vasp-batch-execution

To safely batch submit VASP jobs, stage work under a fresh output tree separate from your input calculation folders. This approach collects results into a final output snapshot, preventing any mutation or corruption of the original input directory structure.

Failure triage for batch VASP calculations relies on reading the _BATCH_STATE.json file alongside focused scheduler, stdout, and stderr evidence. This prioritization preserves failure evidence and enables the selective rerun of only the failed subsets.

Valid VASP calc folders for recursive job dispatch are identified by the presence of INCAR and POTCAR files. Layout validation distinguishes between single-folder and recursive batch modes while rejecting nested or overlapping calculation roots.

Yes, you can rerun failed subsets from a VASP campaign by using vasp_execute_batch to dispatch the failed jobs into a fresh output root. The system reads _BATCH_STATE.json to identify failures and selectively reruns them without touching successful calculations.

The output_dir must be separate from the input_dir to prevent output roots from existing inside input roots. This layout rule ensures the bootstrap dispatcher stages work safely, avoiding nested or overlapping calculation trees that corrupt input data.