2045max/finrl-ppo-dow30-quick

FinRL PPO Agent (Quick Demo, 2000 steps)

Trained on DOW 30 stocks (2014-2025) using FinRL + Stable-Baselines3 PPO.

⚠️ Toy model — only 2000 timesteps, used to validate training pipeline. Not for real trading.

Usage

from huggingface_hub import hf_hub_download
from stable_baselines3 import PPO

path = hf_hub_download(
    repo_id="2045max/finrl-ppo-dow30-quick",
    filename="agent_ppo.zip"
)
model = PPO.load(path)

Training Setup

• Algorithm: PPO (Proximal Policy Optimization)
• Total timesteps: 2,000
• State space: 301 (cash + 30 prices + 30 holdings + 30×8 indicators)
• Action space: 30 (continuous, [-1, 1] per stock)
• Reward: portfolio value change × 1e-4

Source

https://github.com/AI4Finance-Foundation/FinRL