README.md

metadata

title: TalkTuner Probe Training
emoji: 🎯
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 5.0.0
app_file: app.py
pinned: false
python_version: 3.1

TalkTuner Probe Training Interface

This Space provides an interactive interface for training demographic probes on Large Language Models.

Features

• Train reading and controlling probes
• Support for multiple demographic attributes (age, gender, socioeconomic, education)
• Real-time training progress visualization
• Download trained models and results

Hardware Requirements

• CPU Basic: Testing and demonstration (free)
• T4 Small: Full training with GPU (~$0.60/hour)
• A10G: Faster training (~$1.05/hour)

Setup

1. Upload your dataset files to data/dataset/
2. Configure your HuggingFace token in Space settings
3. Select appropriate hardware tier
4. Launch the interface

Based on

"Designing a Dashboard for Transparency and Control of Conversational AI"