Model Card for REVE-small

REVE (project page here) is a transformer-based foundation model for EEG signal processing, designed to be adaptable to any electrode configuration and a wide range of EEG-based tasks. REVE-small is a 4-layer version of REVE-base, with the same layer design.

Model Details

Architecture

REVE (Representation for EEG with Versatile Embeddings), a pretrained encoder explicitly designed to generalize across diverse EEG signals. REVE introduces a novel 4D positional encoding scheme that enables it to process signals of arbitrary length and electrode arrangement.

	REVE-small	REVE-base
Transformer layers	4	22
Width / heads / head dim	512 / 8 / 64	512 / 8 / 64
Feed-forward	GEGLU, ratio 2.66	GEGLU, ratio 2.66
Parameters	12.7M	69.2M

It uses the same patching (200 samples with a 20-sample overlap, at 200 Hz) and the same 4D positional encoding as REVE-base, so it works with the REVE position bank.

Training

REVE-small was pretrained with the REVE masked-autoencoding objective. This checkpoint was saved at epoch 15 of a 200-epoch schedule, so it is an early checkpoint, not a fully trained model. Only the encoder is released; the MAE decoder and the optimizer state are not.

Developed by the BRAIN team and UdeM

Funded by: This research was supported by the French National Research Agency (ANR) through its AI@IMT program and grant ANR-24-CE23-7365, as well as by a grant from the Brittany region. Further support was provided by a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada (NSERC), by funding from the Canada Research Chairs program and the Fonds de recherche du Québec – Nature et technologies (FRQ-NT). This work was granted access to the HPC resources of IDRIS under the allocation 2024-AD011015237R1 made by GENCI, as well as HPC provided by Digital Alliance Canada.

Model Sources

• Repository: github
• Paper : arxiv

Uses

Example script to extract embeddings with REVE, using our position bank:

from transformers import AutoModel

pos_bank = AutoModel.from_pretrained("brain-bzh/reve-positions", trust_remote_code=True)
model = AutoModel.from_pretrained("brain-bzh/reve-small", trust_remote_code=True)

eeg_data = ... # EEG data as a torch Tensor (batch_size, channels, time_points), must be sampled at 200 Hz

electrode_names = [...] # List of electrode names corresponding to the channels in eeg_data
positions = pos_bank(electrode_names) # Get positions (channels, 3)
# Expand the positions vector to match the batch size 
positions = positions.expand(eeg_data.size(0), -1, -1)  # (batch_size, channels, 3)

output = model(eeg_data, positions)

License and Responsible Use

REVE is available for research, commercial, educational, and personal use under the REVE Responsible Use License v1.0.

By downloading, using, modifying, redistributing, or deploying REVE or a derivative model, you agree to comply with the terms of the LICENSE.

In particular, the License prohibits privacy intrusion and re-identification, non-consensual surveillance or profiling, discriminatory or harmful uses, and other uses that violate applicable law. Redistribution and publication of fine-tuned or otherwise modified REVE models are permitted, provided that the License, attribution, and provenance requirements are preserved.