Skip to content

pritamarnab/ECoG-foundation-model

 
 

Repository files navigation

ECoG foundation model

In progress - this repo is under active development in the MedARC discord server.

Run training on local machine

  1. Run "sh setup.sh" to create virtual environment

  2. Activate the virtual environment with "source ecog/bin/activate"

  3. Specify your huggingface user access token in the makefile to authenticate and run "make download-data" to fetch data from huggingface hub

  4. Specify model training parameters in makefile and run "make model-train"

Run training on Google Colab

If you don't have access to GPU's try uploading notebooks/ECoG_Model_Training.ipynb to Colab and use the free tier GPU's. This is a good way to play around with the training code and try training a model to get onboarded, although training is very slow on the T4 GPU's.

About

Self-supervised ECoG foundation model

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages

  • Python 91.9%
  • Jupyter Notebook 7.0%
  • Other 1.1%