File size: 2,699 Bytes
1dbe253 d30950c 1dbe253 d30950c 1dbe253 34ff700 1dbe253 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 | ---
library_name: hd-ppo
tags:
- Pendulum-v1
- deep-reinforcement-learning
- reinforcement-learning
- hyperdimensional-computing
- fractional-power-encoding
- LTU-AI
model-index:
- name: HD-PPO
results:
- task:
type: reinforcement-learning
name: reinforcement-learning
dataset:
name: Pendulum-v1
type: Pendulum-v1
metrics:
- type: mean_reward
value: -125.19 +/- 88.07
name: mean_reward
verified: false
---
# **HD-PPO** Agent playing **Pendulum-v1**
This is a trained **HD-PPO** (Hyperdimensional Proximal Policy Optimization) agent
playing **Pendulum-v1** using **gradient-adaptive Fractional Power Encoding (FPE)**
with a prune-and-fine-tune pipeline.
Published by [LTU-AI](https://huggingface.co/LTU-AI).
## Pipeline
1. Train a teacher at **D=512** with gradient-adaptive single-beta FPE.
2. Prune by actor-weight importance through **D=512 → 128 → 32**.
3. Fine-tune each pruned checkpoint with PPO.
Published checkpoint: seed **2024**, compact **D=32** model
(held-out eval mean reward **-125.19 ± 88.07**).
## Usage
Install dependencies:
```bash
pip install -r requirements.txt
```
Evaluate the local checkpoint:
```bash
python enjoy.py --weights hdppo-Pendulum-v1/weights.npz --episodes 10
```
Render episodes:
```bash
python enjoy.py --weights hdppo-Pendulum-v1/weights.npz --render --episodes 3
```
Record a replay video:
```bash
python record_video.py --weights hdppo-Pendulum-v1/weights.npz --output replay.mp4
```
Load from Hugging Face Hub:
```bash
python enjoy.py --weights LTU-AI/hdppo-Pendulum-v1 --episodes 10
```
## Training pipeline
Reproduce the teacher → prune → fine-tune workflow:
```bash
python run_prune_finetune_5seed.py
```
## Hyperparameters
```python
{
"env": "Pendulum-v1",
"algo": "HD-PPO (gradient-adaptive FPE, continuous)",
"teacher_D": 512,
"pruned_D": 32,
"beta_base": 2.5,
"timesteps_per_stage": 1000000,
"seed": 2024
}
```
## Environment Arguments
```python
{
"render_mode": "rgb_array"
}
```
## Model files
| File | Description |
|------|-------------|
| `hdppo-Pendulum-v1/weights.npz` | Published actor (+ critic if HD) and FPE encoder (D=32) |
| `hdppo-Pendulum-v1/weights_D512_teacher.npz` | Teacher checkpoint (D=512) |
| `replay.mp4` | Sample rollout video from the published min-D checkpoint |
| `results.json` | Evaluation summary for the published checkpoint |
| `results_D512_teacher.json` | Evaluation summary for the teacher |
| `config.yml` | Training hyperparameters |
| `train_hdppo.py` / training modules | Self-contained training code |
## Citation
If you use this model, please cite the HD-PPO / Hybrid-HD-PPO work.
|