User Guide

To use the library, simply include/import cytnx.

• In Python, using import:

import cytnx

• In C++, using the include header:

#include "cytnx.hpp";

Note

In C++, there is a namespace cytnx.

Aliases in Python modules and C++ namespaces can be used equivalently, for example if we want to alias cytnx as cy,

• In Python:

import cytnx as cy

This is equivalent in C++ to:

• In C++:

#include "cytnx.hpp";
namespace cy=cytnx;

Now we are ready to start using cytnx!

Continue reading:

• 1. Objects behavior
  • 1.1. Everything is reference
  • 1.2. clone
• 2. Device
  • 2.1. Number of threads
  • 2.2. Number of GPUs
  • 2.3. GPU status
  • 2.4. Initializing tensors on GPU
• 3. Tensor
  • 3.1. Creating a Tensor
    • 3.1.1. Initialized Tensor
    • 3.1.2. Random Tensor
    • 3.1.3. Tensor with different dtype and device
    • 3.1.4. Type conversion
    • 3.1.5. Transfer between devices
    • 3.1.6. Tensor from Storage [v0.6.6+]
  • 3.2. Manipulating Tensors
    • 3.2.1. reshape
    • 3.2.2. permute
  • 3.3. Accessing elements
    • 3.3.1. Get elements
    • 3.3.2. Set elements
    • 3.3.3. Low-level API (C++ only)
  • 3.4. Tensor arithmetic
    • 3.4.1. Type promotion
    • 3.4.2. Tensor-scalar arithmetic
    • 3.4.3. Tensor-Tensor arithmetic
    • 3.4.4. Equivalent APIs (C++ only)
  • 3.5. To/From numpy.array
  • 3.6. Appending elements
  • 3.7. Save/Load a Tensor
    • 3.7.1. Save a Tensor
    • 3.7.2. Load a Tensor
  • 3.8. When will data be copied?
    • 3.8.1. Reference to & Copy of objects
    • 3.8.2. Permute
    • 3.8.3. Contiguous
    • 3.8.4. Reshape
• 4. Storage
  • 4.1. Creating a Storage
    • 4.1.1. Type conversion
    • 4.1.2. Transfer between devices
    • 4.1.3. Get Storage of Tensor
  • 4.2. Accessing elements
    • 4.2.1. Get/Set elements
    • 4.2.2. Get raw-pointer (C++ only)
  • 4.3. Increase size
    • 4.3.1. append
    • 4.3.2. resize
  • 4.4. From/To C++.vector
  • 4.5. Save/Load a storage
    • 4.5.1. Save a Storage
    • 4.5.2. Load a Storage
    • 4.5.3. Save & load from/to binary
• 5. Scalar
  • 5.1. Define/Declare a Scalar
  • 5.2. Change data type
  • 5.3. Application scenarios
• 6. Tensor notation
  • 6.1. Tensor contraction
  • 6.2. Direction of bonds
• 7. UniTensor
  • 7.1. Print and display
    • 7.1.1. print_diagram()
    • 7.1.2. print() and print_blocks()
    • 7.1.3. print_block()
  • 7.2. Creating a UniTensor
    • 7.2.1. Constructing from Tensor
    • 7.2.2. From scratch
    • 7.2.3. Type conversion
    • 7.2.4. Transfer between devices
    • 7.2.5. Tagged UniTensors and UniTensors with Symmetries
  • 7.3. Changing labels
    • UniTensor.relabel_()
    • UniTensor.relabel_()
    • UniTensor.relabels_()
    • UniTensor.relabels_()
    • 7.3.1. Creating UniTensors with different labels that share data
  • 7.4. Bond
    • Bond()
    • 7.4.1. Symmetry object
    • 7.4.2. Creating Bonds with quantum numbers
    • 7.4.3. Combining Bonds
  • 7.5. Tagged UniTensor
    • 7.5.1. Non-symmetric Tagged UniTensor
  • 7.6. UniTensor with Symmetries
  • 7.7. Accessing the block(s)
    • 7.7.1. UniTensor without symmetries
    • 7.7.2. UniTensor with symmetries
    • 7.7.3. Putting a block
  • 7.8. Get/set UniTensor element
    • 7.8.1. UniTensor without symmetries
    • 7.8.2. UniTensor with symmetries
  • 7.9. Manipulate UniTensor
    • 7.9.1. Permute
    • 7.9.2. Reshape
    • 7.9.3. Combine bonds
    • 7.9.4. Arithmetic
    • 7.9.5. Rowrank
    • 7.9.6. Transpose
  • 7.10. Save/Load a UniTensor
    • 7.10.1. Save a UniTensor
    • 7.10.2. Load a UniTensor
• 8. Contraction
  • 8.1. Network
    • 8.1.1. Network from .net file
    • 8.1.2. Launch
    • 8.1.3. Network from string
    • 8.1.4. Put UniTensors
    • 8.1.5. PutUniTensor according to label ordering
    • 8.1.6. Set the contraction order
  • 8.2. Contract(s)
    • 8.2.1. Contract
    • 8.2.2. Contracts
  • 8.3. ncon
• 9. Linear algebra
• 10. Iterative solver
  • 10.1. LinOp class
    • 10.1.1. Inherit the LinOp class
    • 10.1.2. Example: sparse data structure with mapping function
    • 10.1.3. Prestore/preconstruct sparse elements
  • 10.2. Lanczos solver
    • Lanczos_ER()
• 11. linalg extension
  • 11.1. Tensor decomposition
    • 11.1.1. Singular value decomposition
    • 11.1.2. Eigenvalue decomposition
    • 11.1.3. QR decomposition
• 12. Common APIs
  • 12.1. Data type
    • .dtype()
    • .dtype()
    • .dtype_str()
    • .astype()
    • .real()
    • .imag()
  • 12.2. Device
    • .device()
    • .device_str()
    • .to()
    • .to_()
  • 12.3. Save & Load
    • .Save()
    • .Save()
    • .Load()
    • .Tofile()
    • .Tofile()
    • .Fromfile()
  • 12.4. Operations
    • .clone()