  • Collect feedback from users

API

• Learning Rate Schedules
• Early Stopping
• Sharding support for distributed training
• Clean up rendering API

Features

• More Environments
  • Memory gym
• Multi-agent environments and algorithms
  • MADDPG, MAPPO
  • Pgx environments
• Masked Actions
  • Masked discrete action spaces
  • Masked continuous action spaces
• More complex space support
  • Composite spaces
  • Convex spaces
    • Ellipsoid
    • Polytope
• Robustness verification

Development

• Unit tests for all modules
• Integration tests for training loops
• Developer guide
• GitFlow branching model