aisu-programming / Preprocessor-for-EEG-Signals

NYU CS-GY 9223 E Neuroinformatics (Spring 2024) - Final Project

0 stars 0 forks source link

bci bcic eeg eeg-classification eeg-signals eeg-signals-processing eegnet lstm optuna pytorch tensorflow torch transformer

readme

Preprocessor-for-EEG-Signals

Resource

Tutorial

Clone this project.

For Windows:

git clone --recurse-submodules https://github.com/aisu-programming/Preprocessor-for-EEG-Signals.git

For Mac: (Unclear, need to test)

git clone https://github.com/aisu-programming/Preprocessor-for-EEG-Signals.git
cd Preprocessor-for-EEG-Signals
git submodule init
git submodule update

Download the datasets below and put them in directory "datasets".
Rename the file "_.envsample" to ".env".

Dataset

Automatically download
```
./download_datasets.sh
```

Environment

With pip:

pip install -r requirements.txt
pip install torch  # Based on your environment, see more in the requirements.txt

Can also use conda.

[!TIP] When running MacOS on Apple Silicon, also install tensorflow-metal to utilize GPU usage:
pip install tensorflow-metal
You'll want to run with a higher batch size to get the benefits.

Example: Baselines

For baseline program in TensorFlow version - _baselinestensorflow.py
1. If you are working on a Mac platform, set __TF_USE_LEGACY_KERAS__ to True by:
```
export TF_USE_LEGACY_KERAS=True
```
2. Run the script:
```
python baselines_tensorflow.py
```
For baseline program in PyTorch version - _baselinespytorch.py
Simply run the script:
```
python baselines_pytorch.py
```

For running on HPC Greene
Adjust the command with slurm command prefix

srun -c 16 --mem=64GB --gres=gpu:1 --time=02:00:00 --pty python baselines_pytorch.py