Solid-state SIESTA inputs and exchange–correlation network weights for "Constraint-aware functional cloning for stable and transferable machine-learned density functional theory" (v1.0)

SIESTA input files and trained neural-network exchange-correlation weights used to produce the equation-of-state results (lattice constants a0a_0 a0 and bulk moduli B0B_0 B0) for the 18 solid systems and 8 functional variants (PBE plus seven NN-XC) reported in arXiv:2605.10331. Includes: per-system input.fdf files at 8 lattice-constant volumes per equation-of-state sweep; a uniform TZP basis set (.ion files) shared across all functionals; the NN-XC weight files (4-layer fully-connected networks, 16 neurons per hidden layer); sanitized SLURM-free run_eos.sh drivers; and per-system Murnaghan-fit summaries (CSV, TXT, PNG) so that referees can verify the bulk-property extraction without rerunning SIESTA. See the included README.md for the column-to-directory mapping and the two .fdf flag conventions in use.