Introduction - Neuroencoder

pip install neuroencoder

Neuroencoder serves the EPI-Embedding MRL model — a Matryoshka-distilled version of EPI-250k, our EEG foundation model trained on 250,000 hours of clinical EEG. The distilled model produces a 768-dimensional embedding that you can truncate to any of 768, 384, 192, 48, 16 dimensions in a single forward pass.

import neuroencoder as ne
from neuroencoder import MRL

model = MRL.from_pretrained()
embeddings = model.embed(eeg, sfreq=256, channel_names=ch_names, dim=192)

ne.explore(embeddings)

Embeddings

Use the MRL model end-to-end

Visualization

Interactive Apple Embedding Atlas

Quickstart