| # Huggy - Trained Agent | |
| **Author:** Vishand03 | |
| **Model Type:** Reinforcement Learning (PPO) | |
| **Environment:** Custom Huggy Environment (ML-Agents) | |
| **Framework:** ML-Agents + PyTorch | |
| --- | |
| ## Description | |
| This model is a trained Huggy agent using the PPO algorithm. | |
| It learns to navigate and complete tasks in the Huggy environment. | |
| --- | |
| ## Training Details | |
| - **Trainer:** PPO | |
| - **Steps:** ~800,000 (can be resumed) | |
| - **Reward:** ~3.9 mean reward at the last checkpoint | |
| - **Hyperparameters:** | |
| - Batch size: 4096 | |
| - Buffer size: 40960 | |
| - Learning rate: 0.0001 | |
| - Gamma: 0.995 | |
| - Lambda: 0.95 | |
| --- | |
| ## Usage | |
| ```python | |
| from mlagents_envs.environment import UnityEnvironment | |
| from mlagents_envs.base_env import ActionTuple | |
| import onnxruntime as ort | |
| env = UnityEnvironment(file_name="Huggy.x86_64", no_graphics=True) | |
| # Load model | |
| session = ort.InferenceSession("Huggy-799913.onnx") | |
| # Continue with your inference pipeline... | |