PPO on SnowballTarget
A Unity ML-Agents 1.0.0 policy trained from scratch using seed 42 and the included configuration. Training completed 200,112 environment steps. The final logged training window at 200,000 steps had mean reward 25.386 and standard deviation 2.047. These are training statistics, not a held-out evaluation.
The final policy is SnowballTarget.onnx. Checkpoints, TensorBoard events, configuration and the training log are included. The course requires a trained model for this environment and sets no minimum reward.
To watch the agent, open https://huggingface.co/spaces/unity/ML-Agents-SnowballTarget and select this repository and SnowballTarget.onnx.
Course: https://huggingface.co/learn/deep-rl-course/en/unit5/hands-on
- Downloads last month
- 78