Examples¶

PyTC 0.2.1 ships six numbered examples in the pytc/examples directory. They are small, executable demonstrations rather than production benchmarks.

Get and run the examples¶

Clone the matching release and install it in editable mode:

git clone --branch v0.2.1 https://github.com/nickirk/pytc.git
cd pytc
python -m pip install -e .

Run an example from the repository root, for example:

python pytc/examples/03_dense_xtc_ccsd.py
python pytc/examples/06_rank_m_x_factor_direct_ccsd.py

Each script is runnable on its own unless the table notes a dependency. The scripts print a small expected-value or structural self-check and fail if that check does not pass.

Numbered walkthrough¶

Script	What it demonstrates	Dependency
`01_vmc_optimize_jastrow.py`	Two-phase reference-variance VMC optimization of a Jastrow factor	Writes `h2o_phase_b_hist.h5`
`02_load_and_average_jastrow_params.py`	Polyak–Ruppert averaging of the phase-B trajectory	Reads the file written by 01
`03_dense_xtc_ccsd.py`	Exact dense, non-ISDF xTC-CCSD	Standalone
`04_isdf_xtc_ccsd.py`	ISDF xTC-CCSD compared with the dense reference	Standalone
`05_make_fno_xtc_ccsd.py`	FNO virtual-space truncation scan through ISDF	Standalone
`06_rank_m_x_factor_direct_ccsd.py`	Rank-\(M\) orbital X and direct T2-U-Z factorized ISDF xTC-CCSD	Standalone; PyTC 0.2.1+

Two separate workflows¶

Examples 01 and 02 show how to optimize and average a flexible CompositeJastrow([NuclearCusp, BoysHandy]). Examples 03–05 instead use a fixed, inexpensive REXP correlator so that the dense reference remains practical on a small machine. They do not consume the optimized parameters from 01/02.

Example 06 is a separate H4/STO-3G API demonstration. It confirms that the rank-\(M\) build contains X_tucker = {U, Z} instead of dense X and then runs the dedicated factor-direct solver. See ISDF Efficiency Paths before using that approximation for a new production system.

Choosing a starting point¶

Start with 03 when validating integral conventions on a small molecule.
Compare 03 and 04 before relying on an ISDF rank for a new system.
Use 05 to study the additional FNO truncation error.
Use 06 only after the full-X ISDF calculation is understood and validated; rank-\(M\) X is an explicit approximation, not the default.