- Add config.sh with all pipeline parameters organized by category

  (molecular, crystal structure, compute, run mode, path)
- Refactor search_gen_proc.sh to source config.sh instead of
  hardcoding parameters, with optional config path argument
- Refactor structure_generate.py to load config.sh via exec(),
  replacing hardcoded values with config-driven parameters
- Remove mace-bench (the relaxation part, it will be replaced by updated seperate mace-bench project )

.gitignore

+# =============================================================================
+# Python
+# =============================================================================
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+pip-wheel-metadata/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# IPython / Jupyter Notebook
+.ipynb_checkpoints
+*.ipynb_checkpoints/
+profile_default/
+ipython_config.py
+
+# pyenv
+.python-version
+
+# pipenv
+Pipfile.lock
+
+# poetry
+poetry.lock
+
+# PEP 582
+__pypackages__/
+
+# Celery
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+.conda/
+conda-env/
+
+# Spyder project settings
+.spyderproject.db
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+
+# =============================================================================
+# PyTorch / Deep Learning
+# =============================================================================
+
+# Model checkpoints & weights
+*.pt
+*.pth
+*.ckpt
+*.bin
+*.safetensors
+checkpoints/
+saved_models/
+weights/
+pretrained/
+
+# TensorBoard logs
+runs/
+logs/
+tb_logs/
+tensorboard_logs/
+lightning_logs/
+
+# MLflow
+mlruns/
+mlflow/
+
+# Weights & Biases
+wandb/
+
+# Hydra outputs
+outputs/
+multirun/
+.hydra/
+
+# ONNX models
+*.onnx
+
+# TorchScript
+*.torchscript
+
+# CUDA profiling
+*.nvvp
+*.nvprof
+*.qdrep
+
+# Data directories (customize as needed)
+data/raw/
+data/processed/
+data/interim/
+datasets/
+
+# Large result files
+results/
+predictions/
+evaluations/
+
+# Experiment configs with secrets
+secrets.yaml
+secrets.yml
+.secrets
+
+
+# =============================================================================
+# IDE / Editor
+# =============================================================================
+
+# VSCode
+.vscode/
+*.code-workspace
+
+# PyCharm / JetBrains
+.idea/
+*.iml
+*.iws
+*.ipr
+
+# Vim
+*.swp
+*.swo
+*~
+
+# Emacs
+\#*\#
+.\#*
+
+# Sublime Text
+*.sublime-project
+*.sublime-workspace
+
+
+# =============================================================================
+# OS
+# =============================================================================
+
+# macOS
+.DS_Store
+.AppleDouble
+.LSOverride
+._*
+.Spotlight-V100
+.Trashes
+
+# Windows
+Thumbs.db
+ehthumbs.db
+Desktop.ini
+$RECYCLE.BIN/
+*.lnk
+
+# Linux
+*~
+
+
+# =============================================================================
+# Misc
+# =============================================================================
+
+# Compressed files
+*.zip
+*.tar.gz
+*.tar.bz2
+*.rar
+*.7z
+
+# Temporary files
+*.tmp
+*.temp
+*.bak
+*.orig
+
+# Secrets & credentials
+.env.local
+.env.*.local
+*.pem
+*.key
+config/secrets.*
+
+# =============================================================================
+# Project specific 
+# =============================================================================
+csp_results/
\ No newline at end of file

config.sh

+#!/bin/bash
+# =============================================================================
+# BOMLIP-CSP Configuration File
+# =============================================================================
+# This file contains all configurable parameters for the crystal structure
+# search and generation pipeline. Modify the values below to customize
+# your run.
+# =============================================================================
+
+# -----------------------------------------------------------------------------
+# [Molecular Parameters]
+# SMILES string of the input molecule(s).
+# Use '.' (dot) to separate multiple molecules for co-crystal generation.
+# Example single molecule:  "C1CC2=COC=C12"
+# Example co-crystal:       "C1CC2=COC=C12.CCO"
+SMILES="C1CC2=COC=C12"
+
+# Number of conformers to generate during the conformer search step.
+# Higher values explore more conformational space but take longer.
+# Set to 0 to skip generation and only load existing conformers.
+GENERATE_CONFORMERS=10
+
+# Number of conformers to actually use for crystal structure generation.
+# Must be <= the number of generated conformers.
+# Set to 0 to skip structure generation.
+USE_CONFORMERS=4
+
+# Number of molecules in the unit cell (Z').
+# Use comma-separated values for multiple molecule types (co-crystal),
+# e.g. "1,1" means 1 copy of each molecule in the asymmetric unit.
+# Use space-separated values for multiple packings, e.g. "1 2".
+MOLECULE_NUM_IN_CELL=1
+
+# -----------------------------------------------------------------------------
+# [Crystal Structure Parameters]
+# Space group numbers for structure generation.
+# Use comma-separated values within a packing, and space-separated
+# values for multiple packings.
+# Example: "14,61" means search space groups P21/c and Pbca in one packing.
+# Example: "14 61" means P21/c in packing 1, Pbca in packing 2.
+SPACE_GROUP_LIST="14,61"
+
+# Prefix name added to the output CIF files.
+# Use space-separated values for multiple packings.
+ADD_NAME="XULDUD"
+
+# Number of crystal structures to generate per (space group, conformer) combination.
+# Use space-separated values for multiple packings.
+NUM_GENERATION=100
+
+# -----------------------------------------------------------------------------
+# [Compute Parameters]
+# Maximum number of parallel workers for structure generation.
+# Should not exceed the number of available CPU cores.
+MAX_WORKERS=16
+
+# -----------------------------------------------------------------------------
+# [Run Mode]
+# Execution mode controlling which steps are performed.
+# Available options:
+#   all              - Run conformer search followed by structure generation (default)
+#   conformer_only   - Only perform conformer search
+#   structure_only   - Skip conformer search, use existing conformers to generate structures
+MODE="all"
+
+# -----------------------------------------------------------------------------
+# [Path Parameters]
+# Directory for storing intermediate conformers and output CIF structures.
+# Relative to the project root directory.
+OUTPUT_DIR="csp_results"

mace-bench/3rdparty/SevenNet/CHANGELOG.md

-# Changelog
-All notable changes to this project will be documented in this file.
-
-
-## [0.11.1]
-From here, the version of 'main' branch has 'devX' after it diverges from the latest stable version
-
-CLI interface changed in backward-compatible manner. Now `sevenn` has subcommands for
-inference, train, etc
-
-### Added
-- subcommand with some aliases
-- strict e3nn version requirement from __init__.py
-
-### Changed
-- pre-commit uses python3.11
-- cuequivaraiance optional libraries
-- some gitignores
-
-### Fixed
-- Circular import in sevenn.checkpoint (dev0)
-- Fix typing issues
-
-
-## [0.11.0]
-
-Multi-fidelity learning implemented & New pretrained-models
-
-### Added
-- Build multi-fidelity model, SevenNet-MF, based on given modality in the yaml
-- Modality support for sevenn_inference, sevenn_get_modal, and SevenNetCalculator
-- sevenn_cp tool for checkpoint summary, input generation, multi-modal routines
-- Modality append / assign using sevenn_cp
-- Loss weighting for energy, force and stress for corresponding data label
-- Ignore unlabelled data when calculating loss. (e.g. stress data for non-pbc structure)
-- Dict style dataset input for multi-modal and data-weight
-- (experimental) cuEquivariance support
-- Downloading large checkpoints from url (7net-MF-ompa, 7net-omat)
-- D3 wB97M param
-
-### Changed
-- Sort instructions of tensor product in convolution (+ fix flipped w3j coeff of old model)
-- Lazy initialization for `IrrepsLinear` and `SelfConnection*`
-- Checkpoint things using `sevenn/checkpoint.py`
-- e3nn >= 0.5.0, to ensure changed CG coeff later on
-- pandas as dependency
-- old v1 presets are removed, liquid electrolyte fine-tune yaml is added
-
-### Fixed
-- More refactor for shift scale things + few bug fixes
-- Correctly shuffle training set when distributed training is enabled
-- D3 calculator system swap memory error fixed
-- D3 compile uses $HOME/.cache if package directory is not writable
-
-
-## [0.10.4]
-### Added
-- feats: D3 calculator
-### Fixed
-- bug: info dict sharing (therefore energy stress) when structure_list used
-- torch >= 2.5.0 works
-- numpy >= 2.0 works (need more testing)
-### Changed
-- sevennet_calculator.py => calculator
-- fine tunine preset to use original loss function (Huber) and loss weights
-
-
-## [0.10.3]
-### Added
-- SevenNet-l3i5, checkpoint, preset. (keywords: 7net-l3i5, sevennet-l3i5)
-- SevenNet-l3i5 test
-### Changed
-- Now --help do not load unnecessary imports (fast!)
-- README
-
-
-## [0.10.2]
-### Added
-- Accelerated graph build routine if matscipy is installed  @hexagonerose
-- matscipy vs. ase neighborlist unit test
-- If valid set is not given but data_divide_ratio is given, validaset is created using random split. (shift, scale, and conv_denoiminator uses original whole statistics)
-### Changed
-- matscipy is included as dependency
-- data_divide_ration defaults to 0.0 (not used)
-### Fixed
-- For torch version >= 2.4.0, Loading graph dataset no more raises warnings.
-- Raise error when unknown element is found (SevenNetCalculator)
-
-## [0.10.1]
-### Added
-- experimental `SevenNetAtomsDataset` which is memory efficient, can be enabled with `dataset_type='atoms'`
-- Save meta data & statistics when the `SevenNetGraphDataset` saves its data.
-### Changed
-- Save checkpoint_0.pth (model before any training)
-- `SevenNetGraphDataset._file_to_graph_list` -> `SevenNetGraphDataset.file_to_graph_list`
-- Refactoring `SevenNetGraphDataset`, skips computing statistics if it is loaded, more detailed logging
-- Prefer use .get when accessing config dict
-### Fixed
-- Fix error when loading `SevenNetGraphDataset` with other types of data (ex: extxyz) in one dataset
-
-
-## [0.10.0]
-SevenNet now have CI workflows using pytest and its coverage is 78%!
-Substantial changes in cli apps and some outputs.
-
-### Added
-- [train_v2]: train_v2, with lots of refactoring + support `load_testset_path`. Original routine is accessible: `sevenn -m train_v1`.
-- [train_v2]: `SevenNetGraphDataset` replaces old `AtomGrpahDataset`, which extends `InMemoryDataset` of PyG.
-- [train_v2]: `sevenn_graph_build` for SevenNetGraphDataset. Previous .sevenn_data is accessible with --legacy option
-- [train_v2]: Any number of additional datasets will be evaluated and recorded if it is given as 'load_{NAME}set_path' key (input.yaml).
-- 'Univ' keyword for 'chemical_species'
-- energy_key, force_key, stress_key options for `sevenn_graph_build`, @thangckt
-- OpenMPI distributed training @thangckt
-### Changed
-- Read EFS of atoms from y_* keys of .info or .arrays dict, instead of caclculator results
-- Now `type_map` and requires_grad is hidden inside `AtomGraphSequential`, and don't need to care about it.
-- `log.sevenn` and `lc.csv` automatically find a safe filename (log0.sevenn, log1.sevenn, ...) to avoid overwriting.
-- [train_v2]: train_v2 loads its training set via `load_trainset_path`, rather than previous `load_dataset_path`.
-- [train_v2]: log.csv -> lc.csv, and columns have no units, (easier to postprocess with it) but still on `log.sevenn`.
-### Fixed
-- [e3gnn_serial]: can continue simulation even when atom tag becomes not consecutive (removing atom dynamically), @gasplant64
-- [e3gnn_parallel]: undefined behavior when there is no atoms to send/recv (for non pbc system)
-- [e3gnn_parallel]: incorrect force/stress in some edge cases (too small simulation cell & 2 process)
-- [e3gnn_parallel]: revert commit 14851ef, now e3gnn_parallel is sane.
-- [e3gnn_*]: += instead of = when saving virial stress and forces @gasplant64
-- Now Logger correctly closes a file.
-- ... and lots of small bugs I found during writing `pytest`.
-
-## [0.9.5]
-### Note
-This version is not stable, but I tag it as v0.9.5 before making further changes.
-LAMMPS `pair_e3gnn_parallel.*` should be re-compiled for the below changes regarding LAMMPS parallel.
-This is the first changelog and may not reflect all the changes.
-### Added
-- Stress compute for LAMMPS sevennet parallel
-- `sevenn_inference` now takes .extxyz input
-- `sevenn_inference` gives MAE error
-- Experimental `sevenn_inference` on the fly graph build option
-### Changed
-- **[Breaking]** Parallel LAMMPS model changed, old deployed parallel models will not work
-- **[Breaking]** Parallel LAMMPS takes the directory of potentials as input. Accordingly, `sevenn_get_model -p` creates a folder with potentials.
-- **[Breaking]** Except for serial LAMMPS models, force and stress are computed from gradients of edge vectors, not positions.
-- Separate interaction block from model build
-- Add typing for most of functions
-- Remove clang pre-commit hook as it breaks lammps pair files
-- `torch.load` with `weights_only=False`
-- Line length limit 80 -> 85
-- Refactor
-### Fixed
-- Correct batch size for SevenNet-0(11July2024)
-
-## [0.9.4] - 2024-08-26
-### Added
-- D3 correction (contributed from dambi3613) for LAMMPS serial