Skip to content

Installation

spikeDE supports Python 3.9 through 3.13.

Use a virtual environment

We strongly recommend installing spikeDE in an isolated virtual environment to avoid conflicts with system-wide packages. You can create one using venv, Conda, or Docker.

Install Dependencies

spikeDE relies on several third-party libraries. Please follow the instructions below to install them.

Install PyTorch

The primary dependency is PyTorch. Since installation commands vary by platform and hardware, please refer to the official PyTorch installation guide for the most accurate instructions.

GPU Acceleration

To leverage GPU acceleration with spikeDE, ensure you install the CUDA-enabled version of PyTorch that matches your NVIDIA driver. The CPU-only version will work but will not utilize GPU resources.

Additional Dependencies

spikeDE utilizes neural differential equation solvers requiring the following specific packages:

  • torchdiffeq (for Ordinary Differential Equations - ODEs): 

    pip install torchdiffeq

  • torchfde (for Fractional Differential Equations - FDEs): 

    pip install git+https://github.com/kangqiyu/torchfde.git

Install spikeDE

Once the dependencies are ready, install spikeDE directly from the source repository:

pip install git+https://github.com/PhysAGI/spikeDE.git

GPU Support Enabled

spikeDE automatically detects and utilizes GPU acceleration if a CUDA-enabled PyTorch installation and a compatible NVIDIA GPU are present. No additional configuration is required.