BCI ALVI Challenge

Code repository to prepare the BCI Initiative + ALVI Labs Challenge

The utils folder contains code used to define and train the baseline model as well as to load the data and visualize predictions. The tutorials folder includes example code to load the data, train the baseline model and to submit predictions for the competition.

Please refere to tutorials/04_submit_predictions.ipynb for more details on how to prepare the submission file.

You can access the data on the MUW HPC at the below path (data is stored in the "dataset_v2_blocks/dataset_v2_blocks/"): /msc/home/vsharm64/projects/BCI_Kaggle

For installing mamba-ssm it required me to manually install nvcc onto MUW HPC

Add to your .bashrc the following:

# Adding cuda driver 12 based on Moritz’s fix
export LD_LIBRARY_PATH=/msc/home/mschae83/NVIDIA-Linux-x86_64-535.129.03/
# Adding nvcc
export PATH=/msc/home/vsharm64/cuda_VS/bin:$PATH

Then run

source ~/.bashrc

After entering your environment (i.e. ALVI) run:

pip install --no-build-isolation mamba-ssm