DerivSAC

Trading NVIDIA Derivatives using SAC

12 JULY 2025: Project EOL. Moving on to pastures new.

DerivSAC/
├── config/
│   ├── config.yaml   # Configuration parameters
├── docs/
├── hyperparameters/
│   ├── sac.yml
├── notebooks/
├── tests/
│   ├── __init__.py
│   ├── alpha_test.py
│   ├── conftest.py
│   ├── test_engine.py
│   ├── test_simple.py
├── trading_bot/
│   ├── analysis/
│   │   ├── __init__.py
│   │   ├── technical.py   # Technical analysis
│   ├── core/
│   │   ├── bot.py
│   │   ├── sac_bot.py
│   ├── data/
│   │   ├── local_data/
│   │   │   ├── NVDA_1d_max.pkl
│   │   ├── __init__.py
│   │   ├── data_pipeline.py   # Data fetching and processing
│   │   ├── indicators.py   # Technical indicators
│   │   ├── sentiment_pipeline.py
│   ├── sac/
│   │   ├── zoo/
│   │   │   ├── __init__.py
│   │   │   ├── gradient_clipping_callback.py
│   │   │   ├── hyperparams_opt.py
│   │   │   ├── rl_zoo_env.py
│   │   │   ├── trading_zoo_callback.py
│   │   │   ├── train.py
│   │   │   ├── utils.py
│   │   ├── __init__.py
│   │   ├── test_model.py
│   ├── strategies/
│   │   ├── macd.py
│   ├── trade_env/
│   │   ├── __init__.py
│   │   ├── engine.py
│   │   ├── metrics.py
│   │   ├── nvidia.py
│   │   ├── position.py
│   ├── utils/
│   │   ├── __init__.py
│   │   ├── config.py
│   │   ├── logger.py   # Logging utilities
│   │   ├── output.py
│   ├── __init__.py
│   ├── main.py   # Main entry point
├── trading_bot.egg-info/
│   ├── PKG-INFO
│   ├── SOURCES.txt
│   ├── dependency_links.txt
│   ├── requires.txt
│   ├── top_level.txt
├── utils/
│   ├── README.md
│   ├── __init__.py
│   ├── generate_directory_structure.py
│   ├── update_readme_directory.py
├── .gitignore
├── README.md
├── backtest_results.png
├── pyproject.toml
├── requirements.txt
├── setup.cfg
├── setup.py
├── strategy_comparison.png

Quick start

1. Create and activate virtual environment:

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\Activate.ps1

2. Install dependencies:

pip install -r requirements.txt

NOTE: If the dependencies did not install CUDA toolkit for you, go online to download it. The version used here is CUDA 12.4. This is required to train the model using GPU.

3. Run main.py for current analysis

python trading_bot\main.py

Training

1. Run this command to begin RL-Baseline3 Zoo environment wrapped to ours.

python -m trading_bot.sac.zoo.train --algo sac --env NVIDIATrading-v0 --device cuda --conf-file .\hyperparameters\sac.yml

Training with hyperparameters optimizer

python -m trading_bot.sac.zoo.train --algo sac --env NVIDIATrading-v0 --device cuda --optimize --n-trials 50 --max-total-trials 50 --n-jobs 4 --sampler tpe --pruner median --n-evaluations 5 --eval-episodes 10 --conf-file .\hyperparameters\sac.yml --optimization-log-path logs/optimizer_results --tensorboard-log logs/tensorboard

For Amazon Spot instances:

For g4dn.xlarge (1 T4 GPU, 16GB GPU memory)

python -m trading_bot.sac.zoo.train --algo sac --env NVIDIATrading-v0 --device cuda --optimize --n-trials 100 --n-jobs 6 --sampler tpe --pruner median --conf-file ./hyperparameters/sac.yml --optimization-log-path /opt/ml/model/optimizer_results --tensorboard-log /opt/ml/model/tensorboard --n-evaluations 5 --eval-episodes 5 --storage sqlite:////opt/ml/model/optuna.db

For p3.2xlarge (1 V100 GPU, 16GB GPU memory)

python -m trading_bot.sac.zoo.train --algo sac --env NVIDIATrading-v0 --device cuda --optimize --n-trials 150 --n-jobs 8 --sampler tpe --pruner median --conf-file ./hyperparameters/sac.yml --optimization-log-path /opt/ml/model/optimizer_results --tensorboard-log /opt/ml/model/tensorboard --n-evaluations 5 --eval-episodes 8 --storage sqlite:////opt/ml/model/optuna.db

For p3.8xlarge (4 V100 GPUs, 64GB total GPU memory)

python -m trading_bot.sac.zoo.train --algo sac --env NVIDIATrading-v0 --device cuda --optimize --n-trials 200 --n-jobs 24 --sampler tpe --pruner median --conf-file ./hyperparameters/sac.yml --optimization-log-path /opt/ml/model/optimizer_results --tensorboard-log /opt/ml/model/tensorboard --n-evaluations 5 --eval-episodes 10 --storage sqlite:////opt/ml/model/optuna.db

Unit-Testing

To begin unit testing, run the following command:

pytest

Custom configurations

Running main one time would generate a default config file. From there, you can edit and change certain parameters to your liking.