Skip to content

Examples

This section contains end-to-end LaQX workflows. Each page explains the physics target and the exact python main.py ... commands used by the corresponding scripts under docs/examples/.

The shell scripts are still kept in the repository for reproducibility, but each example page also expands the script into explicit commands so that every model parameter, optimizer choice, boundary condition, and ansatz setting is visible.

Example map

Page Physics task Hamiltonian / objective Main ansatz
Hubbard Model Ground-state VMC Hubbard model tensor, ace, scale
Spin Model Frustrated quantum magnet ground state \(J_1\)-\(J_2\) Heisenberg model cnn_mps
Altermagnetic Hubbard Model Ground-state VMC and momentum-distribution measurement Altermagnetic Hubbard model tensor, ace
Hofstadter Model Many-body Chern-number workflow by flux threading Interacting Hofstadter model with polar observable ace
Excited States (NES) Simultaneous low-lying-state optimization Multi-state Hofstadter VMC ace_nes
Time Evolution (tVMC) Time-dependent variational principle evolution Projected dynamics on the variational manifold ace, ace_peft

How to read the commands

Each command has the same structure:

python main.py \
    --mode <march|spring|adam|test|gfmc|tvmc> \
    --model <hamiltonian-name> \
    --network_name <ansatz-name> \
    --output <checkpoint-and-log-directory> \
    ...

Important groups of arguments:

  • --model, --L1, --L2, --particles, and boundary flags define the Hilbert space and Hamiltonian geometry.
  • Couplings such as --U, --V, --t2, --alpha, --j1, and --j2 define the physics problem.
  • --network_name, --hidden, --layers, and related flags define the neural ansatz.
  • --mode adam, --mode march, --mode spring, --mode test, --mode gfmc, and --mode tvmc choose pretraining, VMC optimization, observable measurement, GFMC projection, or time evolution.
  • --restore reuses a previous checkpoint as initialization for a later stage.