MtM · IBL BrainWideBench

Region-aware masked autoencoder, mixing neuron, causal, inter-region and intra-region masking, pretrained on 423 IBL Neuropixels sessions. Backbone for behavior decoding (TS1), activity prediction (TS2) and, after per-recording calibration, brain-region decoding (TS3). 10.8M parameters in the encoder, plus 59.0M of per-session stitchers.

hf download nerdslab/ibl-bwb-mtm mtm.pt --local-dir ckpt

python src/ts1/train.py trainer=mtm_finetune ckpt.load_from=ckpt/mtm.pt \
    recording_id=<EID> task=choice data_root=<PATH>

Loading a checkpoint with torch.load needs weights_only=False, since it carries its run config alongside the weights.

Paper · Benchmark · Docs · Leaderboard · Training run

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Collection including nerdslab/ibl-bwb-mtm

Paper for nerdslab/ibl-bwb-mtm