Back to News

### TL;DR

NeuralSet is a high-performance Python package for Neuro-AI research by Meta FAIR. It unifies processing of diverse neural recordings (fMRI, EEG, MEG, iEEG, spikes) into PyTorch-ready datasets, with built-in HuggingFace embedding extractors, making it dramatically easier to run brain encoding and decoding experiments at scale.

Key Insights & Metrics

Key Features

• Unified pipeline for fMRI, EEG, MEG, iEEG, spikes — one framework handles all neural recording modalities with consistent event-based data structures
• HuggingFace-native embedding extractors — plug in any HuggingFace model for text, audio, image, or video features directly in the brain data pipeline
• Scalable cluster execution and caching — designed for large-scale Neuro-AI experiments with deterministic splits and experiment reproducibility

Discussion