Torch5 provides a Matlab-like environment for state-of-the-art machine
learning algorithms. It is easy to use and provides a very efficient
implementation, thanks to an easy and fast scripting language
(Lua) and a underlying C implementation.
Torch5 is the official successor of Torch3, and is now developed
at NEC Laboratories America and Google Labs.
Torch5 is distributed under a BSD license. It
makes now heavy use of the QT cross-platform framework,
distributed under the GPL license with
an exception when linked against BSD free software.
You may want to buy a commercial version of QT if it better suits your needs.
Torch5 runs on Linux, MacOS X and native Windows.
It is hosted on Sourceforge.
See the installation manual for download and installation details.
If you use Torch5 on a regular basis, it is recommended to subscribe to the
devel mailing list,
to be aware of bug corrections, possible API changes, or new features.
Tensor and Storage can now address more than 2G of RAM (on 64 bits systems),
as we converted the size type from int to long. This should have few impacts
on the code. Previously saved models should still load, as a versioning procedure has been
introduced in the class system.
Torch5 website is now up. The official version is now Torch 5.1. Please
be careful: the CVS location changed.
Torch 5.0 available on SourceForge CVS.
Torch5 version available on SourceForge CVS.
Torch5 created on SourceForge.