---
license: mit
tags:
  - reinforcement-learning
  - trading
  - stable-baselines3
---
# Trading Agent RL

Kaggle-trained PPO, SAC, and A2C ensemble for AAPL. The Hugging Face Space at https://huggingface.co/spaces/rezvan98/trading-dashboard reads these artifacts and runs CPU inference for live model signals.

## Latest Kaggle Run

- Timesteps per agent: 30,000
- Device: CPU
- Train period: 2018-01-01 to 2023-12-31
- Test period: 2024-01-01 to 2025-12-31

| Metric | Value |
|---|---:|
| Total Return | 2.74% |
| Sharpe | -1.219 |
| Max Drawdown | -2.03% |
| SPY Benchmark Return | 48.95% |

## Artifacts

- `models/ppo_final.zip`
- `models/sac_final.zip`
- `models/a2c_final.zip`
- `models/ensemble_weights.json`
- `data/feature_schema.json`
- `data/backtest_results.csv`
- `data/trade_log.csv`
- `report/metrics.json`

Research and paper trading only. Not financial advice.