Tensor-network experiments focused on PEPO/PEPS compression and time-evolution optimizations. Scripts implement cooling/optimization routines and utilities; notebooks capture exploratory runs and plots.
src/tcompress/: Python package for reusable code.algo_cooling.py: cooling and tensor-network routines used across experiments.p_pepo.py,p_pepo_su.py: PEPO optimization flows (standard and SU variants).Uoptimize.py,gate_arb.py: optimization helpers and arbitrary-geometry tensor utilities.quf.py,register_.py: shared utilities and registrations.
notebooks/pepo/: PEPO/PEPS-focused experiments and trotter reductions.notebooks/mps/: MPS simulations.notebooks/plotting/: plotting/visualization explorations.notebooks/analysis/: analysis/scratch runs.data/outputs/: cached artifacts and reference data (ignored).requirements.txt: Python dependencies for experiments.
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
# Option A: editable install
pip install -e .
# Option B: point Python to the package when running notebooks/scripts
# export PYTHONPATH=\"$PWD/src:$PYTHONPATH\"
# For GPU JAX/PyTorch, install vendor wheels as needed..gitattributesmarks notebooks as binary to avoid noisy diffs; usenbdiffor screenshots for review..gitignoreexcludes checkpoints, caches, anddata/outputs/. Keep transient data there or outside the repo.- Large generated data should stay out of version control or use Git LFS if required.