TensorTrace is an available on Windows, MacOS and Ubuntu
(See GPL for copyright conditions)
In addition to the TensorTrace app each download comes with:
(i) An example project "binaryMERA.ttp" containing the networks needed to optimize a MERA tensor network, together with script files "script_binaryMERA" demonstrating how the contraction code generated from this project can be incorporated into a complete tensor network algorithm (with versions for MATLAB, Python and Julia). See the examples page for more details.
(ii) Copies of the general purpose "ncon" network contraction routine required by TensorTrace (with versions for MATLAB, Python and Julia)
Example project: "binaryMERA.ttp"
Current latest release: v1.03(beta)
Attention Windows users: due to issues with file permissions, TensorTrace is unable to save data in the same directory or a sub-directory of that which contains the "tensortrace.exe" app. In order to avoid this issue the default save folder "SaveData" is set in a different sub-directory to that containing the app executable. Tested on Windows10 Enterprise.
Attention macOS users: you may receive a "game not found" error when first trying to run the TensorTrace app. This is a known issue resulting from how the software used to compile TensorTrace interacts with macOS file permissions. This issue can be resolved by simply moving the TensorTrace app into the "AppData" sub-directory of the main folder after unzipping. The TensorTrace app has been notarized by Apple, so should still run on Gatekeeper enabled systems. Tested on macOS-v10.12.
Attention Ubuntu users: depending on the version of Ubuntu that you have additional dependencies may need to be installed for TensorTrace to work. The required packages are libcurl4 and libopenal1. Tested on Ubuntu 18.04.3 LTS.
The latest windows version can also be downloaded from itch.io (useful if the download below is blocked by you browser):
30/06/2020: release of TensorTrace v1.05 (beta), currently for Windows only.
addition of greedy algorithm solvers, allowing instantaneous determination of contraction orders.
network indices are now drawn with arrows to describe orientation, necessary for representing tensors that contain abelian symmetries.
individual networks can now be exported to code with ctrl-C copy.
inclusion of meta-data strings in exported code, allowing code to be ctrl-V pasted back into tensortrace.
Python versions of exported code are now also compatible with google's tensornetwork code library.
many small UI and quality-of-life improvements
24/09/2019: release of TensorTrace v1.03 (beta)
changes to file management system.
fixed small bug where tensors would sometimes be incorrectly labelled.
17/09/2019: release of TensorTrace v1.02 (beta)
small changes to make compatible with macOS and Ubuntu
11/09/2019: release of TensorTrace v1.01 (beta)
minor graphical and performance improvements
03/09/2019: release of TensorTrace v1.00 (beta)