/sparse-timeseries-transformers

self-supervised Transformer Encoder for sparse time series

Primary LanguagePython

Transformer Encoder for Sparse time series

Referring to https://openreview.net/forum?id=HUCgU5EQluN

Requirements

  • pytorch
  • pytorch lightning

Getting started

Currently supported datasets:

  • Physionet2012 (Has to first downloaded by hand)

The configs are maintained in config directory.

For training, run

DEVICE=cuda python -m src.script.<training_step> <config_name>

where <training_step> is either pretrain or finetune.

Examle:

DEVICE=cuda python -m src.script.pretrain pretrain_physionet_lin_warmup1H