The Nengo OCL team is delighted to announce the release of Nengo OpenCL 1.1.
Nengo OpenCL is a backend for Nengo that uses the OpenCL framework to run large-scale neural models on CPUs, GPUs, and other OpenCL-capable platforms. Using the OpenCL backend can be orders of magnitude faster than the reference backend for large models.
To use Nengo OpenCL, replace instances of
For example, if you have a network called
and you run it with
with nengo.Simulator(model) as sim: sim.run(10)
you would change that to
with nengo_ocl.Simulator(model) as sim: sim.run(10)
and that’s it!
Nengo OpenCL 1.1.0 adds support
for several Nengo features,
Sigmoid neuron types,
and filtering matrices.
This version of Nengo OpenCL also improves compatibility with Nengo, NumPy, and Python 3. Nengo OpenCL works on Python 2.7 and 3.4+, with Nengo version 2.1.2 and later, including the newly released 2.3.0.
To see the full list of changes in Nengo OpenCL 1.1, head to the Github release page.
Installing Nengo OpenCL is more difficult than other parts of the Nengo ecosystem since OpenCL must be properly installed first. See our notes on installing OpenCL for assistance.
Once OpenCL is installed, use
pip install --upgrade nengo_ocl
If something goes wrong during the installation, which can happen due to differing OpenCL installs, refer to our installation notes for assistance.
Please come to the Nengo forum! We welcome any questions and suggestions you might have.