Tokenizing/Running the Poyo model with regular time series data

[omerbe]

Hi,
I have some regular time series neural data, and I am hoping to use it to train a poyo model as described in your paper. However, in the poyo.py file, it only seems to support irregular time series data. Would you be able to give me some guidance as to how I should go about integrating regular time series data into the poyo model?

[linear]

PRO-299 Tokenizing/Running the Poyo model with regular time series data

[mazabou]

Hello @omerbe, we will add support to RegularTimeSeries soon as part of POYO+.
RegularTimeSeries can be easily converted to an IrregularTimeSeries very easily using this method: RegularTimeSeries.to_irregular(). I suggest that you use it temporarily, the results will be identical!

[holmosaint]

Hi, thank you for maintaining this repository. Now that Azabou et al., 2025 has been out for over four months, could you share an anticipated timeline for adding RegularTimeSeries data, such as calcium-imaging recordings, to POYO+? Having this feature in the official repo would greatly simplify benchmarking and improve reproducibility. Thanks!

[vinamarora8]

Hi @holmosaint, POYO+ for calcium-imaging data has recently been merged into this repository!

Model definition: https://github.com/neuro-galaxy/torch_brain/blob/main/torch_brain/models/capoyo.py
Example training setup: https://github.com/neuro-galaxy/torch_brain/tree/main/examples/poyo_plus