This repository builds on top of SENSORIUM 2022 codebase. Clustering loss is effectivelly added in the standard_trainer function in sensorium/training/trainers.py, all the input arguements starting from include_kldivergence relate to DECEMber.

Example usage

standard_trainer(
    ...
    include_kldivergence=True, # turns on DECEMber
    cluster_number=10, # specifies number of clusters
    dec_starting_epoch=5, # defines the pretraining duration (in epochs)
    base_multiplier=4e3, # defines clustering strength \beta
    **kwargs,
)

Below we provide a step-by-step guide for getting started with the competition.

• install docker and docker-compose
• install git
• clone the repo via git clone https://github.com/sinzlab/sensorium.git

You can download the data from https://gin.g-node.org/cajal/Sensorium2022 and place it in sensorium/notebooks/data. Note: Downloading the files all at once as a directory does lead to unfortunate errors. Thus, all datastes have to be downloaded individually.

Start Jupyterlab environment

cd sensorium/
docker-compose run -d -p 10101:8888 jupyterlab

now, type in localhost:10101 in your favorite browser, and you are ready to go!

We provide notebooks that illustrate the structure of our data, our baselines models, and how to make a submission to the competition.
Dataset tutorial: Shows the structure of the data and how to turn it into a PyTorch DataLoader.
Model tutorial: How to train and evaluate our baseline models.
Submission tutorial: Use our API to make a submission to our competition.