Announcing AMD Schola v2.1: state trees, scale, and a richer training stack

AMD Schola v2.1 is released, bringing significant enhancements to the Unreal Engine reinforcement learning framework. Key additions include Unreal Engine StateTree integration for blending learned RL policies with designer-authored NPC logic, Kubernetes support for distributed training at scale, expanded Minari and imitation learning workflows, improved CLI with YAML configs and eval commands for Stable Baselines 3 and Ray RLlib, frame stacking utilities, hardened ONNX export, and async gRPC support in the Python SDK. The release also updates gRPC/Protobuf toolchain versions and modernizes packaging via pyproject.toml. Compatible with Unreal Engine 5.5–5.7 and Python 3.10–3.12, available under the MIT license.