Variational tensor train (or MPS) optimized using zero-site DMRG. Both the Lanczos and Jacobi-Davidson corrections are implemented to avoid local minima. Intelligent tensor contractions are supported:
t2("li")=t1("ijk") * t1("ljk");
together with the MPO automatic construction as in:
h += Fermi(i,L,true) * Fermi(j,L,false) * t(i,j);
h += Fermi(i,L,true) * Fermi(i,L,false) * Fermi(i+1,L,true) * Fermi(i+1,L,false);
int nsweep=8, m=200;
TypicalRunDMRG0(h,nsweep,m);
- lapack, blas and lapacke
- python-numpy for the python library
The other dependencies are automatically downloaded by cmake. They are: armadillo, carma, eigen3, Spectra, Catch2 and pybind11.
mkdir build && cd build
cmake .. -D CMAKE_BUILD_TYPE=Release
make -j4
Several c++ examples are provided at examples/ folder. For python users, there is an example in notebook/