AI4Finance-Foundation/FinGPT
OPEN-SOURCE FINANCIAL LLMS FOR DATA ANALYSIS
FinGPT is a suite of open-source financial large language models from AI4Finance Foundation. The project offers lightweight fine-tuning (under $300) instead of full training costing millions, using RLHF and automated data pipelines.
// KEY FEATURES
- Fine-tuning for <$300 vs $3M full training
- RLHF for personalization to risk profile
- Automated internet-scale financial data pipeline
- Support for cloud APIs and local GPU models
// DEEP DIVE: HOW IT WORKS
FinGPT uses open base LLMs and applies instruction tuning on financial data. Released models for forecasting (FinGPT-Forecaster), sentiment analysis, and multi-task. Two modes: cloud API (OpenAI) without GPU or locally with GPU, transformers and peft.
Unlike BloombergGPT ($3M, 53 days), FinGPT focuses on lightweight adaptation: fine-tuning <$300, RLHF for personalization to risk profile (absent in BloombergGPT), automated data pipeline for weekly updates.
Local run needs GPU, install: pip install -r requirements.txt && pip install -e . Models on HuggingFace, demo on HF Space. Repo is Jupyter Notebook. Cloud mode requires OpenAI API key.
// USE CASES
Best for: For fintech developers and NLP researchers in finance
- Fine-tune model for your trading strategy in hours
- Run financial news sentiment analysis without GPU via API
- Build personalized robo-advisor accounting for risk profile
// QUICK START
git clone https://github.com/AI4Finance-Foundation/FinGPT.git
cd FinGPT
pip install -r requirements.txt
pip install -e .Commands are taken from the repository README — check the current version on GitHub before running.
New repositories every 30 minutes
REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.
Join on Telegram