Crane provides useful components to facilitate the development and training of machine learning models for neural data. It includes tools for data handling and common neural model components, making it easier to build and experiment with models in this domain. It is designed to be compatible with Hugging Face Transformers, PyTorch, and the Insight-Neuro Ecosystem, allowing researchers to share models and preprocessing pipelines seamlessly.
The recommended way to install Crane is via pip:
pip install "crane @ git+https://github.com/insight-neuro/crane"To get started with Crane, you can build your own model by subclassing BrainModel and BrainFeatureExtractor. These classes provide a structured way to define your model architecture and preprocessing steps.