Bring low pass filtering to PyTorch!
This pytorch extension offers a PyTorch alternative for Scipy's
lfilter - with gradient tracking.
Although it's certainly the goal to implement an efficient CUDA lfilter in C++, for now only the CPU version is implemented in C++. That said, the implementation is reasonably fast and doing the filtering on the CPU might be a viable option. Moreover, the pure-python implementation works on all devices.
The library can be installed with pip:
pip install torch_lfilter
Please note that no pre-built wheels exist. This means that pip will
attempt to install the library from source. Make sure you have the
necessary dependencies installed for your OS.
On Linux, having PyTorch installed is often enough to be able install the library (along with the typical developer tools for your distribution). Run the following inside a conda environment:
conda install pytorch -c pytorch
pip install torch_lfilter
On Windows, the installation process is a bit more involved as typically the build dependencies are not installed. To install those, download Visual Studio Community 2017 from here. During installation, go to Workloads and select the following workloads:
- Desktop development with C++
- Python development
Then go to Individual Components and select the following additional items:
- C++/CLI support
- VC++ 2015.3 v14.00 (v140) toolset for desktop
Then, download Microsoft Visual C++ Redistributable from here.
After these installation steps, run the following commands inside a x64 Native Tools Command Prompt for VS 2017, after activating your conda environment:
set DISTUTILS_USE_SDK=1
conda install pytorch -c pytorch
pip install torch_lfilter
© Floris Laporte 2020, GPLv3