CarbideMatterGen: code, model and data for the ε-W2C/WOx CO2-reduction study

Code, fine-tuned model weights and data accompanying "Sacrificial rare-earth extraction closes the C1 channel on an epsilon-W2C/WOx catalyst that converts CO2 into ethanol and acetone" (Cell Reports Physical Science, 2026). CarbideMatterGen is MatterGen fine-tuned on 5,275 transition-metal carbides with nine conditioning properties and used to propose candidate CO2-reduction carbides. The archive contains the fine-tuning and generation code, the checkpoint used for every structure reported in the paper, the conditioning labels, the conditioned and unconditional generation pools, the machine-learning screening outputs, and the plane-wave DFT total energies of all computed systems together with a script that reproduces every free energy in Table 1 and Table S10 of the paper. Notes on the deposited files, which supersede the corresponding statements in README.md: 1. CarbideMatterGen_epoch415_loss0.13.ckpt is the weights-only checkpoint, 329,735,222 bytes, SHA-256 4c3fd2799b7dad06899fdf0392c1f765746c24826bcf3325be4be663bd340e72. The optimizer state has been removed, so the file restores the model for sampling but not a training run. The 488 weight tensors and the full configuration, including the nine conditioning properties, are those of the checkpoint used for every structure reported in the paper. The README describes this file as the complete 873 MB Lightning checkpoint; that description, and the SHA-256 given there, apply to the training checkpoint, which is available from the lead contact on request. 2. dft/verify_tables.py inside dft_energies_structures_inputs.zip labels its output with the supplemental table numbers of an earlier manuscript version. Table S8 in that output is Table S10 of the submitted paper, and Table S11 is Table S13. The energies and the verification itself are unaffected. 3. The copyright line of LICENSE names the project this code was branched from. The licence is the MIT License and applies to CarbideMatterGen.