Roadmap for version 3; new branch announce
We are planning for uni10 v3 and there are big changes on the way. We will use a modified version of Cytnx developed by @kaihsinwu (original Cytnx repo: https://github.com/kaihsin/Cytnx ) to handle low level tensor operations. Cytnx provides a framework that one can go between C++ and Python with ease. In short, all the objects implemented in C++ will be wrapped in either C++ or Python, thus provides a uniform API interface for both C++ and Python. Cytnx provides a PyTorch like API which also allows easy CPU/GPU integration. Lazy evaluation of permutation/reshape is implemented to gain efficiency.
Uni10 will focus on higher level abstraction, such as
UniTensor with symmetries. This decoupling should allow for future integration with other types of container library.
A new branch
v3_cytnx is created for the development of uni10 v3.
Cytnx is now a submodule under
uni10. If you already cloned the repo before, after you pull the updates from gitlab from the
you need to issue
git submodule update --init --recursive
in order to populate the