Skyline works with GPU-based neural networks that are implemented in PyTorch. To run Skyline, you need:
- A system equipped with an NVIDIA GPU
- PyTorch 1.1.0+
- Python 3.6+
Note that Skyline only supports profiling models that are trained using a GPU.
Skyline consists of two components: (i) the Skyline profiler, and (ii) a plugin for Atom. The Skyline profiler has only been tested on Ubuntu 18.04, but should also work on other versions of Ubuntu that have Python 3.6+. The Skyline plugin has been tested on both Ubuntu 18.04 and macOS Mojave (10.14) with the latest version of Atom.
For interactive profiling (i.e. Skyline inside Atom), you need both components. If you only plan to use Skyline for Standalone Profiling, you only need the Skyline profiler component.
The Skyline profiler can be installed using
pip. In your shell, run:
As with most Python packages, we recommend installing Skyline inside a
virtualenv. After installing the Skyline profiler, you will be able to
invoke it by running
skyline in your shell.
Skyline Atom Plugin
The Skyline Atom plugin can be installed using
apm (the Atom package
manager), which should be automatically installed after you install
Atom. In your shell, run:
You can alternatively install Skyline by using Atom's preferences pane and
skyline under the Install section.
After installing the Skyline plugin, the
Skyline:Toggle command should be
available in your command palette. A Skyline sub-menu should also appear
under the "Packages" menu.