Installation

Branch And Dataset

Use the branch that matches your workflow and dataset:

Branch	Use case	Dataset
main	Current FINSABER-2 package workflow and active development	https://huggingface.co/datasets/finsaber-team/FINSABER-V2-Data
reproduce	Reproduce FINSABER-1 experimental results with the v1.0 framework	https://huggingface.co/datasets/finsaber-team/FINSABER-reproduce

Package Install

pip install finsaber

The package installs the reusable backtesting framework only.

Local Development

git clone https://github.com/waylonli/FINSABER
cd FINSABER
conda create -n trading python=3.10 -y
conda activate trading
python -m pip install -U pip setuptools wheel
pip install -e ".[dev,research]"

The research extra includes optional dependencies for baseline research strategies, such as statsmodels, xgboost, datasets, and pandas-datareader.

The documentation dependencies are separate:

pip install -e ".[docs]"

For full local development, install all extras:

pip install -e ".[dev,research,docs]"

Build A Wheel

python -m build --wheel

The generated wheel should include only backtest* packages. It should not include examples/, llm_traders/, rl_traders/, tmp/, build/, or dist/.

Environment Variables

LLM integrations in the repository may require:

OPENAI_API_KEY=...
HF_ACCESS_TOKEN=...

The core backtesting package does not require API keys.