tanlin2013 / tnpy Goto Github PK

View Code? Open in Web Editor NEW

3.0 1.0 1.0 10.95 MB

Tensor Network algorithms implemented in python.

Home Page: https://tanlin2013.github.io/tnpy/

License: MIT License

Python 98.77% Dockerfile 0.49% Makefile 0.18% Shell 0.56%

tensornetwork quantum-mechanics quantum-algorithms quantum-computing quantum-machine-learning

tnpy's Introduction

tnpy: Tensor Network algorithms implemented in python.

Documentation |

This project is a python implementation of Tensor Network, a numerical approach to quantum many-body systems.

tnpy is built on top of quimb, along with TensorNetwork for tensor contractions, with optimized support for various backend engines (TensorFlow, JAX, PyTorch, and Numpy). For eigen-solver we adopt primme, an iterative multi-method solver with preconditioning.

Currently, we support Matrix Product State (MPS) algorithms, with more are coming...

Exact Diagonalization (ED)
Finite-sized Density Matrix Renormalization Group (fDMRG)
Tree tensor Strong Disorder Renormalization Group (tSDRG)

fDMRG & tSDRG are on alpha-release. For others, please expect edge cases.

Requirements

Dependencies are listed in pyproject.toml, and they are supposed to be installed together with tnpy. Here we just list the essential building blocks.

Also, it's required to have lapack and blas installed in prior to Primme. They can also be installed through pip with mkl-devel.

Installation

Using Docker
```
docker run --rm -it tanlin2013/tnpy
```

Using pip

Latest release:
```
pip install tnpy
```

Development version:

pip install git+https://github.com/tanlin2013/tnpy@main

Optional dependencies
- If lapack and blas are missing
```
pip install tnpy[mkl]
```
- For quimb drawing functionality. This will install matplotlib and networkx
```
pip install tnpy[drawing]
```

Getting started

We provide built-in models. Though it's also possible to register your own one.

import numpy as np
from tnpy.finite_dmrg import FiniteDMRG
from tnpy.model import XXZ

model = XXZ(n=100, delta=0.5)
fdmrg = FiniteDMRG(
    mpo=model.mpo,
    chi=60  # virtual bond dimensions
)
fdmrg.update(tol=1e-8)

Compute any physical quantities whatever you want from the obtained state. The resulting MPS is of the type quimb.tensor.MatrixProductState, see here for more details.
```
my_mps = fdmrg.mps
```

License

tnpy's People

Contributors

Stargazers

Watchers

Forkers

10sornetwork

tnpy's Issues

Build: update dependencies

Some dependencies are already expired.

Blockers

quimb 1.3.0 Quantum information and many-body library.
├── cytoolz >=0.8.0
│   └── toolz >=0.8.0 
├── numba >=0.39
│   ├── llvmlite >=0.38.0rc1,<0.39 
│   └── numpy >=1.18,<1.22 
├── numpy >=1.17
├── psutil >=4.3.1
├── scipy >=1.0.0
│   └── numpy >=1.17.3,<1.25.0 
└── tqdm >=4
    └── colorama *

TODO

May potentially have conflicts

wait for poetry 1.2.0 final release
update numba to >=0.54
update quimb to 1.4.0

Feature: Schmidt gap

Characterizing the MBLT

The main quantity we compute, new in the context of MBL, is the Schmidt gap. For two chain halves (or as close to for odd L), A and B, as shown in Fig. 1(a), an eigenvector’s reduced density matrix is ρA,k = TrB(|Ek〉〈Ek|) for a particular sample of the random fields. The disorder-averaged Schmidt gap is then defined as ∆ = 〈λk1 − λk2〉k, where λk1, λk2 refer to the largest eigenvalues of the reduced density matrix ρA,k, 〈·〉k denotes average over eigenstates and · denotes the average over many samples. The Schmidt gap has previously been shown to act as an order parameter for quantum phase transitions [34, 36]. We explore the possibility of using it for characterizing the MBLT. Unlike entanglement entropy, the Schmidt gap ignores most of the spectrum of ρA,k, describing only the relationship between the two dominant states across the A − B cut. This is pertinent in light of the recent finding that while the Schmidt values decay polynomially in the MBL phase [22], finite size corrections are stronger for small Schmidt values. In the ergodic phase we expect strong entanglement to produce multiple, equally likely orthogonal states, thus ∆ ∼ 0. In the MBL phase, however, a single dominant state should appear on either side of the cut, with ∆ rising towards 1 as h → ∞, implying a tensor product. This behaviour is shown in Fig. 2(a) and becomes becomes sharper with increasing L. To see this more vividly, we plot the derivative of ∆ with respect to h in Fig. 2(b). The derivative has a peak at h = ̃ hc, which not only becomes more pronounced but also shifts to the right with L. We infer this to be the finite size precursor to the transition point, which suggests that in the thermodynamic limit, L → ∞, the derivative of the Schmidt gap diverges at the MBLT and ̃ hc asymptotically approaches the transition point hc.

Difficulties

The problem for tsdrg approach is that it's hard to access reduced density matrix in the center of the system.
One may consider to use bipartite fluctuations as a probe to the entanglement.

If it's possible to derive the Schmidt gap in terms of the cumulants of fluctuations.

Reference

http://arxiv.org/abs/1704.00738
https://journals.aps.org/prb/abstract/10.1103/PhysRevB.85.035409

[Bug] supported version of mkl-devel is in-consist on different device

Description

Latest few versions of mkl-devel released in 2022 are in-consist across devices.

On pip, 2022.0.3 seems to be the newest

pip install mkl-devel==2022.0.3

But specifying version 2022.0.0, 2022.0.1, 2022.0.2 or 2022.0.3 does not give consistent result (those after 2021.0.4).

Error Logs

Docker (Ubuntu 20.04)

ERROR: Could not find a version that satisfies the requirement mkl-devel==2022.0.3 (from versions: 2018.0.0, 2018.0.3, 2021.1.1, 2021.2.0, 2021.3.0, 2021.4.0, 2022.0.1, 2022.0.2)

Conda environment (conda 4.10.3) on mac (Monterey 12.3)

ERROR: Could not find a version that satisfies the requirement mkl-devel==2022.0.3 (from versions: 2018.0.0, 2018.0.3, 2021.1.1, 2021.2.0, 2021.3.0, 2021.4.0, 2022.0.0)

Refactor: rename ModelBase to Model1D

models may be annotated in type hinting, and it's therefore better to rename tnpy.model.ModelBase as Model1D, or something similar.

Feature: modified density matrix to overcome local minimum in DMRG

Local minimums are often encountered during the DMRG sweeping, especially for the single-site DMRG. To overcome this issue, it's suggested to bring in the modified density matrix, which considers the small perturbation from the rest of the environment.

References

Feature: modularity in tSDRG

Compute the modularity for the following 2 quantities in tSDRG

Entanglement entropy
Connected correlation function

Reference
—————-
https://journals.aps.org/prb/abstract/10.1103/PhysRevB.105.064202?fbclid=IwAR3wpy-h4oJsyUcjQA7y0UinfHo0JviS1uunMyl70F11FvNpa36xzE5UWw0

Dependency Dashboard

This issue lists Renovate updates and detected dependencies. Read the Dependency Dashboard docs to learn more.

Rate-Limited

These updates are currently rate-limited. Click on a checkbox below to force their creation now.

Open

These updates have all been created already. Click a checkbox below to force a retry/rebase of any.

Detected dependencies

dockerfile

Dockerfile

python 3.10.8

github-actions

.github/workflows/build.yml

docker/setup-qemu-action v2

docker/setup-buildx-action v2

docker/metadata-action v4

docker/login-action v2

docker/build-push-action v4

.github/workflows/codeql-analysis.yml

actions/checkout v3

github/codeql-action v2

github/codeql-action v2

github/codeql-action v2

.github/workflows/doc.yml

actions/checkout v3

actions/setup-python v4

peaceiris/actions-gh-pages v3

.github/workflows/lint.yml

actions/checkout v3

actions/cache v3

actions/setup-python v4

.github/workflows/release.yml

docker/setup-qemu-action v2

docker/setup-buildx-action v2

docker/metadata-action v4

docker/login-action v2

docker/build-push-action v4

actions/checkout v3

.github/workflows/test.yml

actions/checkout v3

actions/checkout v3

actions/cache v3

actions/cache v3

actions/cache v3

actions/setup-python v4

codecov/codecov-action v3

pep621

pyproject.toml

poetry-core >=1.0.0

poetry

pyproject.toml

python >=3.9, <3.11

toml ^0.10.2

tqdm ^4.63.0

numpy ^1.22

scipy >=1.1, <2.0

primme 3.2.1

h5py >=3.6, <4.0

graphviz 0.20.1

tensornetwork 0.4.6

numba ^0.56.4

llvmlite ^0.39.1

autoray ^0.6.0

quimb 1.4.0

mkl-devel 2023.0.0

matplotlib ^3.5.2

networkx ^3.0

pre-commit ^3.0.0

commitizen ^2.35.0

vulture ^2.4

bandit ^1.7.4

safety ^2.3.4

isort ^5.11.0

flake8-bugbear ^23.1.14

Flake8-pyproject ^1.2.2

black ^23.1.0

mypy ^1.0.0

ipython ^8.5.0

pytest ^7.1.2

pytest-cov ^4.0.0

pytest-mock ^3.9.0

pep8-naming ^0.13.0

cruft ^2.12.0

sphinx ^4.5.0

sphinx-book-theme ^0.3.2

nbsphinx ^0.8.8

m2r2 ^0.3.2

pandoc ^2.3

Check this box to trigger a request for Renovate to run again on this repository

bug: tensornetwork.FiniteMPS has no attribute nodes

After version 0.3.x, google/tensornetwork no longer supports the usage of FiniteMPS.nodes. One has to use the method FiniteMPS.get_tensor instead. See the documentation for more details.

Traceback (most recent call last):
  File "/__w/tnpy/tnpy/tests/test_finite_algorithm_base.py", line 20, in test_load
    init_method=f'{Path(__file__).parent}/test.npz'
  File "/__w/tnpy/tnpy/tnpy/finite_algorithm_base.py", line 30, in __init__
    self.reset_mps = init_method
  File "/__w/tnpy/tnpy/tnpy/finite_algorithm_base.py", line 76, in reset_mps
    self._init_envs()
  File "/__w/tnpy/tnpy/tnpy/finite_algorithm_base.py", line 142, in _init_envs
    self._update_left_env(site)
  File "/__w/tnpy/tnpy/tnpy/finite_algorithm_base.py", line 158, in _update_left_env
    M = self._mps.nodes[site-1]
AttributeError: 'list' object has no attribute 'nodes'

Refactor: add ed classmethod from_mpo

It should be better to use a matrix as input to the constructor. The current implementation can be refactor to a classmethod from_mpo.

Feature: bipartite fluctuations of the subsystem magnetization

Implement the bipartite fluctuations of the subsystem magnetization $S_A^z$,

$$ F = \langle (S_A^z)^2 \rangle - \langle S_A^z \rangle^2, $$

where $S_A^z$ is the total Sz number of the subsystem $A$.

tanlin2013 / tnpy Goto Github PK

tnpy's Introduction

tnpy: Tensor Network algorithms implemented in python.

Requirements

Installation

Getting started

License

tnpy's People

Contributors

Stargazers

Watchers

Forkers

tnpy's Issues

Blockers

TODO

Characterizing the MBLT

Difficulties

Reference

Description

Error Logs

References

Rate-Limited

Open

Detected dependencies

References

Recommend Projects

Recommend Topics

Recommend Org