Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 

Repository files navigation

📊 TabFM + Ollama Analyzer

Zero-Shot Tabular Predictions & Local LLM Insights 🧠⚡

A visually streamlined, end-to-end tool blending TabFM (Tabular Foundation Model) and Ollama (qwen3:4b). Input data, get predictions instantly, and receive AI-driven summaries—no dataset training required. This repository acts as an interface and wrapper for Google's TabFM, ensuring fast tabular predictions running 100% locally.


🛠️ Tech Stack

  • 🧠 Core ML: TabFM (PyTorch)
  • 🤖 Local LLM: Ollama (qwen3:4b - zero temp/fast inference)
  • 🖥️ Frontend: Streamlit
  • 💾 Data: pandas & numpy

📂 Files

  • 📄 README.md: Project overview and setup.
  • 🐍 app.py: Streamlit UI, TabFM engine, and Ollama integration.
  • 📦 requirements.txt: Minimal Python dependencies.
  • 📝 outputs.md / ideal_outputs.md: Auto-generated, identically matched markdown logs of your predictions and AI insights.

🚀 Setup & Execution

Copy-paste these commands directly into Windows PowerShell:

1. Install Core Dependencies

git clone https://github.com/google-research/tabfm.git
cd tabfm
pip install -e .[pytorch]
pip install streamlit pandas numpy requests
ollama pull qwen3:4b

2. Run Application

streamlit run app.py

3. Quick Test (Headless)

python -c "from tabfm import tabfm_v1_0_0_pytorch; print('TabFM loaded successfully')"

🎯 5 Core Use Cases

  1. 🏦 Risk Assessment: Instantly predict credit risk based on mixed tabular data.
  2. 🏡 Real Estate Valuation: Zero-shot regression on property features for pricing.
  3. 📉 Customer Churn: Identify subscription cancellations from usage data.
  4. 🏥 Medical Triage: Classify patient risk levels from health records.
  5. 📦 Inventory Forecasting: Predict stock needs using historical inputs.

🔮 5 Future Upgrades

  1. 📁 Batch Upload: CSV support for bulk processing.
  2. 🎛️ Model Selector: UI dropdown for various Ollama models.
  3. 🖨️ PDF Export: Download AI insights instantly.
  4. 📈 Dynamic Charts: Plotly integration for visual metrics.
  5. JAX Support: Toggle PyTorch/JAX backends directly in UI.

SEO Keywords: Google TabFM, Tabular Foundation Model, Local LLM, Ollama, Qwen3, Streamlit AI interface, zero-shot tabular predictions, offline AI inference, Python tabular analysis, data science automation.

About

Visual interface for Google TabFM and Ollama for fast, local tabular data predictions

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages