Suggestion: Add Support for Per-Chromosome Arm Imputation to Reduce Memory Usage #157
Description
While adapting Battenberg for use with Oxford Nanopore (ONT) data, I encountered memory issues when running the imputation step under SLURM. Unlike SGE, where Battenberg ran successfully on 8GB nodes, the same configuration under SLURM consistently resulted in out-of-memory errors.
To resolve this, I modified the package to split chromosomes by arms during the imputation step and forced garbage collection. This change significantly reduced the memory footprint and allowed the pipeline to run successfully on the same hardware. I thought this modification might be useful to you. I have attached the patch. All the best, Marian